摘要
传统的基于K-means聚类算法由于仅考虑裂缝像素值大小,导致了不均匀光照下的路面裂缝提取不精确。为提高不均匀光照下道路裂缝检测的效率和准确性,提出一种改进的K-means聚类算法与区域生长法结合的道路裂缝图像检测算法。该算法首先运用改进的K-means聚类算法对不均匀光照下的路面裂缝进行粗定位;然后采用区域生长法对粗定位的裂缝图像进行准确提取;最后采用形态学算法进行优化处理以提升裂缝分割效果。该算法把路面裂缝位置信息加入到聚类算法中,并结合裂缝的像素值来综合判断裂缝区域,从而克服了图像采集过程中由于光照不均或物体表面反光等原因造成的裂缝信息识读困难的问题。实验结果表明,与传统K-means算法相比,该方法对光照不均的路面裂缝提取在处理效果和性能方面均有明显提升,同时抑制了噪声的干扰,为路面裂缝参数的提取及病害程度的定性分析奠定了基础。
Traditional K-means clustering-based algorithm 0nly considers the size of pixel value of cracks, this leads to inaccurate extraction of pavement cracks under uneven illumination. In order to improve the efficiency and accuracy of road crack detection with uneven illumination, we propose in this paper a road crack image detection algorithm which combines the improved K-means clustering algorithm with the region growing method. The proposed algorithm first uses the improved K-means clustering algorithm to roughly locate the road cracks under uneven illumination; then employs the region growing method to precisely extract the crack images roughly located; at last it uses the morphological algorithm to carry out optimisation process to enhance the cracks segmentation result. The proposed algorithm adds the pavement cracks location information to clustering algorithm and estimates comprehensively the cracks region in combination with the pixel value of cracks, so that solves the problem of cracks information recognition difficulty caused by uneven illumination or light reflection on object surfaces in image acquisition process. Experimental results demonstrate that compared with traditional K-means algorithm, the proposed method has noticeable enhancement in both processing effect and performance on extracting the cracks of pavement with uneven illumination, and the noise interference is suppressed as well. It lays the foundation for the pavement cracks parameters extraction and flae quantitative analysis on the extent of road diseases.
出处
《计算机应用与软件》
CSCD
2015年第7期244-247,300,共5页
Computer Applications and Software
基金
国家自然科学基金项目(31170952)
上海教委科技创新重点项目(11ZZ143)
