摘要
根据立体视觉的原理,设计实现了基于光学体视显微镜图像的立体测量系统。从图像采集、摄像机标定、立体匹配到深度图生成,建立了完整的系统。对系统中的摄像机模型进行了线性标定,达到了亚像素的标定精度。在立体匹配中实现了三种不同的算法:改进的区域相关性匹配算法、基于边界特征的概率松弛迭代匹配方法、新序数测度法(New Ordinal Measure),并对结果的精度进行了分析和比较。
We have designed and constructed a complete stereoscopic system Based on stereoscopy theory using stereo light microscope which includes image grabbing, camera calibration, stereo matching and creating of depth image. The camera is calibrated in the system using linear model and the precision reaches subpixel level. we implement three different algorithms. In the stereo matching that improved area-based matching, iterated relaxation corresponding algorithm based on edge feature, new ordinal measure. By comparing between them, we obtain the conclusion that the new ordinal measure is the most robust.
出处
《中国体视学与图像分析》
2003年第4期220-224,共5页
Chinese Journal of Stereology and Image Analysis
基金
"Motic-北航图像技术研发中心"研发资金资助