基于局部PCA的参数约束Hough直线分层检测

针对经典Hough变换进行直线检测存在计算量和存储量大的问题,提出了基于局部PCA统计分析的分层Hough参数约束直线检测思路。首先对边缘图像进行局部PCA分析，得到边缘的主方向分布信息；然后利用自适应模糊映射统计出图像空间的方向直方图分布信息和约束直线的θ参数空间；最后通过构建局部参数空间的累积矩阵，检测出存在的直线，同时更新原来图像空间，不断重复直线检测过程，直到检测出规定的所有直线。实验表明，该算法计算、存储消耗减少，检测稳定性高。

Improved Hough transform hierarchical line detection method with parameter restriction based on local PCA

To overcome the difficulty in straight line detection that existing huge computing burden and storage space in general Hough transform (HT), a novel scheme which combines local principal component analysis parameter restrained with hierarchical line detection is proposed. The local principal component analysis is performed firstly for the edge image, so the edge orientation distribution information is obtained. Based on the adaptive fuzzy mapping of the edge orientation histogram distribution information, then the range of the line HT parameter is limited in a small scope. Finally the straight line in the image space is detected by using local HT parameter accumulator matrix, and the detected result is introduced to update the image space information.These steps are repeated until all the straight lines are detected over. Simulation results show that the proposed algorithm is efficient, and its robustness is better than the general HT method.