针对夜间动态背景下的行人检测中分割算法受光照条件影响大、误识别多等问题，提出双阈值分割算法和以多目标跟踪为核心的算法框架。新的分割算法解决了行人亮度分布不均时的...

针对夜间动态背景下的行人检测中分割算法受光照条件影响大、误识别多等问题，提出双阈值分割算法和以多目标跟踪为核心的算法框架。新的分割算法解决了行人亮度分布不均时的分割问题，同时在新的框架下可以综合多帧的处理结果进行综合判断，通过将基于支持向量机的识别算法和多目标跟踪算法的融合，降低了系统的计算量，且比一般的系统具有更高的识别率。 -At night the dynamic context of pedestrian detection segmentation algorithm affected by light conditions, and false identification of many other issues, the dual-threshold segmentation algorithm and a multi-target tracking algorithm as the core framework. New segmentation algorithm to solve the uneven distribution of brightness of pedestrians at the time of partition, while in the new framework of an integrated multi-frame can be the result of a comprehensive treatment of the judge, through the support vector machine-based recognition algorithms and multi-target tracking algorithm integration reducing system of calculating the volume, and than a typical system has a higher recognition rate.

一种改进的夜间行人检测算法.pdf