To get a more accurate fusion image with richer information, the Finite Discrete Shearlet Transform (FDST) is used to improve the multi-focus image fusion algorithm. The perfect translation invariance of FDST and its high efficiency of decomposition and reconstruction are used to obtain the coefficient of high- and low- frequency sub-band by multi-scale and multi-direction decomposition. The fusion method of adaptive weighting used together with regional standard deviation matching is used for the high frequency, and the non-negative matrix factorization of improved gradient projection is used for the low frequency, then the high- and low-frequency sub-bands after fusion are obtained respectively. At last, the fusion image is reconstructed by FDST inverse transform. The experiments on multi-focus image show that, with the improved method, the fusion image is more distinct in subjective vision, has better objective indexes with shorter running time. The fusion result not only retains the rich and effective information of the source image, but also has a good effect.

In order to deal with two existing problems of correlation filter based target tracking algorithms, a real-time target tracking algorithm based on kernel correlation filtering is proposed, which combines feature fusion with quasi-residual updating, strategy. Firstly, to solve the problem that correlation filters employ the same coefficient to fuse different kinds of features, a correlation filter based on context-awareness with adaptive feature fusion is proposed according to the average correlation peak energy. Furthermore, the correlation filter based on context-awareness and adaptive feature fusion is combined with a Bayes classifier to construct a robust tracker by ensemble learning. Finally, focusing on the use of high-risk updating strategy in correlation filters, an updating strategy that is similar to deep residual networks and an adaptive learning rate updating model based on step function are proposed to prevent tracking model from drifting. The tracker proposed in this paper is compared with another 9 state-of-the-art trackers on OTB2013 and TC128 benchmarks. The experimental result on OTB2013 benchmark is that the proposed tracker ranks first on precision (0.875) and success rate (0.652), which indicates that the proposed tracker is robust and effective.

A Synthetic Aperture Radar (SAR) target recognition method based on the multi-scale monogenic features is proposed. To fully exploit the discrimination capability of the multi-scale monogenic features, the Multiset Canonical Correlation Analysis (MCCA) is used to fuse the different types of monogenic features from different scales including local amplitude, local phase, and local orientation, which results in a feature vector containing the internal correlations of each kind of feature. In the classification stage, the Joint Sparse Representation (JSR) is employed to classify the feature vector fused by the three kinds of features, and to further exploit the internal correlations of different types of features. Finally, the target type is decided according to the reconstruction errors from JSR.Experiments are conducted on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset to evaluate the performance of the proposed method, and the results prove the validity of the proposed method.

Aiming at the problem that the Sage-Husa adaptive Kalman filter algorithm is easy to diverge and is sensitive to the selection of initial conditions, an adaptive strong tracking algorithm based on Sage-Husa adaptive Kalman filter algorithm is proposed. The algorithm starts from the Sage-Husa adaptive Kalman filter algorithm and introduces the strong tracking technology in it. By online modifying the one-step prediction error covariance matrix with fading factor, the algorithm can cope with the uncertainties such as scene changes, and has stronger robustness. By improving the Sage-Husa adaptive algorithm, the noise covariance matrix is estimated online in real time, so that the algorithm is adaptive to noise changes and has better tracking accuracy. Simulation results show that the proposed filter method can effectively improve the tracking accuracy and robustness of the carrier tracking loop.

Aiming at the problem of poor quality of side-scan sonar image obtained by conventional image mending methods, this paper proposes a method of side-scan sonar image mosaic based on affine transformation method of triangulation. This method divides the strip image into a pair of triangles. While keeping the coordinates of one triangle unchanged, the other triangle is adjusted to fit with the adjacent strip image. Then affine transformation is used to realize the coordinate mapping for completing the seamless mosaic of the image. The experimental results show that:1) This method can effectively complete the seamless mosaic of images with a fine visual effect;and 2) It has higher image quality and better performance in terms of information entropy, signal to noise ratio and average gradient compared with gap filling method.

This paper proposes an improved A* algorithm for 3D spatial path planning. Considering the complexity of UAV flight environment in low-altitude penetration, the algorithm takes the flight altitude and flight length as the weight factors to search the optimal flight path between two waypoints. The variable-step-size search is used to improve the search efficiency of A* algorithm, and the path evaluation function with variable weight is designed to improve the optimization effect. Meanwhile, to satisfy the UAV performance constraints, such as pitch angle, yaw angle and radius of turn, a series of optimization algorithms are proposed to obtain the final flyable path. The simulation results show that the proposed method can provide more satisfactory path-planning for UAVs．

In order to realize real-time multi-target localization with high accuracy, a target localization method based on total least squares method is proposed. The nonlinear multi-point target localization model is linearized, the extended matrix of coefficient matrix and observing vector is built, and the localization result is calculated by using singular value decomposition theory. Then, the affine transform model method is adopted to realize multi-target localization, and the accurate localization information is obtained by adjustment calculation.Flight experiments show that, the proposed approach can improve the localization accuracy and realize the real-time multi-target localization. This approach is easy to implement, has high reliability and considerable application prospect.

When the Autonomous Underwater Vehicle(AUV)executes a task underwater, it faces several problems including few navigation means, positioning error drift of traditional dead reckoning, and that the unitary navigation method is not enough to meet the requirements for implementing precise underwater tasks.An integrated navigation method based on underwater Ultra-Short Base Line (USBL) acoustic positioning system, Global Positioning System (GPS) and Tilt compensation Compass Module (TCM) is proposed, and the coordinate transformation involved is described in detail. The Kalman filter is used to estimate the position of the AUV for obtaining the position and speed information needed for control and navigation.There is no positioning drift. and the navigation and positioning accuracy is better than 10 m, which can satisfy the requirement of AUV when executing underwater tasks for a long time. The method has been tested and verified on an unmanned AUV.

A scheme is proposed for the quad-rotor UAV to implement autonomous tracking with an airborne camera. The machine vision module OpenMV is used to acquire the ground image, and the image is divided into five regions. By calculating and analyzing the black lines in each region, various track elements are identified to obtain the horizontal deviation of the UAV relative to the path and the angle between the UAV and the path.The position outer loop is designed by the Linear Active Disturbance Rejection Controller (LADRC), and the speed inner loop adopts a PID controller, and the two loops form a position-speed cascade control to eliminate the horizontal error. The bang-bang control algorithm is then used to correct the aircraft heading. Experiments were carried out to track the 8-shaped trajectory. The results show that the scheme has a good path recognition effect and can make the quad-rotor UAV accurately and stably track the ground black guide line.

Taking the coaxial double-rotor high-speed helicopter as the research object, a new safety protection and control method based on command constraints is proposed for the safe envelope protection of forward-flight attitude of the helicopter flying at low speed. Firstly, the full-modal nonlinear mathematical model of high-speed helicopter is established. Then, taking pitch-angle protection for the low-speed forward flight as an example, the boundary parameter detection is realized by the dynamic parameter estimation of the restricted parameter and the dynamic trimming algorithm, thus to modify the control input command for safety. Finally, the simulation results show that the control method can protect the flight boundary of high-speed helicopters and improve the safety and reliability of flight.

Aiming at the uncertainty of quad-rotor aircraft model and its sensitivity to external disturbances, a Reduced-order Active Disturbance Rejection Control (RADRC) method is designed to control its position and attitude, and the stability of the control system is analyzed. In this method, the Reduced-order Linear ADRC (RLADRC) is combined with nonlinear tracking differentiator to meet the different requirements of each channel of the quad-rotor aircraft on control performance.Among them, the Reduced-order Linear Expansion State Observer (RLESO) only observes the system output differential signals and disturbances, avoiding repeated observations of known information, so that the system bandwidth can guarantee the estimation accuracy of RLESO, and reduce the noise sensitivity of the system. The trajectory tracking ability and anti-disturbance performance of RADRC are compared with that of the traditional LADRC method through simulation and Qball2 platform experiments, which verifies the effectiveness and superiority of RADRC method.

To get a more accurate fusion image with richer informationthe Finite Discrete Shearlet Transform (FDST) is used to improve the multi-focus image fusion algorithm.The perfect translation invariance of FDST and its high efficiency of decomposition and reconstruction are used to obtain the coefficient of high- and low- frequency sub-band by multi-scale and multi-direction decomposition.The fusion method of adaptive weighting used together with regional standard deviation matching is used for the high frequencyand the non-negative matrix factorization of improved gradient projection is used for the low frequencythen the high- and low-frequency sub-bands after fusion are obtained respectively.At lastthe fusion image is reconstructed by FDST inverse transform.The experiments on multi-focus image show thatwith the improved methodthe fusion image is more distinct in subjective visionhas better objective indexes with shorter running time.The fusion result not only retains the rich and effective information of the source imagebut also has a good effect.

The jamming resource allocation in electronic warfare is a typical combinatorial optimization problem, and is also a NP-hard problem.Besides the typical mathematical programming approaches, all kinds of intelligent optimization methods have been widely used in recent years too. In this paper, the jamming resource allocation mathematic model and its typical solving framework are firstly introduced. Taking the genetic algorithm, immune algorithm, evolutionary computation, particle swarm optimization algorithm, ant colony algorithm, and the various improved methods as typical examples, a review is made to the research progresses of application of such intelligent optimization methods in jamming resource allocation problem. The main points of the application of each method are summarized with its development status.It can be taken as a reference for the engineering application of each method, and for the research of artificial intelligence in electronic warfare.

In order to improve the observability and precision of initial alignment of traditional vehicle-mounted Strap-down Inertial Navigation System (SINS), a multi-position alignment method based on the theory of spectral condition number observability analysis is researched.The error model of SINS stationary base alignment is established, the Piece-Wise Constant System (PWCS)theory based on spectral condition number is used to analyze the observability of single-position alignment, optimal dual-position alignment and three-position alignment.The estimating error and alignment precision of inertial devices are contrastively researched, and time distribution of optimal dual-position alignment is discussed. The conclusions are achieved that the optimal dual-position alignment gets higher precision in the same alignment time, and three-position alignment can improve the estimating precision of orientation gyro for the subsequent navigation. The correctness of theoretical analysis is verified by simulation.

The problem of distributed MIMO radar in range-spread moving target detection under non-homogeneous environment is studied. Due to the fact that the disturbances (clutters and noises) on distributed MIMO radar are non-homogeneous, there is little homogeneous training data to estimate the disturbance covariance matrices. A new knowledge-aided model is proposed to depict the non-homogeneous characteristics of the disturbance in distributed MIMO radar.Specifically, the disturbance covariance matrices corresponding to different Transmit-Receive (Tx-Rx) pairs are modeled to obey the complex inverse Wishart distribution, but have different power levels.Based on the above prior knowledge, a novel Knowledge-Aided Generalized Likelihood Ratio Test (KA-GLRT) range-spread target detector is proposed for distributed MIMO radar without training data. Simulation results show that: compared with conventional covariance matrix-based (with training data) detectors, the proposed KA-GLRT detector has better detection performance in non-homogeneous environment.

This paper focuses on the modal aliasing resulted from the abnormal signal in practical applications of Empirical Mode Decomposition (EMD). A method based on two-scale proportional criterion and local EMD is proposed. First, an analysis is made to the impact of the abnormal signals on EMD algorithm, and the two-scale criterion combining the time characteristic scale with the extreme difference is established. The time domain location of the abnormal signal is determined by the criterion, and the abnormal signal is then separated from inherent signal by means of local EMD during the period. Numerical simulation results and comparative analysis show that the method can settle the problem of modal aliasing caused by abnormal signal more effectively compared with conventional methods.

Considering that the internal parameters of the given-depth electro-hydraulic servo system are uncertain and time-varying, a composite control strategy based on Neural Network (NN) and Sliding Mode Control (SMC) is designed to solve the problem. The strategy makes full use of the strong robustness of the SMC for the control of the given-depth electro-hydraulic servo system, enables it to work under complicated external interference environment, and uses the neural network to weaken the buffeting shortcoming of the SMC itself. In addition, the neural network is also used for identification of the internal parameters of the electro-hydraulic servo system and online adjustment, which can well suppress the internal parameter perturbation and eliminate the time-varying problem of the electro-hydraulic servo system, and thus can ensure the control precision and stability of the system.

A dual-band infrared optical system is designed for the dual-band cryogenic detector of 640×512 array, which can be used in the airborne infrared search and tracking system. The optical system adopts the combination of Ge, Zns, and ZnSe to achieve the achromatic and athermalization design.By introducing aspheric surface and configuration optimization, the high order aberration of the system is well corrected and the system structure is simplified.The optical system consists of only 8 lenseswith the working bands of 3.7~4.8 μm and 8~9.4 μm, F-number of 2, the focal length of 360 mm, and the FOV of 2.54°×2.03°, meeting the 100% efficiency of a cold aperture. The image quality evaluation results show that the imaging quality of the optical system is good in the temperature range of -55 ℃ to +70 ℃. The analysis of non-uniform temperature field shows that the optical system has a fine temperature gradient tolerance.

An object tracking algorithm is proposed based on the deep learning detection algorithm of YOLOv3 (YOLOv3:An Incremental Improvement), which utilizes the advantages of deep learning model in target feature extraction, and extracts candidate targets by using regression-based YOLOv3 detection model. The target color histogram feature and Local Binary Pattern (LBP) feature are also used for target screening, thus to implement object tracking.At the same timea method called K-neighbor searching is presented to improve algorithm performance, which performs neighborhood detection for the selected targets. Experimental results show that the proposed algorithm has a good tracking performance, with an overall performance improved by about 80% in comparison with the four contrast algorithms, and has good robustness in the complex situations of illumination changing, posture changing, size changing and rotation of target object.

Aiming at the deficiencies existing in the DC/DC circuit fault prediction with distance as the fault characteristic parameter, the Piecewise Dynamic Time Warping (P_DTW)distance is proposed as the fault characteristic parameter of DC/DC circuit, which is used together with the Long Short-Term Memory (LSTM) algorithm for the fault prediction of DC/DC circuits. First, the circuit output voltage is chosen as the monitoring signal. Then, the dynamic time warping algorithm is used to calculate its P_DTW distance with the faultless output voltage, which is used as the fault characteristic parameter of the circuit. Finally, the time series prediction of circuit fault characteristic parameters is implemented based on the LSTM prediction model.The Boost circuit is taken as an example to verify the effectiveness and accuracy of the proposed method.

In order to solve the problem of variable task requirements in Multi-State System(MSS), the Markov model is extended, and a performance-demand model of multi-state repairable system is derived based on the basic definition of Markov reward model.On this basis, a reliability evaluation method is proposed for the multi-state repairable system considering variable task requirements by certain transformation of reward matrix and transition intensity matrix. The feasibility of the method is verified by using a case.

A detailed analysis is made to the development of airborne safety-critical system onboard the third and the fourth generation civil aircrafts at home and abroad. The technical characteristics of such typical fault-tolerant technology as the multi-channel voting etc. are summarized.Aiming at the challenges of the new-generation aircrafts on the reliability, fault-tolerant capability, fault-detection/fault-isolation capability, and real-time performance of the airborne safety-critical computer systems, a design idea of a high-integrity computer system based on hierarchical fault-tolerant technology is proposed.The design principle of the hierarchical high-integrity fault-tolerant system is described, and the critical technologies of Lock-step and time-trigger network are given. The strategy can satisfy the development requirements of the next-generation aircrafts, and can improve the safety and reliability of the airborne safety-critical computer systems.