摘要
针对计算机视觉系统在低负载运算平台下的应用,对双目立体视觉系统中的图像特征提取和多分辨率图像匹配处理算法进行分析与研究。在对已有的图像降晰算法进行实验分析的基础上,提出了适用于低负载运算平台的双目视觉快速匹配算法。研究结果表明:该算法与已有的图像特征点匹配算法相比,算法执行时间缩短了约80%。基本满足低负载运算平台如自主移动机器人、车载摄像机等的要求,降低了运算难度,实现了目标定位实时性,同时兼顾了图像匹配精度和高可靠性。
Aimed at the application of computer vision system on low-load calculation platform, the image feature extraction and image matching with multi-resolution ratio are analyzed and researched. On the basis of experimental analysis of the existing image dimension reduction algorithms, a fast image matching algorithm for binocular vision based on low-load calculation platform is presented. Research results show that compared with the existing image feature point matching algorithms, this algorithm＇ s execution time was shortened about 80 %. Basically meet the needs of low-load calculation platform such as autonomous mobile robot, in-car cameras, etc, and reduced the difficulty of computing to achieve real-time target detection and location. The image matching precision and high reliability is also taken into account.
出处
《传感器与微系统》
CSCD
北大核心
2010年第1期45-48,共4页
Transducer and Microsystem Technologies
关键词
双目视觉
图像降晰
特征点匹配
binocular vision
reducible picture
character points matching
