计算机视觉——行人检测方法改进

Computer Vision--Improvement in Pedestrian Detection

该文在阐述行人检测方法改进的主要思路、范围和方法的基础上，进一步对行人检测所使用的统计方法和特征进行了分析比较，提出了基于融合分类器的行人检测算法，并在检测前先进行前景标注，然后再在带标注的图像上进行行人检测。

This passage is about pedestrian detection and its improvement concerning its main idea, scope and method. The pedestrian de- tection based on statistical methods are analyzed and combined with the characteristics. The method of merged classifiers which can im- prove the detection rate on the condition that the performance will not decrease is used in this experiment. A marking foreground method is proposed that the foreground is extracted and marked, so that only the foreground pixels are detected instead of all of the image at the de- tection stage, which can improve the accuracy of detection, and make the detection speed more faster.