基于聚类的直线特征提取算法仿真研究

Research on Clustering-Based Linear Feature Extraction Algorithm Simulation

针对传统的Hough变换和链码法存在的计算效率低和漏检、误检等问题,提出新的直线特征提取算法.该算法在Canny边缘的基础上,使用一个聚类算法将边缘分类,缩小了检测直线的范围;进一步剔除聚类结果中的伪直线;最后使用改进的Hough变换,较快地将直线提取出来.实验结果表明:通过对无背景干扰和有背景干扰的建筑图像,将本文方法和Hough变换和相位编组法进行对比实验,验证了所提方法的有效性.

We give an analysis of the traditional Hough transformation and code methods which are found to be of low computational efficiency and have defects of undetected and false alarms. Therefore, a new method called linear feature extraction algorithm is proposed. According to this method, firstly, edges are detected based on Canny detector. Secondly, an algorithm which follows image clustering categorizes the edges and reduces the range of linear detection. Thirdly, the straight line - identifying rule is used for eliminating false straight lines among the clustered images. Lastly, an improved algorithm of Hough Transformation renders a quick extraction of straight lines. Experimental results show that the proposed algorithm is valid based on a comparison between the proposed method, the Hough transformation and phase in building images with non -background and background interference.