Algorithm And Real-time Processing Implementation Of Target Detection And Recognition For SAR Image

The detection and recognition for synthetic aperture radar(SAR)images play an important role in military and civil fields.SAR images have many difficulties in image detection and recognition because of the geometric characteristics of foreshortening,layover and shadow.On the other hand,due to the large amount of data of SAR image and the complex operation of detection and recognition algorithm,it further increases the difficulty of real-time processing of SAR image detection and recognition.According to the needs of real-time processing of SAR image target detection and recognition,the thesis studies and analyzes the SAR image target detection and recognition algorithm,develops a high-speed signal processing board that can meet the needs of real-time processing,implements the detection and recognition algorithm on the signal processing board,and finally verifies and tests the function and performance of SAR image detection and recognition algorithm on the signal processing board.The main contents of the thesis are as follows:1.The thesis studies the detection and recognition algorithm based on maximally stable extremal regions(MSER)and support vector machines(SVM)and the integrated detection and recognition algorithm based on convolutional neural network,and analyzes the hardware implementation method and computation of the two algorithms.2.According to the algorithm characteristics and hardware implementation requirements,the thesis develops the signal processing board based on ARM+FPGA architecture,and designs the power supply,clock and other key parts of the signal processing board.3.According to the platform architecture and the calculation characteristics of the algorithm,the detection and recognition algorithm is split and deployed on ARM and FPGA respectively.Through the software and hardware collaborative design of ARM and FPGA,the algorithm is implemented on the signal processing board.4.The hardware interface and the implementation of detection and recognition algorithm are verified on the signal processing board,and the performance of the two detection and recognition methods on the target board is analyzed.The above research work has been verified by simulation or measurement,and the results show that the hardware implementation method of SAR image target detection and recognition algorithm proposed in the thesis can effectively reduce the processing time of the algorithm and implement the real-time processing of SAR image target detection and recognition.