作　　者： (熊斌）; (郭海如）;

机构地区： 湖北工程学院计算机与信息科学学院,湖北孝感432000 华中科技大学自动化学院,湖北武汉430074

出　　处： 《计算机工程与设计》 2017年第9期2472-2476,共5页

摘　　要： 为提高红外行人目标的正确检测率,优化检测算法的计算复杂度,提出一种基于图像重组和矩阵恢复的红外行人目标检测算法。将原始图像重组成由图像子块构成的模型图像;将红外行人目标的检测问题转换为求解鲁棒主成分分析的优化问题,运用增广拉格朗日乘子算法将模型图像分解为背景图像和目标图像;通过一种自适应阈值方法,消除噪声并分割出行人目标。实验结果表明,该方法能有效准确地检测出行人目标,较同类方法有更好的实时性。 To improve the correction detection rates of infrared pedestrian target and optimize the computational complexity of de-tection algorithm, an algorithm based on image reconstruction and matrix recovery was proposed. Model images were reorganized using sub-block image from original image. The infrared pedestrian detection problem was converted into solving the optimization problem of robust principal component analysis？ and model images were decomposed into background image and target image using augmented Lagrange multiplier algorithm. Noise was eliminated and the target was segmented using an adaptive threshol-ding method. Experimental results show that the proposed method can effectively and accurately detect infrared pedestrian target and has better real-time performance compared with other similar methods.

关 键 词： 鲁棒主成分分析 图像重组 矩阵恢复 增广拉格朗日乘子法 行人检测