【摘要】 道路损害的自动评估是计算机视觉研究领域的一个难点,因为各种道路的颜色、纹理和光照各不相同以及数据采集的时间不同,所以这些对于评估结果影响特别大。虽然路面损害边缘检测的方法对于光照、纹理等还是相对敏感,但比图像分割方法受到的影响要小得多,如果对于原图像进行多次滤波处理,能有效地增强路面损害,提高自动评估的精度。在传统的道路损害评估方法的基础上,提出了四叉树分离和合并的方法,进行一系列处理后对图像边缘检测、骨骼化和四叉树分离和合并,然后得到评估的结果。并对不同类型的损害进行了实验,实验结果表明,该算法不仅仅能应用于裂缝类的路面损害,同样能应用于龟裂或块裂类的路面损害,并且评估的精度相对较高。更多还原

【Abstract】 Automatic pavement deterioration evaluation is difficult in the field of computer vision study,because various pavements has different colours,textures and light,as well as the time the data are collected,these result in great impacts on the evaluation outcomes.Though the edge detection method for pavement deterioration evaluation is also relatively sensitive to light,texture and so on,but such impact is much smaller than on the evaluation with image segmentation method.If a number of filtering processing is performed on primary images,the accuracy of automatic evaluation of pavement deterioration can be effectively improved.In this paper,we introduce an effective algorithm based on quadtree decomposition and merge method based on traditional pavement deterioration evaluation method.The evaluated outcomes can be attained by performing image edge detection,skeleton and quadtree decomposition and merge after a series of images preprocessing.Experiments of different kinds of deterioration types have been made,the results shown that our algorithm is able to be applied to crack-like deterioration of pavement,and to chap-like and block split ones as well,its evaluation accuracy is relatively high too.更多还原