Research Of The Realization On CRD Target Detection Algorithm Of Hyperspectral Image Based On FPGA

Due to the abundant spectral information in hyperspectral image,the related algorithms has been becoming more and more,hyperspectral image detection technology has been applied in many fields effectively,such as agriculture,medical and satellite reconnaissance,and the demand for real-time image processing in the market is increasing.However,due to the limitation of computing conditions and processing speed,the corresponding real-time image processing cannot achieve the goal of processing in real-time.While the field programmable gate array(FPGA),with many advantages,such as,small,low power consumption,reconfigurable and the characteristics of the radiation resistant has provided the possibility for the real-time processing of hyperspectral image on spaceborne equipment.For the aim of hyperspectral image processing on board,this article has put forward a hyperspectral image target detection of collboration respresention target detection(CRD)algorithm in real time detection based on FPGA parallel structure.The basic concept of the collboration respresention target detection(CRD)algorithm is that the target pixel can be approximated by its spectral signature,while the background pixel cannot.Based on the computational complexity and accuracy of the algorithm,the CRD target detection algorithm is selected to realize on FPGA platform.In the process of realizing this algorithm,high spectral image data is preprocessed on CPU to adapt to the hardware processing on FPGA platform at first.Then,CRD algorithm has been made some optimization with high demand of calculateing the matrix inversion,taking the Sherman-Morrison formula to realize the matrix inversion parallel in order to reduce the hardware complexity of CRD algorithm,and the software algorithm can be implemented on FPGA platform.And we have used the Intellectual Property of Xillinx to improve the frequency of the system.The achieved results demonstrate that the proposed system may obtains shorter processing time of the CRD algorithm than that on CPU.