航拍视频运动目标检测与跟踪  

Moving Object Detection and Tracking Based on Aerial Video

作　　者：罗毅[1] 金立左[1] 

机构地区：[1]东南大学自动化学院,江苏南京210096

出　　处：《工业控制计算机》2017年第3期24-25,28,共3页Industrial Control Computer

摘　　要：提出一种针对航拍视频中的运动目标进行检测与跟踪的算法。首先,采用分块特征检测提取视频帧中的特征点,基于特征点匹配进行帧对齐以后利用帧间差分提取前景并进行时域增强。然后,使用MSER提取前景团块,采用团块匹配以及关联表实现运动目标跟踪。同时,处理团块在跟踪过程中的分裂与合并问题。实验结果表明,该算法能够对航拍视频中的运动目标进行准确地检测与跟踪。An algorithm for moving object detection and tracking based on aerial video is presented in this paper.Firstly,a sub-block feature detection method is applied to extract feature points in a frame.After the frame is aligned based on feature point matching,the foreground is extracted with frame differential method and enhanced in time domain.Then,the foreground blobs are extracted by MSER.The moving object tracking is realized by blob matching and association table.At the same time,a solution for blob splitting and merging is also provided in this paper.

关 键 词：运动目标检测 关联表 运动目标跟踪 分裂与合并 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]