基于自适应阈值区间的动态采样Hough变换直线检测算法  

An Algorithm of Hough Transform Line Detection Based on the Adaptive Threshold Interval Halves Sample

作　　者：宋晓宇[1] 郭寒冰[1] 袁帅[1] 马俊俊[1] 

机构地区：[1]沈阳建筑大学信息与控制工程学院,辽宁沈阳110168

出　　处：《沈阳建筑大学学报（自然科学版）》2014年第5期945-952,共8页Journal of Shenyang Jianzhu University：Natural Science

基　　金：国家自然科学基金项目(61305125);辽宁省教育厅基金项目(L2010450);辽宁省自然科学基金项目(201302816)

摘　　要：目的为解决Hough直线检测方法计算量大、参数空间中峰值点被次峰值点包围导致漏检或误检、阈值难以设定等问题.方法通过对Hough变换原理以及直线表示的特点进行分析,采用阈值区间表示一条直线,通过三个约束定义了该阈值区间,并根据其中第三条约束识别图像中其他像素点是否在该直线上.对被识别的轮廓线采用邻接关系将像素点聚集为不同类,将其依次存放到数组中,根据数组下标对这些像素点进行动态采样,对采样位置上的m个相邻像素点使用阈值区间进行直线判定与轮廓线上的像素点识别.针对GREC2003标准测试集进行实验验证.结果本算法与已有改进算法、Matlab内置优化Hough变换函数进行对比,实验结果验证了本算法的有效性.结论本算法通过过滤噪声信息,提高了Hough变换直线检测的抗干扰能力,增强了算法的鲁棒性.This paper proposed an algorithm for Hough line detection to solve problems such as large amount of calculation, false or missing detection because of first peak surrounded by second peak in the parameter space, and setting the threshold. Firstly, the Hough transform and the charac- teristic of the straight line representation were analized, and the threshold interval was proposed for identifying a straight line. This threshold interval was defined by using the three constraints, and the third constraint can be used to justify whether other pixels are on the line. Secondly, the pro- cessed contour lines were identified into different classes by using adjacency relationship and se- quentially stored in an array. According to the array subscript, the pixels were sampled by dynamicsplit half method, and the threshold interval was used to make line justification of m dacent pixels around the sampling location, and then identification of the pixels on the contour line was per- formed. Based on the GREC 2003, some tests were carried out to prove the algorithm. Results show that the proposed algorithm improves the anti-jamming of Hough transform based line detec- tion, and increases its robustness through filtering noise points. Calculated results are compared with the well known improved algorithm based on RHT and the optimal Hough transform built in Matlab, which illustrate the efficiency of the proposed algorithm.

关 键 词：HOUGH变换 阈值区间 动态采样 直线检测 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]