基于滑动窗的混合高斯模型运动目标检测方法 a moving object detection method based on sliding window gaussian mixture model.pdfVIP

第35 卷第7 期 电 子 与 信 息 学 报 Vol.35No.7 2013 年7 月 Journal of Electronics Information Technology Jul. 2013 基于滑动窗的混合高斯模型运动目标检测方法 周建英 吴小培* 张 超 吕 钊 (安徽大学计算智能与信号处理教育部重点实验室 合肥 230039) 摘 要：在复杂场景下，传统混合高斯模型能较好地检测出运动目标，但随着时间的推移，模型参数收敛缓慢且难 以适应场景中真实背景的实时变化，从而导致运动目标的错误检测率增加。该文利用滑动窗技术的短时历史记忆特 性，提出一种新颖的基于滑动窗的混合高斯模型运动目标检测方法，该方法弥补了传统混合高斯背景模型不能及时 形成新背景的缺点，提高了运动检测的完整性，并进一步降低了算法对场景光照变化的敏感性。多场景下的对比实 验结果表明，该方法能更准确、完整地检测出运动目标并具有更好的环境适应性。 关键词：运动目标检测；滑动窗；混合高斯模型；背景模型 中图分类号：TP391.41 文献标识码：A 文章编号：1009-5896(2013)07-1650-07 DOI: 10.3724/SP.J.1146.2012.01449 A Moving Object Detection Method Based on Sliding Window Gaussian Mixture Model Zhou Jian-ying Wu Xiao-pei Zhang Chao Lü Zhao (Key Laboratory of Intelligent Computing and Signal Processing Anhui University, Hefei 230039, China) Abstract: In complex scenes, the traditional Gaussian Mixture Model (GMM) algorithm is an effective way to extract moving objects. However, after a period of modeling, the model’s updating speed is difficult to keep up with the changes of true background. Consequently, false alarm rate will be increased in moving object detection. This paper proposed a new moving object detection method, which utilizes sliding window technology to retain short-term historical information. To a certain extent, it remedies that the traditional mixture Gaussian background model can not form new background in time, and it improves the integrality of motion detection. Furthermore, it reduces the algorithm sensitivity for the scene illumination change. The experimental results show that the proposed algorithm can more accurately, perfectly detect moving targets, and adapts rapidly to variations in the environment. Key words: Moving object