融合人脸与步态周期模式的行人检测

A pedestrian detection method integrating face information and gait period pattern

提出多自由度的多核跟踪MeanShift算法,其在运动人体目标与背景图像的颜色信息较为接近时仍能鲁棒地跟踪.将所提出的跟踪算法用于融合人脸与步态周期模式的行人检测新算法,将闭环的控制思想引入到行人检测中,即通过步态周期和相应的跟踪反馈验证的理论方法来解决行人检测中误报率高的问题;还对行人部分轮廓存在遮挡的情况提供行人检测的新思路,即通过检测人脸来确定检测对象是否是行人,以解决当前行人检测算法检测率低的问题.

Multi-degree-of-freedom and multi-kernel MeanShift tracking algorithm was proposed in this paper. Whenever the color information of the target is close to that of the background image, the track was still robust. Moreover, this proposed tracking algorithm was applied to a novel moving pedestrian detection integrating face information and gait period pattern. The idea of closed loop control was involved into the pedestrian detection, i6, gait period detection and corresponding tracking feedback control validation were employed to solve the problem of high false positive rate in the pedestrian detection. Furthermore, to resolve the problem of low detection rate in pedestrian detection, the proposed algorithm may provide a novel attacking approach for the parts of contour with occlusion existing in pedestrian detection, and whether a pedestrian existed or not can be determined through human face detection.