基于支持向量机提取感兴趣区域的行人检测

摘要

Considering the characteristics of support vector machine when detecting pedestrians,a pedestrian detection algorithm based on region of interest extracted by support vector machine was proposed.Support vector machine was used for early detection of image and the frames with higher reliability were extracted.According to statistical result of the frame's reliability in the directions of horizontal and vertical,the region of interest was extracted.Pedestrian detection was performed only in the region of interest.Results of tests on PASCAL test set show that this algorithm can retain the recall rate of the compared algorithm and it effectively reduces its detection time and false alarm rate.%综合考虑支持向量机在检测行人时表现出的特点,提出一种基于支持向量机提取感兴趣区域的行人检测算法.利用支持向量机对图像进行初检并提取出置信度较高的检测框,先后在水平和竖直方向统计这些检测框的置信度叠加和并根据统计结果提取出行人感兴趣区域,在提取的感兴趣区域内进行行人的检测.在PASCAL的测试集上的测试结果表明,该算法在保持对比算法召回率的前提下,有效降低了检测时间和虚警率.