基于背景分类的运动目标检测算法

发布时间：2018-07-17 14:59

【摘要】：针对光线暗、对比度和分辨率低的监控视频,提出了一种基于背景分类的运动目标检测算法。首先用视频第一帧图像HSV空间的色度H和亮度V作为背景特征进行初始化,建立两种包含色度和亮度特征的背景模型类,即初始化得到的原始背景类和受光照或者其他因素影响得到的在原始背景周围波动的背景波动类,利用这两个背景模型进行前景检测和背景更新。为提高前景检测的准确率,背景模型的更正加入背景更正机制和权重机制,使得背景中样本的数量根据背景的实际情况处在一种动态的变化中,提高前景分割的效率。用不同场景下的监控视频进行算法对比实验,结果证明,该算法获得的前景完整清晰,视频处理的速度较快。提出的算法简单实用,对噪声干扰表现出良好的鲁棒性。
[Abstract]:A moving target detection algorithm based on background classification is proposed for surveillance video with dark light, low contrast and low resolution. Firstly, the chrominance H and luminance V of HSV space of the first frame of video are initialized as background features, and two kinds of background model classes including chrominance and luminance are established. That is initializing the original background class and background fluctuation class affected by illumination or other factors. The two background models are used for foreground detection and background updating. In order to improve the accuracy of foreground detection, the correction of background model adds background correction mechanism and weight mechanism, which makes the number of samples in the background in a dynamic change according to the actual situation of background, and improves the efficiency of foreground segmentation. The results show that the foreground of the algorithm is complete and clear, and the speed of video processing is fast. The proposed algorithm is simple and practical and has good robustness to noise interference.
【作者单位】：山东建筑大学信息与电气工程学院;山东省智能建筑技术重点实验室;
【基金】：国家自然科学基金(No.61303087) 山东省科技发展计划(No.2013 GGX10131) 济南市高校院所自主创新项目(No.201202002)
【分类号】：TP391.41

【相似文献】