一种基于时空兴趣点和光流法的行人检测方法

为解决视频监控系统中帧差光流法在光照变化大时对目标检测存在误差大的问题，提出一种基于时空兴趣点的目标检测算法。从光流的行人目标提取着手，首先对图像序列进行高斯滤波、Gabor滤波，提取到兴趣点集，然后对兴趣点集提取运动目标区域，接着在一定区域进行光流计算，识别并跟踪目标。实验结果表明，该方法的准确率比传统的帧差光流法的准确率高，且在简单的实际场景中更有效。

A Personal Detection Method Based on Spatio-temporal Interest Points and the Optical Flow Method

Big deviation will be generated by using frame-differential optical flow method on target detection. In order to solve the problem, this paper proposes a spatio-temporal interest point target detection algorithm. The authors extract optical flow from a pedestri- an target, apply the method based on spatio-temporal interest point first, and for image sequences Gauss filter and Gabor filter are used to extract interest points set, then the moving objects are extracted from the interest point set, and the optical flow is calculated in a certain area so that the targets are identified and tracked. The experimental results show that the accuracy of this method is higher than that of traditional frame differential optical flow one and the method is relatively simple and effective for actual scenes.