基于矩阵低秩稀疏分解的运动目标检测

发布时间：2018-02-27 19:29

本文关键词： 运动目标检测矩阵低秩稀疏分解边缘检测帧间差分　出处：《西北大学》2017年硕士论文　论文类型：学位论文

【摘要】：运动目标检测是一种将一个或者多个运动物体从视频中准确提取出来的技术,是图像处理和计算机视觉学科中的一个重要分支。运动目标检测的准确性直接影响后续运动目标追踪、运动目标识别以及运动目标行为分析等的结果。目前,运动目标检测在交通监控、天气预报以及卫星云图分析等方面有着广泛的应用。因此,对运动目标检测技术进行研究有重要的理论研究意义和应用价值。常用的运动目标检测方法有:帧间差分法、背景减除法以及光流法。近年来提出的矩阵低秩稀疏分解理论引起了相关学者的高度关注,并将其应用于运动目标检测。本文围绕基于矩阵低秩稀疏分解的视频运动目标检测展开研究,以提高运动目标检测的准确率和算法效率。创新性工作主要体现在:提出了一种改进的基于矩阵低秩稀疏分解的视频运动目标检测方法,更进一步综合考虑前后视频帧之间的差别,结合边缘提取和帧间差分法对检测到的运动目标边缘进行补偿,提高运动目标检测的准确率。本文研究工作主要包括以下几个方面:1.研究和学习了几种运动目标检测方法:帧间差分法、光流法、背景减除法和矩阵低秩稀疏分解法,并数值实现了部分算法,给出相应的实验结果与分析。2.在研究矩阵低秩稀疏分解理论及其求解算法的基础上,提出了一种改进的基于矩阵低秩稀疏分解的运动目标检测方法。该方法采用对视频等间隔隔帧提取,并对矩阵稀疏分解模型中的稀疏项加以约束,提高运动目标检测的准确率和算法实现效率,实验结果表明该方法能够准确地提取运动目标,是一种行之有效的方法。3.考虑边缘是图像中最重要的信息,结合帧间差分法的优势,在改进的矩阵低秩稀疏分解方法的基础上,融合边缘与帧间信息,对检测出的运动目标边缘补偿,提高检测目标的准确率,并给出相应的实验结果及分析。总之,围绕运动目标检测问题,本文提出并实现了改进的基于矩阵低秩稀疏分解的运动目标检测方法,进一步融合边缘与帧差信息对检测的运动目标边缘进行补偿。该方法不仅能够以较高的算法效率准确地提取运动目标,而且占用较少的运行内存。
[Abstract]:Moving target detection is a technique that accurately extracts one or more moving objects from a video. It is an important branch of the subject of image processing and computer vision. The accuracy of moving target detection directly affects the results of subsequent moving target tracking, moving target recognition and behavior analysis of moving object. Moving target detection is widely used in traffic monitoring, weather forecast and satellite cloud image analysis. The research of moving target detection technology has important theoretical research significance and application value. Background subtraction method and optical flow method. The theory of low rank sparse decomposition of matrix proposed in recent years has attracted great attention of scholars. This paper focuses on video moving target detection based on low rank sparse decomposition of matrix. In order to improve the accuracy and efficiency of moving target detection, the innovative work is as follows: an improved video moving target detection method based on matrix low rank sparse decomposition is proposed. Furthermore, considering the difference between video frames, the edge of the detected moving object is compensated by edge extraction and inter-frame differential method. To improve the accuracy of moving target detection, this paper mainly includes the following aspects: 1. Research and study of several moving target detection methods: inter-frame difference method, optical flow method, background subtraction method and matrix low-rank sparse decomposition method. Numerical results and analysis of some algorithms are given. 2. On the basis of studying the theory of matrix low rank sparse decomposition and its solving algorithm, An improved method of moving target detection based on low rank sparse decomposition of matrix is proposed, in which the frame interval of video is extracted at equal intervals, and the sparse items in the sparse decomposition model of matrix are constrained. The accuracy of moving target detection and the efficiency of the algorithm are improved. The experimental results show that this method can extract moving target accurately and is an effective method. Considering edge is the most important information in the image. Combining the advantages of the inter-frame difference method and the improved matrix low-rank sparse decomposition method, the edge and inter-frame information are fused to compensate the edge of the detected moving object, and the accuracy of the detection target is improved. The corresponding experimental results and analysis are given. In a word, an improved method of moving target detection based on matrix low rank sparse decomposition is proposed and implemented around the problem of moving target detection. The edge of the detected moving object can be compensated by further fusion of edge and frame difference information. This method can not only extract the moving object accurately but also occupy less running memory.
【学位授予单位】：西北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】