摘要
针对行星车视觉导航系统,提出一种立体匹配算法。使用Census变换得到一组二进制码流,通过色彩权重和距离权重完成权重计算,采用双通道的方法累加匹配代价。实验结果表明,该算法在时钟频率为100 MHz的情况下,每秒可处理2幅23帧左右的图像,平均误匹配率低于6%。
Aiming at the application of planetary rover vision system, this paper proposes a stereo matching algorithm. It uses Census transformation to get a group of binary code flow, gets through the colour weight and distance weight to complete weight calculating, and uses the method of double channel to accumulate the matching the price. Experimental results show that this algorithm can process 23 frames stereo images per second at 100 MHz, and the average error rate is lower than 6%.
出处
《计算机工程》
CAS
CSCD
2012年第3期159-162,共4页
Computer Engineering
基金
国家自然科学基金资助项目(60805030)
关键词
行星车
视觉导航
立体匹配
自适应权重
Census变换
planetary rover
vision navigation
stereo matching
adaptive weight
Census transformation