改进的SAD-Census立体匹配算法研究

Investigation of Advanced SAD-Census Stereo matching algorithm

在实际的三维工件测量环境中,外界光照、噪声均会对立体匹配结果产生巨大影响,从而影响最终三维重构的结果。本文提出了一种基于自适应权重的SAD-Census立体匹配算法,首先对Census立体匹配算法进行了改进,然后将代价聚合中构建的十字代价聚合区域最小臂长与空间阈值和颜色阈值的比值作为权重,融合SAD立体匹配算法和改进后的Census立体匹配算法计算视差图。最后通过对Middlebury立体匹配测试平台提供的标准图像对进行测试。测试结果表明,改进后的Census立体匹配算法具有更好的鲁棒性、在弱纹理区域匹配精度更高,基于自适应权重的SAD-Census可以得到准确的视差图。

In the actual 3D workpiece measurement environment, the external light and noise will have a great impact on the stereo matching results, thus affecting the final 3D reconstruction results. In this paper, a SAD-Census stereo matching algorithm based on adaptive weight is proposed. Firstly, the Census stereo matching algorithm is improved.Then, the ratio of the minimum arm length of the cross cost aggregation region constructed in the cost aggregation to the spatial threshold and color threshold is taken as the weight. The SAD stereo matching algorithm and the improved Census stereo matching algorithm are used to calculate the parallax map. Finally, the standard image pairs provided by Middlebury stereo matching test platform were tested. The test results show that the improved Census stereo matching algorithm has better robustness and higher matching accuracy in the weak texture area, and the accurate parallax map can be obtained by SAD-Census based on adaptive weight.