Abstract: This paper presents an algorithm of pedestrian detection and recognition by the integrate of 2D laser scanner and infrared image system. The algorithm can be divided into two main parts. Firstly, 2D laser scanner data is used to detect Regions Of Interest（ROI） of target by Robust Principal Component Analysis（RPCA）, and a window filter is designed for filtering the foreground matrix obtained by RPCA. The result of the first part is the positions of ROIs in 2D laser scanner data. Secondly, the position obtained in the first part is projected into the plane of infrared image and the ROIs in the infrared image can be obtained. Based on ROIs, feature vector is extracted by sparse coding spatial pyramid matching and pedestrian detection is finished by support vector machine. The experimental results show the effectiveness of the algorithm proposed in this paper.

Key words: 2D laser scanner, infrared image, pedestrian detection, robust principal component analysis, window filter

摘要： 研究了一种利用激光雷达数据引导红外图像进行行人检测与识别的方法。首先针对激光雷达数据，提出了一种利用鲁棒主成分分析进行目标感兴趣区域检测的方法，进而设计了一种窗口滤波算法对前景矩阵进行滤波处理，得到目标感兴趣区域的位置信息。在此基础上，将该位置信息投影到红外图像中获取红外图像中的目标感兴趣区域，进而在红外图像感兴趣区域内利用稀疏编码金字塔算法和支持向量机完成行人识别。实验结果表明了该算法能够有效地完成行人识别。

关键词: 激光雷达, 红外图像, 行人检测, 鲁棒主分量分析, 窗口滤波