【作者】 吴晓阳；

【导师】 章专；

【作者基本信息】 浙江大学 ， 电路与系统， 2008， 硕士

【摘要】 计算机视觉和数字图像处理技术可以广泛地应用于工业、医疗保健、航空航天、军事等各个领域,其中针对视频连续图像中运动物体的分析是其中应用前景最为广泛的一个方向,在机器人导航、智能视觉监控系统、医学图像分析、工业检测、视频图像分析以及军事雷达视频信号的处理上都有占有重要地位。其中基于视频图像的运动目标分析也是计算机视觉和数字图像处理技术最为复杂的一个方向。它是一个庞大的工程,既要对视频信号进行采集,又要对采集的图像进行处理,还要针对具体的应用编写程序进行开发。对于要完成某项应用的工程人员来说,如果所有底层的算法都要自己编码实现,所有的图像处理函数都要从头编写,既造成时间和精力上的浪费,又难以保证稳定性、实用性和通用性。OpcnCV(Open Source Computer Vision Library)是一种用于数字图像处理和计算机视觉的函数库,由Intel微处理器研究实验室(Intel’s MicroprocessorResearch Lab)的视觉交互组(The Visual Interactivity Group)开发。采用的开发语言是C++,可以在Windows系统及Linux系统下使用,该函数库是开放源代码的,能够从Intel公司的网站免费下载得到。OpenCV提供了针对各种形式的图像和视频源文件(如:bitmap图像,video文件和实时摄像机)的帧提取函数和很多标准的图像处理算法,这些函数都可以直接用在具体的视频程序开发项目中。针对在复杂背景中检测出多批特定运动目标并实施分配批号实行标记跟踪,本文利用OpenCV的运动物体跟踪的数据结构、函数以及基本框架,建立了一个由人机交互界面模块;运动物体的前景检测模块;运动物体的团块特征检测模块;运动物体的团块跟踪模块轨迹生成模块;轨迹后处理模块组成的视频图像运动目标分析系统。在本文的最后阶段,使用此视频图像运动目标分析系统进行了大量实验,并全面分析了实验现象和数据。通过这些现象和数据可以得出结论:本文基于OpenCV设计的视频图像运动目标分析系统具有良好的实时性,能够正确的进行运动物体的实时检测和跟踪,并具有良好的鲁棒性。由于该系统在Windows下开发,如何将该系统移植到其他的系统或者是嵌入式平台并进一步提高系统的通用性和鲁棒性就成了今后研究工作的重点。更多还原

【Abstract】 Detection and tracking of moving object in image sequence is one important area in the domain of digital image processing and computer vision. It’s applications include fields like robot navigation, vision-based supervision, ecurity surveillance, medical image analysis, industrial automation, video processing, etc...Detection and tracking of moving object in image sequence is also the most complicated direction of computer vision and digital image processing techniques. It is a ample project that including video collection, digital image processing and application development.It’s too complex for a engineer to accomplish all these works, at the same time the stability, practicability and general availability is difficult to guaranteed.OpenCV(Open Source Computer Vision Library) is a library for digital image processing and computer vision. It is developed by the visual interactivity group in Intel’s microprocessor research Lab with C++ language. It can be used in window system and Linux system. The library is open source and can be free download from Intel’s Web site. OpenCV provides various functions to extract frame from image sequence and video source ( such as: bitmap images, video files and real-time video camera ) and many of the standard image processing algorithms, these functions can be directly used in the video program for specific development projects.Aim at detecting, tracking and marking multipule specific targets in complex background. We use the basic framework of object tracking of OpenCV to establish a multi-modules system. The system is composed of Human-computer interaction interface module, foreground of moving objects detection module, blocks feature of moving objects detection module, blocks tracking module and track analysis module.We use this system doing a lot of experiment under various contexts, and tht experiments results are well wxamined. The results demonstrate that the system can do automatic motion detection and tracking with high real-time performance and moderate robustness. Since we develop this system under windows platform, how to migrate to other systems or embedded platform and further improve the versatility and robustness has become the focus of future research work.更多还原