人体运动自动检测与跟踪是人工智能系统的重要研究方向.提出了一种简单高效的人体感兴趣区域——ROI的自动检测方法.该方法运用帧差法和图像显著图算法模拟人眼视觉系统对运动人体注视点的形成过程,并通过联合显著区域的划定自动检测形成相应的运动人体ROI.实验结果表明,该方法检测出的人体ROI符合人眼观察运动目标时的视觉关注区域,可运用于后续基于颜色特征区域的运动人体跟踪.
Automatic detection and tracking of human motion is an important research direction of artificial intelligence. The paper presents a simple and effective method to automatically detect a region of interest (ROI) in human body. The frame difference of image and saliency-map algorithm are used to simulate the process of the focus of human body movement in human visual system, and automatically detect a ROI in human body by demarcating the region of joint saliency. The experimental results show that the ROI in human body detected by this scheme is consistent with the target area of the visual attention. It can be used in tracking of moving body based on the characteristics of color.