摘要
基于非刚性目标运动跟踪问题进行分析,针对运动过程中目标会出现旋转、比例变化及形变等情况,在复杂的背景中通过特征识别匹配实现跟踪。以运动目标的检测识别方法及特征识别方法为出发点,对目前几种经典方法进行分析、比较。结果表明:与光流法及背景差分法相比,MHI在非刚性目标运动的检测准确率、实时性以及鲁棒性方面均占优势;特征匹配技术SURF则可更好地平衡处理速度和效果性能。以SIFT使用不同尺度确定稳定特征点的思路为基础,提出多小波变换图像分析方法,用于实现对序列图像运动目标的特征匹配。
The analysis based on non-rigid object motion tracking. In a complex background by the feature,we identified the tracking match through the goal conditions of rotation,scaling and deformation changes in the campaign course. The starting point was to analyze the detection and identification method and feature recognition method of moving targets and to comparatively analyze on the current development of several classical methods. Compared with the optical flow method and background subtraction,MHI has advantages in non-rigid object motion detection accuracy,timeliness and robustness dominant. Feature matching technology,SURF can better balance processing speed and effectiveness of performance. Based on the idea of the SIFT of different scales to determine the stability feature points,we proposed multi-wavelet transform image analysis,which was used to implement the feature match of image sequences moving targets.
出处
《重庆理工大学学报(自然科学)》
CAS
2015年第4期81-85,共5页
Journal of Chongqing University of Technology:Natural Science
基金
重庆市教委科学技术研究项目(KJ120807)
关键词
非刚性目标
目标检测
运动跟踪
运动识别
多小波变换
non-rigid target
target detection
motion tracking
motion recognition
multiwavelet transform