一种基于倾斜聚集窗口的高效双目匹配算法研究

An efficient binocular matching algorithm research based on tilt aggregation window

目前双目匹配算法多采用正面平行的聚集窗口假设(Frontal-Parallel Assumption),但是这种假设不符合实际场景中的倾斜平面.针对实际场景多为倾斜平面的问题,提出一种基于交叉倾斜聚集窗口的局部匹配算法,将实际场景中的斜面和视差空间中的视差平面建立一种数学模型,该算法能够快速找到视差空间中聚集窗口的视差平面信息,通过在KITTI公共评估网站上进行测试,实验结果表明提出的算法能够很好地提高对场景中倾斜平面的匹配精度且有较高的执行效率.

Now many matching algorithms have made the frontal-parallel assumption, but this assumption does not conform to the actual scene of tilt surfaces.In order to address this issue, a mathematical model is illus- trated to conclude that tilt surfaces in the environment have corresponding slanted disparity surfaces in the DSI （ Disparity Space Image） , which helps to find the proper disparity plane parameters of slanted support window.An improved algorithm based on cross aggregation is proposed.The algorithm is evaluated using the Toyota Technical Institute at Chicago （KITTI） benchmarks.The results show that the proposed algorithm can efficiently improve the tilt surfaces in the scene matching precision and execution efficiency.