摘要： 现有算法精度低，且线段被完全遮挡时会出现误匹配甚至无法匹配等情况，为此提出改进的直线匹配算法。该算法在 匹配阶段改进搜索方法，通过删除缺失目标直线对应的模板图像直线，建立新的模板图像轮廓再进行匹配，解决直线段缺失时无法匹配问题；在 匹配阶段引入RANSAC算法剔除误匹配直线对，解决误匹配问题。实验表明，该方法能解决目标直线被遮挡问题，并提高匹配精度。

关键词: Hough变换, 目标检测, RANSAC算法, sobel算子, 最小二乘

Abstract: As to the previous algorithms, the accuracy of matching is low and when the line segments are occupied, the matching is failure or even not to be implemented. Focusing on these instances, this paper presents a modified method. In matching, a new search method establishes a new contour of the model image via removing its lines corresponding to the missing lines, so it solves the problem that the procedure of matching can not be implemented when several line segments in model image are absence; and in matching, the Random Sample Consensus(RANSAC) algorithm is introduced to settle the problem of mismatch by removing the pairs of mismatching lines. Experiments demonstrate that the proposed algorithm can settle the problem brought by occupied lines, and promote the accuracy of the matching.

Key words: Hough transform, object detection, Random Sample Consensus(RANSAC) algorithm, sobel operation, least squares

中图分类号: 

• TP391