基于自适应滑动窗的模糊场景行人快速检测

智能场景监控是公共安全领域研究的热点.然而当前公共场所监控视频大都比较模糊,如何能够在模糊场景下快速检测行人目标是一个亟待解决的问题.通过引入一个自适应尺寸的行人检测窗,以特定策略滑动扫描前景二值图像,并根据一系列行人轮廓判别条件获取行人目标,实现了对模糊监控场景的实时行人检测与人数统计.实验结果表明,与传统算法相比,该算法可以有效快速地检测出指定区域内的行人目标.

Pedestrian Fast Detection of Fuzzy Scene Based on Adaptive Sliding Window

Intelligent scene supervision is the focus of public security sector. However, most of surveillance video for public is not clear. Thus, fast pedestrian detection of fuzzy scene will be instantly resolved. In this paper, a new pedestrian detection algorithm based on adaptive sliding window is proposed. It builds a searching scheme on foreground binary map. By setting a series of outline discriminant conditions, it achieves the goals for pedestrian detection and people counting on fuzzy scene. Experiments results show that, this algorithm can fast effectively achieve the goals compared with some other algorithms.