Battlefield dynamic scanning and staring imaging system based on fast steering mirror
CHANG Tianqing1,WANG Quandong1,2,*,ZHANG Lei1,HAO Na3,and DAI Wenjun1
1.Department of Weapon and Control,Army Academy of Armored Forces,Beijing 100072,China;
2.Center for Assessment and Demonstration Research,Academy of Military Sciences,Beijing 100091,China;
3.Department of Science Research,Army Academy of Armored Forces,Beijing 100072,China
Abstract: This paper presents the design of an experimental battlefield dynamic scanning and staring imaging system based on a fast steering mirror(FSM),which is capable of real-time monitoring of hot targets and wide-area reconnaissance of hot regions.First,the working principle and working sequence of the FSM are briefly analyzed.The mathematical model of the FSM system is built by modeling its dynamic and electrical properties,and the rationality of the model is validated by means of model identification.Second,the influence of external sources of disturbance such as the carrier and moment on the control precision of the FSM is effectively suppressed by the jointly controlling of proportional integral(PI)and disturbance observer(DOB),thus realizing a high precision and strong robustness control of the FSM system.Then,this paper designs an experimental prototype and introduces a special optical structure to enable the infrared camera to share the FSM with the visible light camera.Finally,the influence of the velocity difference between the mirror of the FSM and the rotating platform on the imaging quality of the system is experimentally analyzed by using the image sharpness evaluation method based on point sharpness.A good dynamic scanning and staring imaging result is achieved when the velocity of these two components correspond.
Keywords: fast steering mirror(FSM);dynamic scanning and staring; backscanning compensation; disturbance observer(DOB);point sharpness.
1.Introduction
The battlefield environment of information warfare based on a modern command and control system is complex and changeable,with frequent scheduling of personnel and weaponry.Information about the enemy’s battlefield resource allocation and force deployment is not only the core information the reconnaissance and situational awareness equipment is required to obtain,but also the key factor to consider when making tactical decisions and issuing commands.Recent years have seen a significant increase in the maneuverability of new armored vehicles,a variety of unmanned vehicles,and other land warfare weapon platforms.This has made it necessary to accurately obtain real-time images of targets and their additional orientation,location,and other battlefield intelligence information in a wide field of view(FOV)to enable an evaluation of the enemy’s combat situation,an assessment of the target threat,and our firepower plan in the future cooperative combat of armored units.This requires the ground reconnaissance equipment to be capable of real-time monitoring and tracking of hot targets,and the capability to provide wide-area reconnaissance of hot regions would also be necessary,such that high resolution and large FOV battlefield images of the target area can be obtained dynamically,stably,and in real time in the wide-area reconnaissance mode of the equipment.
At present,there are mainly three kinds of imaging methods for high resolution and large FOV imaging systems,namely staring imaging[1–3],scanning imaging[4–7]and step-stare imaging[8,9].Among these methods,the staring imaging method is based on a large-sized focal plane detector[1],which can realize a wide FOV and high-resolution imaging,but its large number of pixels result in a longer charge read cycle and a low frame rate.Therefore,this method cannot be used to carry out realtime monitoring and tracking of moving targets,and image distortion readily occurs in the wide FOV.The scanning imaging method is based on a linear-array detector[4]and expands its FOV by means of scanning or sweeping the detector by a motor,but the system frame rate remains limited.Since there is a delay of the exposure time between different lines of the image,it is mainly applied to wide-area imaging of static areas,and it still cannot realize real-time monitoring and tracking of moving targets.The step-stare imaging method uses a small-scale array detector[8,9],and the step and stare of the boresight is achieved by a motor-controlled rotating platform,which can remain stationary during exposure time and allows the detector to record an image of the current scene in a static state,before the platform moves to the next scene and repeats the same process until the entire area is scanned.This method can use a small-scale array detector to achieve wide-area reconnaissance and target monitoring at the same time.However,because the inertia of the rotating platform is too large,the start and stop operations of the platform consume considerable time.Thus,the traditional step-stare imaging method does not take full advantage of the high frame rate of the small-scale array detector,thereby resulting in a low scanning efficiency.
中段音乐之后是对前面音乐的全部反复，即又重复了A和B段的音乐。奏完B段之后，音乐直接进入“结尾”（Coda）。结尾的八小节音乐，完全重复引子的音乐，在弱音量中结束全曲，结束在温馨甜美的意境中，余音绕梁。这是真正的“首尾呼应”。
Continuous rotation of the rotating platform for the purpose of scanning the area would avoid the problem of excessive stable time between frames and this would improve the frame frequency and scanning efficiency of the system.However,the relative motion between the scene and the detector during exposure time can produce image motion,thus causing images to appear blurred and smeared.There are two solutions to the image motion problem,one of which is the software compensation method that uses an image degradation algorithm and other image processing algorithms to deblur the image and compensate for the image motion.Although this method is less costly and less flexible,the real-time performance,adaptability,and compensation performance of the algorithm are relatively limited.The other is a physical compensation method in which the scene and the detector remain static relative to each other during the exposure time by controlling the motor,which rotates or moves the optical elements in the optical path to finally compensate for image motion by achieving a“dynamic balance”.However,when the size and weight of the camera are large,this method is difficult to be realized because of the large inertia moment.In order to reduce the moment of inertia,the usual practice is to add a fast steering mirror(FSM)[10,11]before the optical camera system,which compensates for the image motion by rotating the mirror.Therefore,this step-stare imaging method is also known as dynamic scanning and staring imaging.
Dynamic scanning and staring imaging technology achieves the boresight of the detector“stepping”and“staring”at a fixed cycle by enabling the FSM to“step”and“backscan”in the same cycle.Combined with the technology of image mosaic,it is possible to simultaneously realize high-resolution imaging at a wide FOV and high frame rate staring in an imaging system.Therefore,it has become one of the key technologies of new battlefield surveillance equipment.By using the periodic backscanning motion of the FSM to compensate for the image motion caused by the continuous uniform scanning of the rotating platform,it becomes possible not only to shorten the settling time between the frames,but to maximize the advantage of the small-scale array detector and to effectively improve the frame rate and scan efficiency of the imaging system.The extent to which the angular velocity of the rotating platform matches that of the FSM considerably affects the image quality of the system.Therefore,the high precision and strong robustness control of the FSM system are key factors to determine the performance of the dynamic scanning and staring imaging system.Optimization of the FSM system also presents the main difficulties in the design of the system.
2.Working principle and working sequence analysis of FSM
2.1 Structure of FSM
FSM has been widely used in laser communication,precision tracking,boresight stabilization and other optical systems for the optical device to control the direction of the light beams between the target and the detector[12–15].There are usually two kinds of driving devices for the FSM:a piezoelectric ceramic motor[16–18]and voice coil motor[19–21].Compared with the voice coil motor,the piezoelectric ceramic motor has a higher resolution and frequency response,but its stroke and load capacity are limited,so that it is often used in FSM systems with small aperture and high bandwidth.In this work,the aperture and stroke of the mirror are relatively larger,so the voice coil motor is chosen as the driving device of the FSM.Since the mirror of the FSM system only needs to perform backscan compensation in the horizontal direction,we design a two-drive and single-shaft structure FSM,as shown in Fig.1(a).
The structure is mainly composed of a mirror,a mirror holder,a base,a spherical hinge,and a pair of voice coil motors and eddy current sensors that are used for obtaining the deflection angle of the mirror in real time.Among these components,the mirror is rigidly fixed to the mirror holder,the two identical motors are symmetrically fixed to the base,and a flexible support is constituted between the mirror holder and motors through a spherical hinge.The movement velocity and direction of the mirror can be quickly controlled by controlling the magnitude and direction of the voltage applied to the motors.The actual three-dimensional structure and specific parameters of the FSM used in this study are shown in Fig.1(b)and listed in Table 1.
Fig.1 Structure of FSM
Table 1 Parameters of FSM
2.2 Working principle of backscanning compensation
In order to obtain clear imaging,it is necessary to ensure that the detector and the scene remain stationary or relatively stationary during the exposure time.As shown in Fig.2,in order to maintain a relatively static environment between the scene and the detector,to enable stare imaging in the state of motion,the rotation velocity of the detector moving along with the rotating platform must be compensated for by controlling the FSM.Within the length of time required for a full exposure from t0 to tτ,the boresight of the detector is rotated from OD to OD′,and the position of the mirror of the FSM system is rotated from AOB to A′OB′.ΔθD,Δθm are the rotation angle of the detector and mirror,respectively.
5.1 走品牌兴薯之路 临沭县委县政府将地瓜产业作为富民兴沭的着力点，制定印发了《关于加快现代农业园区创建 深入推进优质农产品基地品牌建设的意见》，重点扶持地瓜产业发展。采取“稳规模、增储存、延链条、育品牌”等有效措施，最大程度地维护临沭地瓜的品牌质量，进一步提高临沭地瓜在全国乃至国际市场上的品牌知名度。
From Fig.2 we can determine that
Fig.2 Working principle of backscanning compensation
Substituting(1)–(3),we obtain
穿袜子时，易非发现脚上起了几个拇指大小的燎泡，原来是昨晚太冷，她灌了热水袋，当时脚太冰了，搁在脚上不觉得，后来迷迷糊糊睡着了，竟慢慢烫起了泡。
样品测试在武汉综合岩矿测试中心进行，采用的测试方法为等离子体发射光谱法、X射线荧光光谱法、原子荧光光谱法、发射光谱法、石墨炉原子吸收法、离子选择性电极法、催化极谱法、化学发射光谱法等。样品测试质量采用了标准样、密码样、监控样等多种监控手段，并通过了中国地质调查局区域地球化学分析质量监督检查组的验收，分析质量为优秀级。
In summary,when the detector rotates along with the rotating platform at an angular velocity of ωD,the FSM control system causes the mirror to rotate at an opposite angular velocity of ωm.By ensuring that the angular velocity of the detector corresponds to that of the mirror,the staring imaging of the boresight can be realized in the inertial space.The position of the detector changes constantly with the rotation of the rotating platform during the“staring”operation,but the angle of the boresight is stable,thus we can achieve continuous and stable imaging of a wide FOV.
2.3 Continuous imaging process analyses
The continuous imaging process of the system is shown in Fig.3,where θD,θS,and θm are the horizontal angular positions of the detector,boresight,and mirror,respectively.When the detector is at position M and the boresight is at a,the boresight position of the FSM is at the center of the FOV of#a.At this time,the angular velocity of the mirror ωm is required to match the velocity of the detector ωD,so that stare imaging of the boresight can be realized in the inertial space.Meanwhile,it is necessary to ensure that the staring time or backscanning time is longer than the integration time of the detector to achieve a clear imaging of the field of view#a.Once the detector has completed the integral imaging,the mirror of the FSM system must be quickly rotated back to the starting position of the backscanning operation.When the system judges that the position of the detector is at M+1,the boresight is at position a+1,where the two images just meet the designed overlap angle that is convenient for later image mosaic processing.The mirror undergoes backscanning,and the boresight again performs staring in the inertial space,to complete the clear imaging of field#a+1.Subsequently,it repeats the above process of view#a+2,...,and#a+n until the entire regional area is scanned or panoramic imaging is completed.The rotating platform and the mirror of the FSM system work together and cooperate accurately to enable the detector to efficiently scan and perform staring imaging of the reconnaissance area.
Fig.3 Continuous imaging process of the system
3.Modeling and control system design
The closed-loop control structure of the FSM system,which is shown in Fig.4,mainly consists of a digital signal processing(DSP)position controller,a motor driver,a motor,a mirror,and a position sensor.The angle error Δθ between the input angle θr and the output angle θm is inputted into the DSP position controller to obtain the position control signal Pθ of the mirror,after which Pθ is converted into a voltage signal Uθ that powers the motor to rotate the mirror by the motor driver,and through the real-time feedback of the mirror position by the sensor,the closed-loop control of the position of the mirror can finally be realized.
Fig.4 Closed-loop control structure of FSM system
In order to realize accurate control of the FSM system,it is necessary to model and analyze the system precisely.
预测是指利用已经获得的书面信息，如文章的标题、生词表中的单词和短语、练习中的问题和选项等,结合自身掌握的语言知识、背景知识、生活常识和逻辑推理能力,对听力材料的内容以及问题的答案做出较为合理的猜测。预测或推测技巧的使用贯穿整个听的过程，即听前预测，听中预测和听后推测。
3.1 Modeling of FSM system
As shown in Fig.4,the FSM is mainly composed of a motor driver,a motor,a mirror,and a position sensor.Among these components,the motor driver adopts a linear power amplifier with a high control accuracy and response velocity,and can be modeled as a proportional component.The position sensor uses an eddy current sensor with wide bandwidth and high linearity,and because there is a fixed proportional relationship between the angular position of the mirror and the feedback voltage of the sensor,it can also be modeled as a proportional component.Therefore,building a model of the FSM requires the motor and the mirror as well as the mechanical connection between them to be modeled.
在伯努力大数定律和辛钦大数定律的理论支持下,设计蒙特卡洛实验,来计算一些复杂积分的近似值,例如下文记为积分Ⅰ。
3.1.1 Dynamic model
The equivalent dynamic model of the voice coil motor and the mirror is shown in Fig.5,which also shows a representation of the mechanical connection between them.In Fig.5,M is the moment applied by the motor in the x-axis direction,J is the rotational inertia of the mirror and flexible support,θm is the deflection angle of the mirror,and x is the axial displacement of the motor.Since θm is relatively small,x ≈lθm,Kθ is the torsional stiffness of the flexible support in the x-axis direction.cA and cB are the equivalent damping coefficients of the flexible support and motor,mcA and mcB are the mover mass of the motors,and lA and lB are the distance between the force acting point of the motor and the rotating shaft of the mirror.Because motor mA and mB are symmetrically arranged,we can assume that cA=cB=c,lA=lB =l and mcA=mcB=mc to simplify the model.
Fig.5 x-axis dynamic model of FSM
The moment balance equation of the single-axis FSM model can be expressed as
As shown in Fig.6(a),the energized coil windings fixed to the coil holder of the voice coil motor will be forced to move by the permanent magnetic field,and the magnitude of the force is proportional to the magnitude of the current flowing through the coil.Assuming the moment coefficient of the motor is Km,the moment of the motor is Tm=Kmim.Since the moment applied in the direction of the x-axis comes from a pair of symmetrically distributed motors,we obtain
In order to simplify the model,we set Jτ=J+2mel2 and Fm=2cl2.Then the moment balance equation of(6)can be simplified as follows:
3.1.2 Electrical model
警员说，遭遇这样的不幸，处置起来比较棘手，我们理解甚至同情，我们只能依法办事，实事求是。希望你们配合。
The electrical structure of the voice coil motor and its equivalent circuit model(including the load)are shown in Fig.6.In the figure Uθ,im,Rm,and Lm represent the voltage,current,resistor,and inductor of the motor,respectively,and vm is the movement velocity of the coil windings that produces a reverse electromotive force(EMF)that is proportional to the cutting velocity when the coil windings are cutting the magnetic induction line.Assuming that the reverse EMF coefficient of the motor is ke,the reverse EMF of the motor is Em=kevm.
Fig.6 Electrical structure and equivalent circuit of voice coil motor
In this study,the dynamic scanning and staring imaging system based on FSM is mainly applied in the vehicle carrier environment of the battlefield.Contrary to the laboratory environment,the external environment is complex and volatile,and the system is easily affected by various obstacles and disturbances.In order to meet the special requirements of applications intended for use in external environments,the control system of the FSM also requires a strong anti-disturbance ability under the premise of ensuring robustness and bandwidth control.The closed-loop control structure of the FSM system with the PI controller in consideration of disturbance and measurement noise is shown in Fig.13.
Let Ke=kel,then the voltage balance equation of the motor can be expressed as
Subjecting the moment and voltage balance equations that are shown in(8)and(10)to a Laplace transform gives
The transfer function of the output angle position to the input voltage of the voice coil motor is
Assume the proportional coefficient of the motor driver is Kd.As shown in Fig.4,with an input signal Pθ,the output Uθ(s)of the motor driver can be expressed as Uθ(s)=KdPθ(s).Under the premise of ignoring the DSP controller,the open-loop transfer function of the system is
From(13)we can see that the FSM system can be modeled as a third-order system with a constant molecule term.
3.2 Model identification of FSM system
As shown in Fig.7,a dynamic signal analyzer is used to identify the model of the FSM system.This analyzer directly inputs its generated excitation signal V1 to the motor driver of the FSM and its own input channel in1 directly.At the same time,the analyzer collects the response signal V2 of the FSM system through its input channel in2.Then the analyzer acquires the amplitude and frequency response data of the FSM system by processing V1 and V2.Finally,the amplitude-frequencycurve measurement result of the FSM system is achieved and shown in Fig.8.
Fig.7 Amplitude and frequency response measurement of FSM system by a dynamic signal analyzer
Fig.8 Amplitude-frequency curve measurement results of FSM system and its fitting curve
It can be seen from Fig.8 that the amplitude-frequency curve of the FSM system has obvious resonances.The main reason for the lower order resonance is that there is a resilient connection between the two rigid connections of the mirror;that is to say,there is a flexible support structure.The lower order resonance ω1 is the first order resonance caused by the flexible support,and there are also higher order resonances due to other mechanical coupling in the high frequency band.The third-order system in(13)is used to fit the amplitude-frequency curve of the FSM system,and the fitting results are shown in Fig.8 and(14).It can be seen from Fig.8 that the fitting result is very close to the measured value(except for the high-frequency resonant part,since the actual system has high order resonances due to coupling,which is not considered during the modeling process),which shows that the previous model is reasonable and it can reflect the true characteristics of the system.Therefore,when the higher order resonances are neglected,the open-loop transfer function of the FSM system is considered as
3.3 Control algorithm design of FSM system
3.3.1 Proportional integral(PI)control
In Fig.20(b),the visible light camera is not shown in the actual structure because it is shielded by the spectroscope.
Fig.9 Zero-pole distribution of closed-loop system
The third-order system shown in(14)can be expressed as the product of an inertial system and a second-order system:
where T=4.016×10-4,K=3.12,the natural or undamped resonance frequency ωn=247.6 rad/s(39.6 Hz),and the damping coefficient ξ=0.078.Because the parameter T of the inertia system is much smaller than 1,the low-frequency characteristic of the FSM system can be approximated by a second-order system,and the resonant frequency of this system is
This coincides with the frequency of the lower order resonance ω1 as shown in Fig.8.Therefore,the bandwidth of the FSM system can be approximated by the bandwidth of the second-order system:
It can be seen from(17)that ωb is proportional to ωn and inversely proportional to ξ.Here,ξ=0.078,which indicates that the second-order system is an underdamping oscillation resonant system,and the resonance of the FSM system at ω1 is mainly caused by this second-order oscillation system.In order to suppress or eliminate the influence of the resonant peak,it is possible to add a velocity feedback loop in the inner loop of the FSM system to increase ξ of the second-order system.At the same time,an increase in ξ can also cause the conjugate poles position of the system to move gradually to the left half of the S plane to finally stabilize the system.After the velocity feedback loop is added,the closed-loop control structure of the FSM system is shown in Fig.10,and the new open-loop transfer function of the FSM system is
Fig.10 Closed-loop control structure of FSM system with velocity feedback loop
When b is equal to 0,0.000 1,0.001,0.01,and 0.1,respectively,the Bode diagram of the open-loop system shown in Fig.11 is obtained.As the value of b increases,the lower order resonance gradually disappears,and the amplitude margin and phase margin of the system increase gradually from negative to positive(h1 ＜ h2 ＜ 0 ＜h3＜h4＜h5,γ1＜γ2 ＜0 ＜γ3＜γ4＜γ5).However,the cutoff frequency of the system decreases gradually(ωc1＞ωc2＞ωc3＞ωc4＞ωc5),and the bandwidth frequency and control bandwidth are also reduced accordingly,which is consistent with(17).The control bandwidth is a key parameter for the FSM system,and a larger bandwidth means improved dynamic performance.Therefore,our comprehensive consideration including the amplitude margin,phase margin,bandwidth,and stability of the system finally leads us to choosing the parameter b=0.02,which not only effectively eliminates the lower order resonance,but also to a great extent retains the original bandwidth of the system.After adding the velocity feedback loop,the control bandwidth of the system is reduced from 119.8 Hz(752.5 rad/s)to 118.2 Hz(742.0 rad/s).Generally speaking,the control bandwidth is limited by the mechanical structure of the FSM,and the larger the stroke and the aperture of the mirror are,the smaller the control bandwidth is.In this work,the stroke and aperture of the mirror are large.Hence,the control bandwidth of the FSM system is limited.The new open-loop transfer function of the system is
2 结果示特发性鼻出血患者共210例，其中A型病例21例，B型病例82例，O型病例97例，AB型病例10例，患者血型分布结果见表2。对照组血型统计结果显示血型比例为：O型（34%）＞A型（31%）＞B型（27%）＞AB型（8%），见表3。将特发性鼻出血患者血型分布与对照组血型分布比例进行比较有显著差异后（χ2=39.38，P＜0.01），各血型再分别进行比较，具体结果见表4。特发性鼻出血患者B型和O型血病例构成比高于损伤类患者（P＜0.01），A型血构成比低于对照组（P＜0.01），AB型血构成比与对照组构成比差异无统计学意义（P＞0.1）。
Since the base of the FSM and the detector are fixed to the rotating platform,the rotation angle of the base is the same as that of the detector.It can be seen from(4)that the rotation angle of the detector is twice that of the mirror at time tτ.Therefore,it is necessary to take control of the mirror backscanning operation,which rotates the mirror in the opposite direction to the rotating platform to compensate for the angular velocity of the rotating platform.Assuming the rotation direction of the detector is positive,the relation of the angular velocity between the mirror and the detector or rotating platform is
Fig. 11 Bode diagram of the open-loop system with different values of b
The zero-pole distribution of the system after adding the velocity feedback loop(b=0.002)is shown in Fig.9,and all three poles are located in the left half of the S plane,indicating that the system is stable.At this point,the step response curve of the system is shown in Fig.12.Although the system is stable,there is an error in the output value of the steady state and the input value,which means there is an error in the feedback angle and the input angle.Using the PI controller to further adjust the system performance,and with the help of the MATLAB single input single output(SISO)tools,the final designed PI controller parameters are as follows:
与意大利许多其他产区一样，威尼托产区有着相当悠久的酿酒历史。早在千年前，威尼托的葡萄种植及葡萄酒酿造就已经开始。到了中世纪，随着中心城市威尼斯商业的蓬勃发展，这里主导了意大利的葡萄酒贸易。
The step response curve of the system is shown in Fig.12 with and without the PI controller.After the controller is added,the output of the system is consistent with the input,the settling time ts is approximately 30 ms and the overshoot is significantly reduced(σ1=30.3%,σ2=9.0%).Therefore,the control precision and stability of the system can both be improved effectively.
1.3.3 简明健康状况调查量表(SF-36)评分[10] 该量表评估内容主要包括8个维度(一般健康状态、生理机能、生理职能、躯体功能、情感职能、社会功能、精神健康、精力)，36个项目，每个维度评分范围0～100分，评分越高患者生存质量越高。
Fig.12 System step response with and without PI controller
3.3.2 PI+disturbance observer(DOB)control
Since x ≈lθm as mentioned above,the reverse EMF of the motor can be calculated as follows:
Fig.13 Control structure of the FSM system including disturbance and measurement noise
In Fig.13,θr(s)is the input,θm(s)is the output,Gv(s)is the model of FSM including the velocity feedback loop,C(s)is the PI controller,Pθ(s)is the control quantity calculated by the controller,n(s)is the measuring noise,d(s)is the disturbance,which involves both disturbance factors such as friction,resonance,and vibration and interference caused by velocity feedback.The relationship between the input and output of the system is
where
The influence of external disturbance on the control precision of the system is suppressed by introducing a DOB[22–26]into the internal loop of the control system.The structure diagram of the FSM control system containing the DOB and PI controller is shown in Fig.14,in which the DOB can estimate the external disturbance d(s)and then apply an equivalent compensation
d(s)at the input of the control system to ensure that the output of the system is not affected by external factors.
Fig.14 FSM control system with DOB and PI controller
Among them,G(s)is the actual model of the FSM system,Gv(s)is the nominal or modeling model of the FSM system,which includes a velocity feedback loop,
is the inverse of the nominal model,Q(s)is a low-pass filter and
is the observation disturbance.
Assuming that the modeling error is ignored,that is,Gv(s)=G(s),the relationship between the input and output of the system is
where
By comparing(26)and(22),we can see that the introduction of DOB in the internal loop does not change the input and output relationship of the original system.A comparison of(27),(28)with(23),(24)indicates that the disturbance suppression ability and the sensitivity to noise of the system are not only related to the low-pass filter Q(s),but are also influenced by the PI controller C(s).In the low-frequency band,Q(s)=1.Thus,Gdm(s)=0 and Gnm(s) = -1.It shows that the DOB can completely suppress the external disturbance at low frequencies and ensure the robustness of the system to an external disturbance,but the system is more sensitive to the measurement noise.In the high-frequency band,Q(s)=0 and C(s)Gv(s)≈0.Thus,Gnm(s)≈0 and Gdm(s)≈Gv(s).This shows that the measurement noise of the system is basically filtered out in the high-frequency band,but the disturbance suppression ability of the DOB is reduced.Therefore,Q(s)can be regarded as a measure of external interference and measurement noise suppression.Assuming the disturbance suppression coefficient ηdis=Q(s),and the noise suppression coefficient ηnoi=1-Q(s).ηdis+ηnoi=1.It can be concluded that the suppression ability of disturbance and noise has an antagonistic relation,and we must make a compromise.Considering that the disturbance in the actual control system is generally located in the low-frequency band,and the measurement noise is located in the high-frequency band,we can make ηdis sufficiently large in the low-frequency band and ηnoi large enough in the high-frequency band through the design of Q(s).
The key for designing the DOB is the design of Q(s),in which the currently widely used low-pass filter is the binomial filter,and its specific form is
where αk is the coefficient of the binomial and αk =N!/[(N-k)!k!],N is the order of the denominator,M is the order of the numerator,M=0,1,2,...,N-1,and τ is the time constant with values ranging from 0 to 1.
From(29)we can see that the design of Q(s)involves three parameters:M,N,and τ.The choice of M and N should ensure that the relative order of Q(s)is not smaller than the relative order of the nominal model G(s)so as to make 
physically realizable.Moreover,from the point of view of system stability and real-time performance,the order of the low-pass filter should not be excessively high.The time constant τ determines the bandwidth of Q(s):the smaller the value of τ,the wider the bandwidth of Q(s),the stronger the ability of the system to suppress external disturbance.However,it also means a higher sensitivity to noise and a reduction in the robustness of the system.On the other hand,the larger τ is,the narrower the bandwidth of Q(s)is,the lower the sensitivity to noise is,and the more robust the stability is.However,the ability to suppress external disturbance becomes correspondingly weak.Therefore,the suppression of external disturbance and the robust stability of the system are mutually exclusive.Thus,a compromise is necessary by choosing the best time constant τ during design.
According to the relative order of the FSM system model Gv(s)shown in(19),the relative order of Q(s)should be greater than or equal to 3;hence Q(s)is designed as
When τ equals 0.000 01,0.000 1,0.001,0.01,0.1,and 0.99,respectively,the amplitude frequency characteristic curve of ηnoi and ηdis are shown in Fig.15.We can see that ηnoi is larger in the low-frequency band,and ηdis is larger in the high-frequency band,which indicates that the lowpass filter Q(s)is effective.Moreover,it can effectively suppress the disturbance in the low-frequency band and the measurement noise in the high-frequency band.When τ decreases,the bandwidth of ηdis increases,and the amplitude of ηnoi decreases in the low-frequency band,which means the noise suppression effect of the system can be expected to decrease.When τ equals the six respective values ranging from 0.000 01 to 0.99,there are six intersections between ηnoi and ηdis,as listed in Table 2.
Fig.15 Amplitude frequency characteristic curve of ηnoi and ηdis by varying τ
Table 2 Intersection frequency of ηnoi and ηdis by varying τ
From Table 2,we can see that wQ×τ ≈0.326,which indicates that wQ has almost an inversely proportional relation with τ.Furthermore,the intersections are basically located at the maximum of ηnoi and ηdis.That is,before the intersection frequency,the control system has almost the best suppression effect on external disturbance,whereas after the intersection frequency,the control system has almost the best suppression effect on the measurement noise.It can be seen from Fig.8 that the high-frequency resonance occurs after ωn=1 640 rad/s(261.1 Hz),which may come from the coupling of the rigid link in the mechanical connection with the resilient connection part,or the measurement noise of the sensor,which needs to be suppressed.Because ωn is located between Q2 and Q3,we take τ between 0.000 1 and 0.001.Under the condition of satisfying the robust stability of the system[27–29],we chose ωQ=1 500 rad/s(238.8 Hz)as the intersection frequency of ηnoi and ηdis,and then τ=0.000 217.Finally,Q(s)is designed as follows:
A comparison of the results of the suppression effect on the carrier(the mobile platform carrying the FSM system)disturbance Cd(unit N/m)and the moment disturbance Fd(unit N/m)are shown in Fig.16 and Fig.17 before and after the disturbance observer is used and when the input of the system is θr=sin t unit°/s,where
It can be seen that the ability to suppress the disturbance to Cd and Fd is obviously enhanced by the control of PI+DOB compared with using PI only.As shown in Fig.16,when the carrier disturbance increases abruptly at 10 s,the angle error between the input angle and the output angle is significantly increased under PI control,whereas there is no obvious increase in the angle error under the control of PI+DOB.The reason is that the DOB makes an equivalent compensation for the disturbance at the input of the system,such that the output of the system is not affected by external factors.The moment disturbance contains a constant component,and when Fd increases suddenly at 5 s as shown in Fig.17,there is a large instantaneous angle error at the time of mutation with PI control,whereas the instantaneous angular position error under PI+DOB control is significantly smaller than that for PI control.This indicates that the FSM system not only has a superior control precision,but also has a strong adaptability to the mutation of external disturbance under PI+DOB control.As a result,it is more suitable for practical applications in the external environment.
Fig.16 Comparison of the suppression effect of the system on carrier disturbance
Fig.17 Comparison of the suppression effect of the system on moment disturbance
4.Experimental prototype design
4.1 Overall design
The main components and overall control structure of the battlefield dynamic scanning and staring imaging system based on FSM are shown in Fig.18 and Fig.19.As shown in Fig.18,the servo control unit that contains the rotating platform control circuit and backscanning control circuit is the core part of the system.The control system should not only ensure the high-speed stepping and staring of the boresight,but also have sufficient high-accuracy control.
Fig.18 Main components of the battlefield dynamic scanning and staring imaging system
Fig.19 Overall control structure of the battlefield dynamic scanning and staring imaging system
As shown in Fig.19,the system adopts the compound axis control,including the position control loop of the rotating platform and the FSM.The position control loop of the rotating platform uses a double loop control structure,which consists of both position and velocity loops,and a gyroscope angular velocity sensor is applied to isolate external moment disturbances to ensure that images are scanned at a given velocity.
The imaging system mainly has three kinds of working modes:regional mode,panoramic mode,and manual mode.In the regional mode,the rotating platform is first rotated to the starting position of regional scanning by the transfer command,and then it rotates at a constant velocity in the scanning direction.Meanwhile,in order to obtain dynamic scanning and staring imaging,we control the velocity of the mirror by the backscanning command to compensate for the image motion caused by the scanning of the rotating platform.In the panoramic mode,the control system directly takes the existing position of the rotating platform as the starting point of scanning,and with the cooperation of the backscanning command,panoramic imaging can be completed.In the manual mode,the movement of the rotating platform is controlled by the handle;that is,the system can complete the monitor imaging when the rotating platform is either static or slowly rotated by the handle.In this mode,the position control loop of the FSM takes the residual velocity between the rotating platform and mirror as input,and functions as a secondary stability mechanism of the system.Due to the fast response velocity of the FSM system control loop,it is possible to compensate for the high-frequency image motion that exceeds the bandwidth frequency of the rotating platform control loop,which can further improve the stability accuracy of the boresight and realize clear imaging of the target.
4.2 Optical structure design
The optical path is shown in Fig.20,and the incident light reflected by the mirror of the FSM is divided into visible light and infrared light by using a spectroscope,after which the light is sent to the visible light and infrared cameras,respectively.The image motion caused by the motion of the rotating platform is compensated for by the same mirror,thereby resulting in the FSM being shared between the visible light and infrared cameras.
Fig.20 Optical structure design
It can be seen from Fig.8 that the lower order resonance of the FSM system occurs at ω1=246.1 rad/s(39.2 Hz)and its resonant peak gain reaches 25.8 dB,whereas the higher order resonances occur after ωn=1 640 rad/s(261.1 Hz).The lower order resonance at ω1,which is also known as the structural resonance,causes the amplitude margin h and the phase margin γ to be too small(h=-5.52 dB,γ=-5.22°),thereby seriously affecting the control bandwidth and stability of the system.Thus,the lower order resonance should be suppressed.The zero-pole point distribution of the closed-loop system is shown in Fig.9(without velocity feedback).As we can see,there are three poles:s1=-2.56×103,s2,3=16.626 4±490.46j,among which the pair of conjugate poles s2,3 is located in the right half of the S plane.This closed-loop system is unstable.
仿真结果表明：该电路设计和补偿网络设计合理，电路能够正常运行，并具有一定的抗干扰能力。系统运行稳定，静态误差小，动态性能良好。
4.3 Electrical design
First,the two-gigabit network signals are converted into optical signals through an upper fiber transceiver,and then the optical signals are transmitted to the lower optical transceiver via a slip ring.The received optical signals are re-converted into the two-gigabit network and finally transmitted to the information-processing computer.
Fig.21 Electrical connection diagram of experimental prototype system
The electrical connection diagram of the imaging system is shown in Fig.21.The management computer is responsible for receiving various operating instructions from the information-processing computer and transmitting them to the constituent units,such as the visible light and infrared cameras,laser range finder,and servo control board.In addition to this,it is also responsible for sending the video image,boresight angle,and other information back to the information-processing computer.The image video data obtained by the cameras is transmitted through two-gigabit networks,respectively.
式中：F驱min是清洁机器人的最小驱动力；μ是履带节与其滑轨之间的摩擦系数；α光伏面板与水平面的夹角；G是清洁机器人的重力(N)。
The final designed experimental system and its terminal interface are shown in Fig.22.Among the components of the system,the photoelectric sensors,servo control unit,system management unit,and other components are integrated into the rotating platform.The terminal interface can display the battlefield image in real time,and we can control the rotation and pitch movement of the rotating platform and distance measurement by using the control handle.
Fig.22 Final designed experimental prototype
The resolution of the visible light and infrared cameras is 1 600×1 200 and 640×480,respectively.Their FOV values are the same,the horizontal direction is 5.4°,and the vertical direction is 7.2°.Their optical axis consistency is less than 0.3 mrad.
The horizontal FOV of the system can be extended by the transverse scanning of the rotating platform,but the vertical FOV is still limited.The pitch angle of the rotating platform ranges from–10°to 60°.As shown in Fig.22(c),in order to compensate for the lack of vertical FOV,we can set a reasonable number of steps in the vertical direction.In regional mode,the rotating platform can automatically return to the original scanning point immediately after regional imaging is completed,following which it makes a 5°step in the vertical direction and performs regional imaging again in the transverse direction at the new vertical angle.Assuming that the vertical angle of the point at which scanning starts is 0°,after three steps,the height of the vertical FOV reaches 1 009.7 m and 1 682.7 m at the typical distance of 3 km and 5 km,respectively,which enables the system to provide a basic reconnaissance and surveillance capability for helicopters,unmanned aerial vehicles(UAVs),and other low-altitude and low-speed air targets.
5.Experiment and analysis
5.1 Imaging effect test under different imaging modes
The imaging effect of the visible light and infrared cameras is tested under three different imaging modes:stationary imaging,mobile imaging,and backscanning imaging.Stationary imaging refers to imaging that occurs when both the rotating platform and the mirror of the FSM system are in the stationary state during imaging.Mobile imaging means the rotating platform rotates continuously while the mirror of the FSM system remains stationary.Backscanning imaging is dynamic scanning and staring imaging when both the rotating platform and the mirror are rotated,and their speed is precisely“matched”by controlling the FSM system.
The experimental results are shown in Fig.23 and Fig.24,where it can be seen that extensive motion-blur is caused by the image motion in the mobile imaging mode.The reason is that there is relative movement between the camera and the scene during the exposure time.However,the backscanning imaging mode compensates for image motion by controlling the mirror of the FSM system,and remains relatively static between the scene and camera during exposure time.Finally,stable and clear imaging is realized,with no significant difference in the imaging quality compared with stationary imaging.
Fig.23 Comparison of imaging sequences of visible light camera in different imaging modes
Fig.24 Comparison of imaging sequences of infrared camera in different imaging modes
The edge and local details are the important features and information of an image that affect the visual quality of an image.Generally speaking,the narrower the transitional zone of the edge is,the clearer the image is.The point sharpness method[30–33]is an improved algorithm of edge sharpness[34,35],which mainly uses the changing information of a gray edge to judge the clarity of the image.This method is easy to implement and is suitable for evaluating the clarity of an image with rich details and texture features.In this work,the imaging effects are quantitatively analyzed in different imaging modes by using the point sharpness algorithm to evaluate the imaging quality of the system.
Suppose Pm×n is the evaluation function of the point sharpness algorithm:
新零售服务体系，是河北销售落实中国石油全面建成世界一流综合性国际能源公司、创建一流能源服务商和销售公司开创新时代销售企业转型发展新局面的决策部署，且围绕高质量服务而推出的一项服务标准。为此，河北销售推出了《新零售精益服务体系管理手册》，包括总则（企业文化、服务承诺、精益服务理念、总体目标）和10个章节，共涉及 3 6项工作标准。
As shown in Fig.25,m and n represent the length and width of the image;df is the grayscale value change;dx is the distance increment between pixels.
Fig.25 Schematic diagram of point sharpness algorithm
For a grayscale image of the size m×n,its point sharpness is Pm×n.For a color image of the size m×n,we first extract the Rm×n,Gm×n,and Bm×n components of the image,then their respective sharpness is evaluated,such that the final point sharpness value of the color image is
The calculated results of the point sharpness of the imaging sequences shown in Fig.23 and Fig.24 are presented in Table 3 and Table 4.Wherein,PRGB(a),PRGB(b)and PRGB(c)represent the point sharpness of the imaging sequences of the visible light camera in the stationary imaging,mobile imaging,and backscanning imaging modes,respectively,and PGRAY(a),PGRAY(b)and PGRAY(c)are the point sharpness of the imaging sequence of the infrared camera in the three imaging modes mentioned above.Since the point sharpness of the image is affected by the image content,in order to eliminate the impact,taking the point sharpness of the imaging sequences in stationary imaging mode as a reference,we normalize the point sharpness of the imaging sequences in the mobile imaging mode and the backscanning imaging mode,and the processing results can be concluded in Table 4 and Table 5,in which
where ηb/a and ηc/a denote the normalized point sharpness of the imaging sequences of the visible light camera in the mobile imaging and the backscanning imaging modes,respectively.
and
denote the normalized point sharpness of the imaging sequence of the infrared camera in the mobile imaging and the backscanning imaging modes,respectively.
Table 3 Point sharpness and normalized point sharpness of visible light images in various imaging modes
Table 4 Point sharpness and normalized point sharpness of infrared image in various exposure modes
As shown in Fig.26,the normalized point sharpness
and
are plotted as curves,respectively.The normalized point sharpness of the imaging sequence in the backscanning imaging mode is significantly higher than that in the mobile imaging mode,which indicates that the image clarity in the backscanning imaging mode is significantly higher than that of the mobile imaging mode.Moreover,the point sharpness values of the imaging sequence in the mobile imaging mode are more than 90%of the values in the stationary imaging mode,which indicates that there is no significant difference in image clarity between these two imaging modes and that the imaging quality of dynamic scanning and staring imaging of the system is good.This shows that our system not only is capable of monitoring hot targets in real time,but also has wide-area reconnaissance capability of hot regions.
Fig.26 Normalized point sharpness curves of imaging sequences in different imaging modes
5.2 Dynamic scanning and staring imaging effect test under different stall degrees
As shown in(5),in the normal backscanning imaging mode,the relationship between the rotation velocity of the mirror and the rotating platform is ωm=-0.5ωD,where ωD=2 093.3 mrad/s,and it will take 3 s to capture a complete panoramic view.In this paper,during the exposure time,we refer to the situation in which the velocity of the rotating platform and the mirror does not match as“stall”.Assuming that the stall coefficient is ζ,the velocity difference between the rotating platform and the mirror during exposure time is
By using a simple adjustment to the FSM control system,we set the stall coefficient ζ equal to 0,0.2,0.4 0.6,0.8,and 1.0,respectively.The effect of different stall degrees on the imaging quality of the system has been experimentally investigated.The imaging results are shown in Fig.27 when testing in the same scene.Taking the point sharpness of the same scene image in stationary imaging mode as reference,the point sharpness and its normalized point sharpness of Fig.27 are shown in Table 5,in which
Fig.27 Comparison of imaging effect of visible light camera under different stall degrees
Table 5 Point sharpness and normalized point sharpness of same visible scene under different stall degrees
As shown in Fig.28,the normalized point sharpness ηd/a#4 is fitted to a curve,and then the relationship between the normalized point sharpness and the stall degree can be obtained:
where ζ is the stall degree coefficient and f(x)is the normalized point sharpness.We can see that the greater the difference between the rotation velocity of the rotating platform and the mirror is,the higher the stall degree is.Further,the smaller the normalized point sharpness,the lower the image quality,the more blurry the image.
Fig.28 Fitting curve of normalized point sharpness under different stall degrees
If the secondary term of f(x)is omitted,the stall degree has a negative exponential relation with the clarity of the system imaging.As shown in Fig.28,when the stall coefficient ζ is 0.035 12(Δω=36.76 mrad/s),the normalized point sharpness of the image is reduced to 0.8.Because the velocity of the rotating platform is fast,the stall degree between the rotating platform and the mirror of the FSM system has a serious impact on the imaging quality of the system.Therefore,the control accuracy of the mirror of the FSM system is a particularly important aspect of the dynamic scanning and staring imaging system.
6.Conclusions
In this study,we design a dynamic scanning and staring imaging system based on FSM which is capable of realtime monitoring of hot targets and wide-area reconnaissance of hot regions.The design process resulted in the following ideas and conclusions:
(i)The FSM system can be mathematically modeled as a third-order system with a constant molecule term by ignoring the high-order resonance caused by the mechanical coupling between the rigid structure and the flexible support.
(ii)The influence of external disturbance on the control precision of the FSM system can be suppressed by introducing a disturbance observer into the internal loop of the control system.This can make an equivalent compensation for the disturbance at the input of the control system to ensure that the output of the system is not affected by the influence of external factors.The experimental results show that in the control of PI+DOB,the system not only has improved control accuracy,but also has stronger adaptability to the mutation of the disturbance.
(iii)With the help of the spectroscope,the image motion caused by the motion of the rotating platform can be compensated for by the same mirror,and the visible light and infrared cameras can share the same FSM.
(iv)The mobile imaging mode is adversely affected by extensive motion blurs.However,in the backscanning imaging mode,we can compensate for image motion by controlling the mirror of the FSM to realize stable and clear imaging,and there is no significant difference in the imaging quality compared with that of the stationary imaging mode.Moreover,the stall degree between the rotating platform and the mirror has a negatively exponential relationship with the imaging clarity of the system.As the velocity of the rotating platform is fast,the stall degree between the rotating platform and the mirror has a significant impact on the imaging quality of the system.Therefore,the control accuracy of the FSM system to a great extent determines the imaging quality of the dynamic scanning staring imaging system.
References
[1] QI G,LI J,LIU L,et al.Development of focal plane plate of optical remote sensor with large field of view.Optical Technique,2011,37(6):675–678.
[2] GORIN B A.Performance of imaging spectrometer methods for airborne reconnaissance with wide area coverage.Proceedings of SPIE,2002,48(24):88–101.
[3] ZHANG S,LIU B Q,HUANG F Y,et al.Super wide field of view staring infrared imaging technology and its application.Laser&Infrared,2016,46(10):1176–1182.(in Chinese)
[4] XU Y Y,XU R,LI F F,et al.Verification of programmable,large-FOV spectral imaging technology based on a staring/scanning area-array detector.International Society for Optical Engineering,2014,9263:65–77.
[5] QIU M P.Optical design of wide FOV infrared scanning imaging system.Infrared Technology,2012,34(11):648–651.
[6] JIAN F U,LIU H N.Research of large field of view scan mode of industrial CT.Chinese Journal of Aeronautics,2003,16(1):59–65.
[7] COURTNEY B.Scanning mechanisms for imaging probe.Journal of the Acoustical Society of America,2016,133(1):615–622.
[8] SUN C,DING Y,WANG D,et al.Backscanning step and stare imaging system with high frame rate and wide coverage.Applied Optics,2015,54(16):4960–4965.
[9] LAVIGNE V,CHEVRETTE P C,RICARD B,et al.Step-stare technique for airborne high-resolution infrared imaging.International Society for Optical Engineering,2004,5409:128–138.
[10] LONG Y,WEI X,WANG C,et al.Modeling and design of a normal stress electromagnetic actuator with linear characteristics for fast steering mirror.Optical Engineering,2014,53(5):054102–054110.
[11] TAPOS F M,EDINGER D J,NI M S,et al.High bandwidth fast steering mirror.Proc.of the Optics&Photonics International Society for Optics and Photonics,2005: 587707–587721.
[12] SUITE M R,MOORE C I,VILCHECK M J,et al.Fast steering mirror implementation for reduction of focal-spot wander in a long-distance free-space optical communication link.Proc.of the SPIE’s 48th Annual Meeting on Optical Science and Technology,2003:439–446.
[13] TANG T,HUANG Y,LIU S.Acceleration feedback of a CCDbased tracking loop for fast steering mirror.Optical Engineering,2009,48(1):510–520.
[14] LU Y,FAN D,ZHANG Z.Theoretical and experimental determination of bandwidth for a two-axis fast steering mirror.International Journal for Light and Electron Optics,2013,124(16):2443–2449.
[15] TIAN J,YANG W,PENG Z,et al.Inertial sensor-based multiloop control of fast steering mirror for line of sight stabilization.Optical Engineering,2016,55(11):111602–111608.
[16] NURHADI H,KUO W M,TARNG Y S.Study on controller designs for high-precisely linear piezoelectric ceramic motor(LPCM).Proc.of the IEEE Industrial Electronics and Applications,2010:1276–1281.
[17] SHENG S,LIU C F,YUAN Z Y,et al.Design of double-sided fast steering mirror based on piezoelectric actuating.Optics&Precision Engineering,2016,24(11):2777–2782.
[18] GENG W,LEI F,HONG Z.Current status and future development of precision tracking control technology of piezoelectric fast steering mirror.Piezoelectrics&Acoustooptics,2016,38(1):5–10.
[19] WU Q Y,WANG Q,PENG Q,et al Wide bandwidth control of fast-steering mirror driven by voice coil motor.Opto-electronic Engineering,2004,31(8):15–18.
[20] DONG L,CHEN J,ZHANG C,et al.A novel voice coil motor used in nano-positioning device.Proc.of the 19th International Conference on Electrical Machines and Systems,2016:1997–2002.
[21] WU X,CHEN S.High-powered voice coil actuator for fast steering mirror.Optical Engineering,2011,50(2): 23002–23007.
[22] OH Y, WAN K C, OH Y. et al. Disturbance-observer.IEEE/ASME Trans.on Mechatronics,1999,4(2):133–146.
[23] JO N H,JEON C,SHIM H.Noise reduction disturbance observer for disturbance attenuation and noise suppression.IEEE Trans.on Industrial Electronics,2017,64(2):1381–1391.
[24] DENG C,TANG T,MAO Y,et al.Enhanced disturbance observer based on acceleration measurement for fast steering mirror systems.IEEE Photonics Journal,2017,19(4):1–5.
[25] CHEN W H,YANG J,GUO L,et al.Disturbance-observerbased control and related methods-an overview.IEEE Trans.on Industrial Electronics,2016,63(2):1083–1095.
[26] PAN H,SUN W,GAO H,et al.Disturbance observer-based adaptive tracking control with actuator saturation and its application.IEEE Trans.on Automation Science&Engineering,2016,13(2):868–875.
[27] SARIYILDIZ E,OHNISHI K.Stability and robustness of disturbance-observer-based motion control systems. IEEE Trans.on Industrial Electronics,2014,62(1):414–422.
[28] JOO Y,PARK G,BACK J,et al.Embedding internal model in disturbance observer with robust stability.IEEE Trans.on Automatic Control,2016,61(10):3128–3133.
[29] EOM M,CHWA D.Robust swing-up and balancing control using a nonlinear disturbance observer for the pendubot system with dynamic friction.IEEE Trans.on Robotics,2017,31(2):331–343.
[30] WANG Z,XIE Z.A fast quality assessment of image blur based on sharpness.Proc.of the 3th International Congress on Image and Signal Processing,2010:2302–2306.
[31] XUE W X,BIAN C J,CHEN H Z.Image clarity evaluation based on point sharpness and square gradient.Electrical Design Engineering,2017,25(8):163–168.(in Chinese)
[32] HASSEN R,WANG Z,SALAMA M M A.Image sharpness assessment based on local phase coherence.IEEE Trans.on Image Processing:a Publication of the IEEE Signal Processing Society,2013,22(7):2798–2810.
[33] LIN J W,WENG Q,XUE L Y,et al.A retinal image sharpness metric based on histogram of edge width.Journal of Algorithms&Computational Technology,2017,11(3):292–300.
[34] DAS A,RANGAYYAN R M.Enhancement of image edge sharpness and acutance.The International Society for Optical Engineering,1997,30(26):133–142.
[35] LIN W S,GAI Y L,KASSIM A A.Perceptual impact of edge sharpness in images.IEE Proceedings Vision Image&Signal Processing,2006,153(2):215–223.
DOI: 10.21629/JSEE.2019.01.05
Manuscript received November 22,2017.
*Corresponding author .
This work was supported by the National Defense Pre-research Project of China during the 12th Five-year Plan Period(4040570201)and Innovation Project of Military Academy(ZYX14060014).
Biographies
CHANG Tianqing was born in 1963.He received his Ph.D.degree in concurrent engineering from Tsinghua University in 1999.Since 2000,he has been a professor in Academy of Armored Force Engineering.His current research interests include target detection and recognition,as well as navigational guiding and controlling.
E-mail:diegorevilo@163.com
WANG Quandong was born in 1989.He received his B.S.and M.S.degrees in electrical engineering from Beijing Jiaotong University in 2012 and 2015,respectively.He is a Ph.D.candidate in School of Academy of Armored Force Engineering.His current research interests include target detection and recognition,as well as navigational guiding and controlling.
E-mail:08291025@bjtu.edu.cn
ZHANG Lei was born in 1975.He received his Ph.D.degree in electrical engineering and automation from Academy of Armored Force Engineering in 2010.Since 2013,he has been an assistant professor in Academy of Armored Force Engineering.His current research interest is target detection and recognition.
E-mail:13611377718@163.com
HAO Na was born in 1982.She received her Ph.D.degree in electrical engineering and automation from the Academy of Armored Force Engineering in 2016.In 2002,she became a lecturer in Academy of Armored Force Engineering.Her current research interest is target detection and recognition.
E-mail:oliver chan1214@126.com
DAI Wenjun was born in 1992.He received his B.S.and M.S.degrees in electrical engineering from Academy of Armored Force Engineering in 2012 and 2015,respectively.He is a Ph.D.candidate in School of Academy of Armored Force Engineering.His current research interest is intelligent technology of fire control system.
E-mail:daiwenjun1993@163.com
标签：fast论文; STEERING论文; MIRROR(FSM)论文; dynamic论文; scanning论文; and论文; staring论文; backscanning论文; compensation论文; disturbance论文; OBSERVER(DOB)论文; point论文; sharpness论文; Department of Weapon and Control论文; Army Academy of Armored Forces论文; Center for Assessment and Demonstration Research论文; Academy of Military Sciences论文; Department of Science Research论文;