视频图像序列中运动目标自动跟踪及其应用研究

发布时间：2018-05-14 01:35

本文选题：视频图像序列 + 边缘检测　；参考：《沈阳理工大学》2017年硕士论文

【摘要】：视频图像序列中的运动目标跟踪一直是计算机领域中的热点问题,在动态场景中运动目标的检测和跟踪技术通常可以利用在视频监控、人机交互、汽车辅助驾驶、运动行为分析等方面。在实际的生活生产过程之中,目标跟踪技术仍然是计算机视觉中富有挑战性的工作。基于视频图像序列中运动目标的特征和其跟踪方法和应用,本文作出工作如下:(1)文中提出一种新的边缘检测方法用来改善图像边缘检测噪声点过多的问题。图像的边缘被用来进行图像分割、模版匹配和图像识别等方面的探讨和研究,是图像最基本的特征和构成因素。在图像的边缘检测内容中,由于检测图像边缘的方法式多样化的,但其中适用性最广最为快速的方法还是基于模糊梯度算法中的边缘检测方式。文中对现有的模糊梯度算法进行评估和讨论,希望能够解决掉其中存在的缺点问题并且对之进行进一步的完善,最终得到一种更新型更好的图像边缘检测方法。(2)在基于现有的运动目标跟踪方法中,对于图像的复杂背景有噪声干扰的情况下,改善帧间相减法算法来是目标跟踪方法更加完善。很多物体都会随着时间而进行运动,在实际生活中,不同的运动物体对于不同的群体传达着非常不一样的视觉信息,而这些视觉信息又会给群体们带来非常重要的现实意义,人们用视觉所捕捉到的信息往往会对其有实际意义与使用价值。在研究基于视频图像的运动目标的检测与追踪具有着重大意义,目前实验中已经实现了对运动物体的检测与追踪,在探究理论的过程中,发现帧间差分法在获取运动目标并跟踪的这种形式是最为简便和完整可靠的一种。在运动目标追踪检测方面,运用的复杂度的分块搜索法来进行计算,能够将其进行仿真模拟实验与编程实现。(3)针对图像在一些光暗程度变化较大时和图像的背景复杂的情况下本文改善两种图像清晰化算法,以及和原有的卡尔曼滤波方法进行比对。在公共环境的安全和防护成为人们关注的热点问题同时,而视频监控即为最常用的公共安全保护措施,在马路、银行、医院、学校等公共区域,视频监控技术被广泛应用与生产生活之中。虽然高清摄像头已经逐渐开始普及,在图像受到一定的光暗程度变化干扰和复杂背景时会使原有的算法没有明显效果,本文使用维纳滤波算法和同态滤波算法来进行图像的清晰化工作。
[Abstract]:Moving target tracking in video sequence has always been a hot issue in the computer field. Detection and tracking techniques of moving targets in dynamic scenes are usually used in video surveillance, human-computer interaction, vehicle driving, and motion behavior analysis. In the actual production process, target tracking technology is still the same. A challenging job in computer vision. Based on the features of the moving target in the video sequence and its tracking methods and applications, the following work is made as follows: (1) a new edge detection method is proposed to improve the problem of excessive noise points in the image edge detection. The edge of the image image is used for image segmentation and template matching. Research and Research on image recognition are the most basic features and components of images. In the content of image edge detection, the method of detecting the edge of the image is diversified, but the most widely and most fast method is based on the edge detection method in the fuzzy gradient algorithm. The algorithm is evaluated and discussed. We hope to solve the shortcomings of the existing problems and further improve it, and finally get a new and better image edge detection method. (2) in the current moving target tracking method, the interframe subtraction is improved for the complex background of the image with noise interference. The algorithm is that the target tracking method is more perfect. Many objects will move along with time. In real life, different moving objects convey very different visual information to different groups, and these visual information will bring very important realistic meaning to the group, people use the information captured by the vision. It is of great significance for the detection and tracking of moving objects based on video images. At present, the detection and tracking of moving objects have been realized in the experiment. In the process of exploring the theory, it is found that the form of inter frame difference method in obtaining moving targets and tracking is the most simple form. It is a complete and reliable one. In the field of tracking and detection of moving targets, the complexity of the block search method is used to carry out the calculation. (3) to improve the image sharpening algorithm, two kinds of image sharpening algorithms are improved in this paper, when the light and dark degree of the image is varied and the background of the image is complex. Compared with the original Calman filtering method, the security and protection of the public environment have become the hot issues of attention, and video surveillance is the most common public security protection measures, in the public areas such as roads, banks, hospitals, schools and other public areas. Video surveillance technology is widely used and in production life. It has gradually become popular. When the image is disturbed by a certain degree of light and dark, the original algorithm has no obvious effect. In this paper, the Wiener filtering algorithm and homomorphic filtering algorithm are used to make the image clear.

【学位授予单位】：沈阳理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】