论文部分内容阅读
稀疏表示是一种有潜力的图像信息表示方法,已应用于图像目标检测。正交匹配追踪算法(OMP)求解稀疏系数过程计算复杂,不能满足快速处理的要求,因此引入Kalman滤波器的递归思想,提出了一种计算稀疏系数的快速OMP(FastOMP)算法。利用Hermitian引理,从上一时刻的状态更新当前信息,避免了高维矩阵数据的重复计算。为提高算法的执行效率,提出了基于GPU/CUDA(图形处理器/统一计算设备架构)的并行计算方法,充分利用GPU的并行计算能力,提高了FastOMP算法的计算速度。实验结果表明,与传统OMP算法相比,FastOMP算法可大幅度缩短计算时间并提高检测精度。
Sparse representation is a promising method of image information representation and has been applied to image target detection. The orthogonal matching pursuit algorithm (OMP) solves the sparse coefficient process complexly and can not meet the requirement of fast processing. Therefore, the recursive idea of Kalman filter is introduced and a fast OMP (Fast OMP) algorithm for calculating sparse coefficient is proposed. Using Hermitian Lemma, the current information is updated from the state of the last moment, avoiding the repeated calculation of high dimensional matrix data. In order to improve the execution efficiency of the algorithm, a parallel computing method based on GPU / CUDA (GPU / CUDA) is proposed, which makes full use of the parallel computing power of GPU and increases the computational speed of FastOMP algorithm. Experimental results show that, compared with the traditional OMP algorithm, FastOMP algorithm can greatly shorten the calculation time and improve the detection accuracy.