摘要： 为防止运动阴影在视频图像序列中被错误地检测为目标，必须提高阴影检测算法的准确性和普适性。为此，从独立分量分析(ICA)的原理及其特性出发，提出一种基于空间变换技术的运动阴影检测算法。该算法通过对视频序列建立高斯混合背景模型产生自适应背景，利用ICA技术对其进行空间变换提取特征，再通过背景与当前帧图像对应像素点在特征空间的位置特征来分类运动阴影与前景目标。实验结果表明该方法能够较好地抑制噪声，减少光照变化的影响，准确地检测出阴影。

关键词: 阴影检测, 独立分量分析, 特征空间, 高斯混合模型

Abstract: Moving cast shadows generally exist in video sequence. To prevent moving shadows being misclassified as moving objects or parts of moving objects, it is critical to improve the accuracy of shadows detection algorithm. On the basis of the theory and principle of Independent Component Analysis(ICA), this paper presents an algorithm for detecting moving cast shadows based on ICA transformation. An adaptive background is generated by building Gaussian mixture background model, and it extracts the feature using ICA-based transformation. The moving cast shadows are detected utilizing the space coordinate of shadow and object pixels. Experimental results show that the proposed algorithm is robust to noise, relieves influence of illumination change and can detect moving shadows correctly.

Key words: shadow detection, Independent Component Analysis(ICA), feature space, Gaussian Mixture Model(GMM)

中图分类号: 

• TN911.73