摘要： 高斯混合模型背景差法的难点在于对背景模型进行有效更新。针对该问题，提出一种高斯混合模型的自适应更新算法——HCGMM。通过量化帧间灰度直方图的差异，得到图像的亮度变化值，并依据亮度变化值对高斯混合模型参数进行调整。实验结果表明，即使在画面光强剧烈变化的情况下，该算法也能够准确地重构背景，避免过度检测现象，从而实现对运动目标的完整提取。

关键词: 直方图比较, 全局分析, 高斯混合模型, 运动目标检测, 光线突变

Abstract: An updating method called Histogram Comparison Gaussian Mixture Model(HCGMM) is proposed to reduce the affection of target detection in the presence of large illumination changes and background variations. To begin with, the method constructs the intensity histograms for each frame in the time window, and then, calculates the values of comparison between the two intensity histograms of images in the time window. Using the values just calculated, it updates the parameters of the Gaussian Mixture Models(GMM). Experimental result shows that the method can reconstruct background accurately and extract the target integrally in the presence of large illumination changes and background variations.

Key words: histogram comparison, global analysis, Gaussian Mixture Model(GMM), moving object detection, sudden change of light

中图分类号: 

• TP391.41