摘要
开展基于视差和尺度不变特征变换(SIFT)的双目视觉移动目标识别和追踪的研究。首先采用基于梯度的立体匹配算法得到较准确的左右视图视差映射,其次通过视差映射提高基于SIFT特征的左右视图运动目标的匹配精度,最后利用视差映射和区域增长的方法相结合分别在左右视图完成运动目标的追踪。实验结果表明,基于视差信息和SIFT的双目视觉移动目标识别与追踪算法具有很好的准确性,能够在连续视频中完成左右视场中对同一物体的追踪。
In this paper, we proposed a method for identifying and tracking moving objects in binocular vision based on disparity computing and Scale-Invariant Feature Transform （SIFT） . Firstly, we employ the gradient-based stereo matching algorithm to construct the disparity mapping between the left and the right views. Based on SIFT feature matching, the disparity mapping is applied to improve the object matching accuracy in binocular vision. The disparity mapping and regional growth are used to track moving objects in the left and right views respectively. Experimental results show that of the proposal has higher accuracy and can track the same object in the left and right views of a video.
出处
《计算机工程与科学》
CSCD
北大核心
2015年第10期1947-1951,共5页
Computer Engineering & Science
基金
湖北省自然科学基金资助项目(2015CFC782)
湖北省教育厅科研计划资助项目(Q20131904)
文化部科技提升计划资助项目(201307)
国家自然科学基金资助项目(61562025)
关键词
视差立体匹配
尺度不变特征变换
目标识别
物体追踪
区域增长
disparity stereo matching
scale-invariant feature transform （SIFT）
target recognition
object tracking
regional growth
