基于图像的动目标跟踪技米研究与系统实现

发布时间：2018-03-19 13:51

本文选题：目标跟踪　切入点：Cam　出处：《西南科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：伴随着视频监控智能化的步伐,目标跟踪等图像处理技术逐步与普通应用场景对接,对跟踪算法的综合性能提出了新的需求。本文以相关项目为背景,在现有网络监控系统的模式上,结合智能图像分析算法和云台控制技术,以实现对监控画面内移动物体的检测和锁定跟踪功能。文中首先对基于图像的动目标检测和跟踪技术及相关原理进行了分析,结合实际需求,选取简单、高效的Cam Shift跟踪算法进行深入研究。针对Cam Shift算法环境依赖性强和容易出现跟踪漂移等问题,提出了改进思路。通过引入Kalman滤波、局部差分方法和最优特征的思想,实现了跟踪效果的明显优化。实验证明改进后算法在增加较少计算量的情况下减少跟踪过程中出现的不稳定现象,提高算法的准确性和鲁棒性。在此基础上根据项目需求,系统采用嵌入式平台和云台摄像头的结构,在保障视频采集、压缩和传输功能稳定的前提下,采用多线程编程方式,完成了目标检测和跟踪程序的设计。最后通过实际验证和分析证明了系统的可用性和跟踪算法的可移植性,完成了设计目的。同市场上现有跟踪摄像头相比,该系统具有明显的技术优势,因此也具有更稳定的跟踪效果和应用价值。
[Abstract]:With the intelligent pace of video surveillance, image processing technology such as target tracking is gradually docking with the common application scene, which puts forward new requirements for the comprehensive performance of tracking algorithm. In the mode of the existing network monitoring system, combining the intelligent image analysis algorithm and the cloud head control technology, In order to realize the detection and locking tracking function of moving objects in the monitoring screen, this paper first analyzes the moving target detection and tracking technology based on image and related principles, combining with the actual demand, the selection is simple, The efficient Cam Shift tracking algorithm is studied in depth. Aiming at the problems of strong environmental dependence and easy to occur tracking drift in Cam Shift algorithm, an improved approach is put forward. By introducing Kalman filter, local difference method and the idea of optimal feature, the author proposes a new method, which is based on the following methods: the Kalman filter, the local difference method and the optimal. The experimental results show that the improved algorithm can reduce the instability in the tracking process and improve the accuracy and robustness of the algorithm. The system adopts the structure of embedded platform and cloud head camera. Under the premise of ensuring the stable function of video collection, compression and transmission, the system adopts multi-thread programming method. Finally, the usability of the system and the portability of the tracking algorithm are proved by the actual verification and analysis. The purpose of the design is accomplished. Compared with the existing tracking camera in the market, The system has obvious technical advantages, so it also has more stable tracking effect and application value.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】