基于改进的PBAS算法的前景目标检测

发布时间：2018-08-01 11:14

【摘要】：针对像素层自适应分割算法(Pixel Based Adaptive Segmenter,PBAS)在动态背景下检测准确率低、静止或运动缓慢的前景目标被更新为背景以及出现鬼影干扰的问题,提出了一种结合像素级信息和区域级信息的改进的前景检测算法。首先,提出一种融合区域结构信息和区域颜色信息的背景复杂度衡量方式;然后,采用改进的背景复杂度来控制判定阈值和学习率,并检测前景;其次,对像素层的检测结果使用区域窗口进行空间邻域对比,以消除鬼影;最后,引入前景计数机制来保证静止前景不被更新为背景。实验结果表明,该算法对光照条件和前景运动速度不敏感,能有效地从背景中检测出完整的前景目标,并迅速地消除鬼影干扰,准确率达到了92.7%。
[Abstract]:The pixel layer adaptive segmentation algorithm, (Pixel Based Adaptive Segment Center, has low detection accuracy in dynamic background, static or slow-moving foreground targets are updated to background, and ghost interference occurs. An improved foreground detection algorithm combining pixel level information and region level information is proposed. Firstly, a method to measure the background complexity of the region structure information and the region color information is proposed. Then, the improved background complexity is used to control the decision threshold and the learning rate, and to detect the foreground. The detection results of pixel layer are compared with the region window in order to eliminate ghost. Finally, the foreground counting mechanism is introduced to ensure that the static foreground is not updated to the background. The experimental results show that the algorithm is insensitive to illumination conditions and foreground motion speed, and can effectively detect the complete foreground target from the background, and eliminate the ghost interference quickly, and the accuracy reaches 92.775%.
【作者单位】：武汉理工大学信息工程学院;内河航运技术湖北省重点实验室;
【基金】：内河航运技术湖北省重点实验室基金(NHHY2014004) 国家自然科学基金(51579204)资助
【分类号】：TP391.41

【相似文献】