论文部分内容阅读
目的道路裂缝的等级评定是公路养护的基本任务之一,目前国内相关部门主要通过线阵列相机采集道路影像,由于道路影像裂缝的识别会受到多种因素干扰(树木及车辆的投影、光照变化、油渍、树枝与稻草等条状物、各类垃圾),降低了基于道路影像自动识别裂缝算法的准确率,导致对于路面等级评价依旧采用人工的方式进行,为此提出一种道路影像裂缝鲁棒识别方法。方法由于采集的图像尺寸较大,同时为了避免光照不均匀带来的问题,首先对图像进行分块,采用改进的CV模型对分块影像进行预处理,获得初步的分割结果。其次,通过以下4个特点识别线阵列CCD道路影像的裂缝:1)裂缝在分块区域中占据较小的面积比;2)裂缝在影像中呈现的连续性较差;3)裂缝的宽度与长度比值较小;4)同一段裂缝的走向基本一致。为了利用裂缝的后两项特点,采用椭圆拟合的方法计算初步分割区域的方向,并以此为基础将这些区域分为4类。在每个分类中,分别计算各区域内的质心位置,建立质心间的矢量表,设计递归算法计算其共线性,获得裂缝检测结果,并以此为基础构造活动模型的初始距离矩阵,通过在原图中迭代求解更为精确的裂缝区域。结果从2 000幅道路影像中挑选包含道路裂缝的影像100幅,并按序号等间隔分别取出5组未含有裂缝的影像100幅,每组200幅组成数据集进行测试,采用分类指标统计的方法评测本文算法性能,在正确率、灵敏度、特效度、精度上均达到95%以上,道路裂缝的检测与提取时间约为1min。结论该方法不仅可以有效地识别裂缝,同时可以克服了环境中多种因素的干扰,误识别率较低,具有较高的实际应用价值。
The grading of the purpose road cracks is one of the basic tasks of road maintenance. At present, relevant departments in the country mainly collect road images through line array cameras. Since the identification of road image cracks will be disturbed by many factors (the projection of trees and vehicles, the change of illumination, Oil stains, branches and straw bars, and various types of rubbish), the accuracy of the algorithm for automatically identifying cracks based on road images is reduced, which leads to the artificial evaluation of the road surface grade. Therefore, a road image crack robust recognition methods. Methods Because of the large size of the captured image, in order to avoid the problem caused by uneven illumination, the image was first segmented, and the improved CV model was used to preprocess the block image to get the initial segmentation result. Secondly, the cracks in line image CCD road image are identified by the following four features: 1) the fracture occupies a smaller area ratio in the patch area; 2) the continuity of the fracture in the image is poor; 3) the width of the fracture and Length ratio is smaller; 4) the same section of the crack is basically the same direction. In order to make use of the latter two characteristics of fractures, the ellipse fitting method is used to calculate the direction of the initial segmentation region, and based on this, the regions are divided into four categories. In each category, calculate the centroid position of each region, set up a vector table of centroid, recursive algorithm is designed to calculate the collinearity, to obtain the results of crack detection. Based on this, the initial distance matrix of activity model is constructed, Figure iterative solution for more accurate crack area. Results A total of 100 images of road cracks were selected from 2 000 road images. Five groups of 100 images without cracks were taken at equal intervals. Each group consisted of 200 composition data sets for testing. Statistical methods of classification were used Evaluation of the performance of the algorithm in this article, in the correct rate, sensitivity, special effects, accuracy reached more than 95%, road crack detection and extraction time is about 1min. Conclusion The method not only can effectively identify cracks, but also can overcome the interference of many factors in the environment, and has a low false positive rate and high practical value.