基于深度摄像机的行人检测方法研究与实现
首发时间:2012-11-21
摘要:本文介绍了一种基于深度摄像机的行人检测方法,致力于快速实时地检测出视频中的行人。该方法充分利用了深度摄像机所特有的深度信息,经过预处理去除噪声和背景,使用模板匹配在目标图像上进行寻找,确定了最匹配点之后,结合深度信息进行漫水填充,进一步增强行人区域的准确性,最后再使用凸包分割对结果进行优化。经实验验证,该方法的检测准确率可达90%以上,在检测效率上亦可达到25帧/秒左右,相较于传统的行人检测方法,该方法可以兼顾检测的准确性以及实时性,因此可以很好地应用于实时视频中的行人检测。
关键词: 计算机应用技术 深度图 行人检测 模板匹配 漫水填充 凸包
For information in English, please click here
The Research and Realization of Pedestrian Detecting Based on Depth Camera
Abstract:In this paper, a pedestrian detecting method based on Depth Camera is proposed, it aims to label out the pedestrians in videos accurately and instantly. By taking full advantage of the depth information, the method can easily get the outline of the interested object, then uses template matching and optimizes the matching result to finish the detecting. By experiments, the accuracy rate can exceed 90 percents, and the detecting speed is about 25 frames per second. Therefore, the method can be perfectly used for pedestrian detecting in real-time videos.
Keywords: Computer Application Technology Depth Image People Detecting Template Matching Flood Filling Convex Hulll
基金:
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
基于深度摄像机的行人检测方法研究与实现
评论
全部评论0/1000