一种改进的高光谱端元提取算法及其FPGA实现

Improved algorithm for hyperspectral endmember extraction and its FPGA implementation

自动目标生成算法在迭代更新正交子空间投影矩阵时,面临复杂的求逆运算和不断增长的迭代规模等问题,给算法的高速硬件实现带来困难.在对经典自动目标生成算法深入研究的基础上,采用固定规模的矩阵乘加运算代替复杂的矩阵求逆来更新正交投影算子,提出一种全新的自动目标生成算法的高速硬件实现架构.此外,首次使用高层次综合工具在现场可编程门阵列平台上完成了算法验证.实验结果表明,该硬件实现架构的处理速度达到了严格的实时性,同时保持了与传统算法相同的高检测精度.

The automatic target generation process (ATGP) is unsuitable for being deployed for high speed hardware implementation due to enormously complex inverse operations and the increasing scale of iterative operation problems.On the basis of an intensive study of the ATGP algorithm,a novel implementation framework is proposed in this paper,which takes advantage of simple regular matrix operations instead of increasingly complicated matrix inversion to update the orthogonal projection operator matrix.Furthermore,an accelerated design of the algorithm is implemented on FPGA by using high-level synthesis (HLS) for the first time.Experimental results demonstrate that the processing time achieved by the FPGA implementation is strictly real-time while retaining the same high detection accuracy.