Abstract: This paper proposes a detection algorithm of moving object based on self-adaptive threshold configuration, unlike previous approaches to object detection which detect objects by global threshold, it uses a local threshold to reflect temporal persistence which combines global threshold and local thresholds. The proposed approach can handle scenes containing gradual illumination variations and noise and has not bootstrapping limitations. Experimental results on different types of videos show the utility and performance of the proposed approach.

Key words: self-adaptive, threshold, kernel density estimation

摘要： 提出一种基于自适应阈值设置的运动目标检测算法，该算法不同于传统的全局阈值设置方法，而是利用核密度对背景像素点进行密度估计，给出一种新的全局和局部阈值相结合的自适应阈值设置方法。该方法考虑不同位置的像素颜色分布复杂度不同，针对每个像素点自适应设置局部阈值，能克服全局阈值的不足，提高检测的精度。对多个标准视频进行实验，实验结果证明了该算法的有效性。

关键词: 自适应, 阈值, 核密度估计

CLC Number: 

• TP18