基于卡方检验的SAR图像道路检测算法

摘要

To solve the threshold optimization problem in road detection algorithm of SAR effectively, a novel algorithm which need not fixed threshold is proposed. Firstly, a group of edge detection results are computed using combinations of the Tupin de-tector's parameters. And edge correspondence maps are obtained with pixel by pixel's statistics. Then correspondence threshold is optimized by chi-square test which act as similarity measurement between edge correspondence maps and edge detection results. Finally, "ideal" road edge detection result is obtained using optimal correspondence threshold. The experimental result shows that the proposed method possesses openness, object suitability and high automation, and can assist in practical implementations.%为有效解决SAR图像道路检测算法中阈值优化的问题,提出了一种不需要固定阈值的检测方法.该方法由不同阈值组合的Tupin边缘检测算法得到边缘检测结果图集,并逐像素统计得到边缘密度图；引入卡方检验计算边缘密度图与边缘检测结果图集之间的相似性度量参数,优化得到最佳边缘相关等级；根据最佳等级得到最终的道路边缘检测结果.实验结果表明,该方法有较强的开放性与目标适应性,自动化程度高,有很强的工程实用性.