危化品堆垛双目测距中改进的SURF匹配算法研究

Research on improved SURF matching algorithm in binocular distance measurement for hazardous chemicals stacking

堆垛安全"5距"是保证危化品存储安全的重要因素,双目视觉测距是解决该问题的一个非常有前景的技术。本文针对危化品堆垛测距中双目立体匹配算法受货物特征点相似和光照影响的特点从而影响精度的关键问题,提出一种自适应邻域的改进SURF匹配算法,该算法对特征点进行位置约束,解决传统SURF算法容易受局部区域像素梯度方向影响的问题,大大提升了匹配效果。实验表明,该算法使得匹配准确率基本保持在98%,特征点数量相对较少时,准确率能达到100%。

The "5-distance" of stacking is an important factor to ensure the storage safety of hazardous chemicals. Binocular vision distance measurement method is a very promising technology to solve this problem. In this paper, in the view of features that binocular stereo matching algorithm is affected by similarity of goods features points and features of illumination influence in distance measurement of hazardous chemicals stacking, an improved SURF matching algorithm based on self-adaptive neighborhood is proposed, which constrains the location of feature points, solves the problem of traditional SURF algorithm is easily affected by pixel gradient directions in local areas, and greatly improves the matching effect. Experiments show this algorithm can maintain 98 % of matching accuracy rate and this rate can be 100 % when the number of feature points is relatively few.