To solve the problem that the information dynamics in constant weight summing is not flexible enough in air-combat situation assessment,a dynamic variable weight method is proposed for the situation assessment in close-range air combat.Based on the advantage function,the evaluation index system is improved by adding the air-combat capability index.Moreover,according to the Bayesian theory,the situation type is determined in real time,the characteristics of air-combat situation changes are integrated,and the weight of the evaluation index is dynamically adjusted according to the variable weight theory.The feasibility and effectiveness of this method in air-combat situation assessment are verified by the simulation experiments with real air combat data.Simulation results show that the dynamic variable weight evaluation method can effectively prevent the problem of poor information dynamics caused by constant weight summing.

Aiming at the highly dynamic and nonlinear characteristics of ballistic missiles,this paper introduces a nonlinear Cubature Kalman Filter (CKF) algorithm based on the third-order spherical-radial cubature rule.In addition,according to its characteristics,the Adaptive Cubature Kalman Filter (ACKF) algorithm for BDS/SINS deeply integrated navigation under the launch inertial system is studied.The algorithm is based on the idea of the Sage filter windowing method and the evanescent idea.By introducing multiple suboptimal fading factors into the CKF filter,the covariance matrix of observation error of the CKF filter can be adaptively adjusted online,and the fast-changing state can be effectively tracked while improving the filtering accuracy.Simulation results show that:Through the introduction of multiple suboptimal fading factors,the CKF filter can make better use of the systems prior information and make a stronger tracking of rapidly changing conditions;and the systematic errors can converge in a short period of time.Thus the dynamic performance of the integrated navigation system is improved greatly.

To improve the accuracy and robustness of the target tracking algorithm under illumination changes, a joint optimization algorithm combining illumination compensation with multi-task sparse representation is proposed based on the sparse representation theory.First,the algorithm compensates for the illumination of the template according to the average brightness difference between the template and the candidate target.Then,the candidate target is used to construct an over-complete dictionary to represent the template after illumination compensation,and the problem is transformed into a multi-task optimization problem.Moreover,the sparse coding matrix is used to quickly eliminate unrelated candidates.Finally,based on the reconfiguration error,a local structured assessment is carried out on the remaining candidates,so as to realize accurate target tracking.Simulation results show that the proposed algorithm can significantly improve the accuracy and robustness of target tracking under severe illumination changes compared with the existing state-of-the-art algorithms.

In order to implement tracking control for nonlinear systems with unknown dead-zone inputs,an adaptive dynamic surface control algorithm is proposed based on Extended State Observer (ESO).The function approximator is replaced by ESO to online estimate the uncertain function of each step in backstepping control.An adaptive dynamic surface output feedback controller is constructed based on the backstepping technology.The Tracking Differentiator (TD) is employed to eliminate the problem of large computational load in the conventional backstepping control algorithm.Finally,the stability proof is given and the proposed control method is verified by simulations.The results show that the algorithm has good tracking effects,strong robustness and good anti-interference ability.

To solve the problems of control surface redundancy and severe channel coupling of the tiltrotor aircraft in its transitioning process,we studied the rudder distribution strategy and the control channel switching strategy of the aircraft in the transition mode,and designed an attitude controller based on active disturbance rejection control.According to the characteristics of flight control,the structure of the active disturbance rejection controller was optimized,the order of the extended state observer was reduced,and the differential output of the tracking differentiator was used as the attitude angular rate instruction to simplify the structure of the attitude controller.Through the whole-envelope flight control simulation of the tiltrotor aircraft,the validity of the control system and the rationality of the rudder distribution strategy and the channel switching strategy are verified.

In order to accurately identify the aircrafts in remote sensing images,K-means algorithm is used to carry out a clustering analysis on the dataset based on YOLO V3 algorithm.By referring to Densenet theory,the two residual network modules in YOLO V3 are replaced by two dense network modules,and a Dense-YOLO Deep Convolution Neural Network (DCNN) is developed.The networks before and after improvement are trained respectively,and the weight files with which the two networks have the best recognition results are selected.The high-quality remote sensing images and the low-quality remote sensing images having the problems of over-exposure and cloud occlusions are tested and analyzed respectively.The results show that the application of the improved DCNN to the two kinds of images improves the recognition performance.In high-quality remote sensing images,the accuracy rate of the improved algorithm is as high as 99.72%,which is improved by 0.85%;the recall rate of the improved algorithm is as high as 98.34%,which is improved by 1.94%.In low-quality remote sensing images,the accuracy rate of the improved algorithm is as high as 96.12%,which is improved by 5.07%;the recall rate of the improved algorithm is as high as 93.10%,which is improved by 19.75%.

To address the issue of poor robustness and slow convergence speed in the calculation of optical flow under the influence of noise,a fast robust method for the optical flow field estimation in noisy environment is proposed.Based on the estimation method of optical flow in noisy environment,a penalty factor is introduced to enhance the robustness of the calculation of optical flow,a momentum factor is added to the iterative formula of optical flow calculation to shorten the convergence time of optical flow calculation,and then the calculation of the optical flow field is accelerated.The Euler-Lagrange equation is solved by minimizing the energy function of optical flow on the basis of the variation principle.Finally,the velocity field is obtained by using the iterative method.Simulation results show that,as compared to the M algorithm and the ML algorithm,the proposed algorithm can enhance the robustness of the optical flow considerably,shorten the convergence time of optical flow calculation and speed up the calculation of the optical flow field,after adding two different Gaussian noises to two consecutive frames in the video.

In order to address the issue of detection for weak and small moving targets in dynamic background,a classification method was proposed based on fast-LOF and optical flow trajectory.Weak and small moving objects only occupy a few pixels and are lack of features.To solve the problem,an idea of optical flow trajectory was introduced,and the abnormal optical flow trajectories were detected in high-dimensional space,so as to realize the detection of weak and small moving targets in dynamic background.To solve the problem of high complexity of the traditional LOF algorithm,fast-LOF was introduced to reduce the complexity of the anomaly detection link and ensure a good detection efficiency of the system.Experiments were carried out using videos captured by handheld cameras.The results showed that the algorithm can achieve rapid detection of weak and small moving targets with complex large field-of-view in white-light scenes.The combination of optical flow trajectory with fast-LOF effectively improved the performance and detection efficiency of the algorithm,which has certain value in visual detection systems.

In order to improve the tracking precision of the fDSST algorithm when the target is moving fast,deforms or even disappears,a long-term tracking algorithm based on TLD and fDSST is proposed.Based on the fDSST algorithm,a detector and an online learner are used to modify and learn from the tracking results.To solve the problem of learning wrong parameters when the fDSST algorithm fails,the positive and negative samples of the detector and the learner are used to evaluate the tracking results.The experiment results indicate that the long-term tracking algorithm based on TLD and fDSST solves the problem that the fDSST algorithm fails to track the target for a long time when the target is in fast moving,deforms or even disappears.This approach could also improve the tracking precision of the TLD algorithm.

In order to solve the problem of tracking failure of the Kernelized Correlation Filtering (KCF) tracking algorithm in the case of target scale changes and severe occlusion,an adaptive tracking algorithm is proposed based on kernelized correlation filtering.The algorithm uses a scale estimation strategy to adapt the tracking frame to target scale changes,and uses polynomial kernel functions to reduce the computational complexity.The FHog target feature is used to replace the original Hog feature to obtain more target feature information.In the experiment,50 sets of video sequences based on the OTB-2013 evaluation benchmark were tested and compared with other 31 tracking algorithms to verify the effectiveness of the proposed algorithm.The experimental results show that:the success rate of this algorithm is 0.549 and the accuracy is 0.736,ranking first,which is improved by 3.8% and 1.0% respectively compared with the KCF algorithm.The algorithm has strong steadiness and robustness under complex conditions such as target scale changes and severe occlusion.

This paper focuses on the study of UAV formation keeping based on velocity vector field.Firstly,the Leader-follower formation is taken as the research object,and the mathematical model of the UAV is built.Secondly,by referring to the artificial potential field method,the relative velocity vector is introduced,the attractive force function and the repulsive force function between the UAVs are given,and the modeling of the velocity vector field is completed.Under the influence of the velocity vector field,the UAVs can keep the formation and do not collide with each other.Finally,a simulation verification platform is built to verify the feasibility and effectiveness of the algorithm.

Group targets have fixed structure,particular motion patterns,large quantity,dense spatial distribution and serious mutual occlusion.Traditional multi-target tracking algorithms may lead to wrong data association and even target loss.Three kinds of typical research ideas for group target tracking are introduced,including the centroid group tracking algorithm,the tracking algorithm based on the random finite set and the extended state tracking algorithm.Then,the research achievements at present are analyzed and summarized.Finally, based on the existing theory and the development of advanced technologies in similar research fields,the development trends of group target tracking are discussed.

For the positioning and navigation in complex environment,a type of All-Source Positioning and Navigation (ASPN) system needs to be constructed to achieve multi-sensor plug-and-play and data fusion at different frequencies.This paper studies a data fusion method based on factor graphs.This method uses the factor graph method to represent the recursion and update of the state,and uses the Gauss-Newton iteration method to complete the data fusion task by solving the optimization equation in the integrated navigation.Then,taking the SINS/GNSS integrated navigation system as an example,the principle of the factor graph is analyzed and a corresponding information fusion framework is designed.Finally,the feasibility of the method is verified by simulation.The experimental data show that the three-axis position RMSE is 1.53 m,1.55 m and 1.53 m respectively,which proves the feasibility of the method.Based on this,the sensor can be extended to build an ASPN system.

To solve the problems of low ADS-B trajectory tracking precision and low matching efficiency between the target tracking model and the target motion model, the adaptive algorithm is used to improve the motion model set of the classic IMM algorithm. The uniform accelerated motion (CA) model in the standard IMM motion model set is improved to the “current” statistical (CS) model and the modified turn (MCT) model. The improved model set is used to filter the target's current position,velocity, and acceleration.The model transition probability is revised, so as to improve the adaptive ability of the IMM algorithm and realize rapid target tracking.The simulated track data and the measured data received by actual equipment are both used to verify the algorithm. The experimental results show that the improved IMM algorithm based on the motion model set has better filtering results than the classic IMM algorithm, and the tracking results are stable, which can adapt to the real-time tracking of complex target tracks.

To address the issue of effective evaluation of Reconnaissance/Strike UAV(R/S UAV) combat effectiveness in complex battlefield environment,the main factors affecting the operational efficiency of the UAV are analyzed based on the UAV's operational process.A combat effectiveness evaluation model of R/S UAV is built by using the Dynamic Bayesian Network (DBN) method,and the calculation methods of the variable attribute level and the key parameters are given.The Netica tool is utilized to dynamically evaluate and simulate the combat effectiveness of the UAV,and the validity and feasibility of this model are proved.The model can serve as a reference for the combat operation and equipment development of the UAV.

The allowable guidance handover area of Air-to-Air Missile (AAM) under the cooperative guidance of AWACS is studied.On the basis of analyzing the process of cooperative guidance,the mechanical models and motion models of the AWACS,the launching vehicle and the missile are set up respectively,and the restrictive factors that affect the handover area are analyzed from the aspects of the AWACS,the launching vehicle and the missile and their interrelationship.In the process of calculating the cooperative guidance handover area,the golden section method is used to simulate the model.In the simulation,different parameters of the AWACS and the missile are set respectively.According to the simulation result,the influence of these parameters on the guidance handover area is analyzed.The results show that,under certain situation,the guidance handover area has a positive correlation with the communication distance,the guidance distance,the beam width of the guidance antenna,and the beam width of the missile receiving antenna.However,the influence becomes less when the beam widths of the guidance antenna and the missile receiving antenna exceed a certain value.

The infrared detection range is an important indicator of the Infrared Searching and Tracking (IRST) system.How to quickly and accurately measure the infrared detection range has become a key part of IRST development.This paper analyzes the advantages and disadvantages of the current measuring method in the flight test,and proposes a new method to test the infrared detection range.This method uses the principle of contraction detection to introduce the real atmospheric path and real infrared scenes.The measurement of the infrared detection range is realized by accurately controlling the energy from a simulated radiating source.In view of the complicated and variable atmospheric transmittance near the ground,the correction factor of atmospheric transmittance is introduced to calibrate the atmospheric transmittance near the ground.Tests for two IRST systems of Type A and Type B were carried out.The maximum error between the measured infrared detection range and the flight test data was 12%.Compared with flight tests,this method greatly shortens the development cycle and reduces the development costs.

Temperature can change the polarization characteristics of liquid crystal materials,and the change will directly affect the display performance of Liquid Crystal Display (LCD).Therefore,if we obtain the LCD polarization characteristics in high-temperature environment,the changes of its display characteristics when it works in such environment are predictable.A method is proposed for the measurement of LCD polarization characteristics in high-temperature environment.First,a backlight LED and a polarizing film are placed in a temperature test chamber to simulate a high-temperature environment.The polarization state of the linearly polarized beam after passing the glass door of the chamber is measured by using a surface imager,so as to obtain the polarizing characteristics of the glass door.Then,the LCD is placed in the test chamber.The temperature and polarization characteristics of the LCD in the high-temperature state are obtained via the Stokes vector that is compensated by the temperature and polarization characteristics of the glass door.

To improve the accuracy and robustness of the cooperative robot Baxter when identifying and grasping target objects,the relevant image processing algorithms are studied.An algorithm for detecting the contour of the workspace is proposed based on HSV color model and morphological treatment,and the traditional Hough transform algorithm for detecting circles is improved.The experimental results show that,the proposed and improved algorithms can improve the accuracy and robustness of the cooperative robot Baxter when it is identifying and grasping target objects.

In order to solve the problem of boundary effects of the correlation filter,an object tracking algorithm is proposed based on similar background and color histogram in HSV color space.By using the Best-Buddies Similarity principle,similar backgrounds with higher similarity to the target are selected in the real background as the negative sample to train the correlation filter,so as to reduce the boundary effect.In order to improve the success rate of object tracking in complicated environment,the color histogram in HSV color space is combined with Bayes classifier for color tracking.Experiments are carried out on 16 videos selected from OTB-50 and OTB-100,and the results are compared with that of the current six tracking algorithms.The success rate and accuracy of the proposed algorithm are 0.593 and 0.467 respectively, which is superior to that of the other six algorithms.The proposed algorithm can effectively improve the success rate and accuracy of object tracking and has good real-time performance.

Based on the traditional PI control micro-nano manipulated imaging system,the parameters can be online adjusted in real time.In order to obtain better control effects,the simulation of the micro-nano manipulated imaging system of adaptive fuzzy PI control is studied.The principle that the atomic force between the probe and the sample remains unchanged is used,the dynamic process of the micro-nano manipulated imaging system is modeled,and the fuzzy PI control module is added to establish a complete system simulation platform.An adaptive fuzzy PI controller is designed,which can adjust the control parameters online in real time by the fuzzy reasoning of the voltage error and error variations,so as to optimize the control process.At the same time,Matlab is used for simulation.The results show that the adaptive fuzzy PI control algorithm can improve the dynamic and steady-state performance of the system more effectively than the traditional PI control algorithm.

In order to achieve light intensity compensation and reduce measurement error of fiber displacement sensor,a model of light intensity compensation and correction was proposed based on BP neural network optimized by Genetic Algorithm (GA).First,through the calibration experiment to the optical fiber displacement sensor,the original data was obtained.Then,the GA-BP neural network was used for modeling.Through the study on the encoding method,fitness function and parameters of GA,the global optimization capability of GA was used to optimize the weights and thresholds of traditional BP neural network,which made it less easier to fall into local extreme.Finally,the measured data was used to train the GA-BP network and the traditional BP network.The experimental results show that:compared with BP network,the GA-BP network has much smaller prediction error and higher compensation accuracy,and thus can realize the intensity compensation of the optical fiber displacement sensor.