Implementation Of Hyperspectral Target Detection Algorithm Based On FPGA

Hyperspectral target detection technology has always been one of a hot research area in the field of hyperspectral image processing.However,the large amount of sets and high-dimensional signatures make the implementation of target detection algorithm a technical problem,which restricts the development of hyperspectral target detection.But,with the optimization of detection algorithms and the advancement of detection platforms,especially the development of field programmable gate array(FPGA),many image processing algorithms can be implemented by hardware.FPGA contains a large number of storage and logic resources,and the data processed scale continues to grow.Moreover,it has the advantages of high speed,low power consumption and high reliability,making it occupy an important position in the field of hyperspectral image processing.This thesis implements the hyperspectral target detection algorithm based on FPGA.Since the OSP algorithm does not depend on the statistical model of the data when the target is detected,and can be directly derived from the linear mixed model,it is choosen as the target detection algorithm to be implemented in this paper.Firstly,the hardware system of OSP algorithm is designed.From top to bottom,the system is divided into interference spectrum signature matrix multiplication module,matrix inversion module,detection operator operation module and detection quantity operation module.Furthermore,every module is described by Verilog HDL language.The function simulation of each module and system is completed by using Modelsim software,and gives the corresponding simulation results.And then,the detection quantity of the system output is read and displayed in MATLAB software,and the detection performance of system is assessed by ROC curve.The correctness of the system is verified by comparing with the detection results of the corresponding algorithm in MATLAB software.Secondly,the matrix inversion of OSP algorithm is more complicated in hardware implementation and consumes more resources.However,the calculation of the improved ROSP algorithm is relatively simple,the hardware implementation of ROSP algorithm based on FPGA is further studied.There gives the overall design of the system and the specific design method of each module.And then the target detection result of the system is got by compiling and simulation.After that,the result of the hardware implementation system of the ROSP algorithm are read and analyzed,and compared with the result realized by the MATLAB software to verify the function of the system.The implementation of hyperspectral target detection OSP algorithm and ROSP algorithm based on FPGA have completed the target detection of hyperspectral images well,the OSP algorithm implements a faster system,while the ROSP algorithm implements fewer logic resources.