Abstract: This paper proposes a background model based on fusion of color and edge information. Color information is exploited to perform pixel-level background and foreground classification, and a novel edge based background subtraction method is applied to separate the foreground/background edges. The foreground-pixels are clustered to form hypnosis moving targets, which are further classified into true moving targets and false moving targets by incorporation of edge information. The method effectively removes the false moving target caused by sudden illumination change and the relocation of background objects.

Key words: background modeling, removal of false moving targets, edge based background subtractioninformation fusion

摘要： 提出一种颜色和边缘信息融合的背景建模方法。在像素级利用基于颜色特征的背景差方法，将像素分类为前景像素和背景像素。给出一种新的基于边缘特征的背景差方法，将边缘分类为前景边缘和背景边缘。将前景像素聚类为不同的目标区域，利用前景边缘信息识别出真实运动目标和虚假运动目标。实验表明，该方法可以有效地去除由于局部突然光照变化和背景静止目标的移动造成的虚假运动目标，提高运动目标检测的精确率。

关键词: 背景建模, 虚假运动目标去除, 基于边缘的背景差, 信息融合

CLC Number: 

• TP391