航拍图像的路面裂缝识别

发布时间：2018-04-28 06:24

本文选题：图像处理 + 航拍目标检测　；参考：《光学学报》2017年08期

【摘要】：针对航拍沥青路面图像识别的噪声和干扰问题,提出一种应用于航拍图像的路面裂缝识别算法。根据路面区域与路旁景观区域灰度级数分布不同,采用多方向拟合的区域生长方法联合HSV颜色空间阈值进行路面区域分割,提取包含完整裂缝信息的单通道路面;再通过改进的形态学滤波剔除面积较大的干扰区域,利用结合显著性分析的边缘检测算法识别路面的裂缝片段,实现复杂裂缝与路面纹理噪声的区分;自动筛选存在裂缝的图像,针对裂缝可疑区域,结合人眼辅助观察标记并计算其长度。结果表明,该算法可有效剔除图像中的噪声和干扰,较好地识别沥青路面的裂缝,裂缝宽度的识别精度能达到9.7mm,分类识别准确率大于80.0%,长度测量准确率大于75.0%。
[Abstract]:In order to solve the problem of noise and disturbance in aerial image recognition of asphalt pavement, an algorithm of pavement crack recognition is proposed. According to the different grayscale series distribution between the road surface area and the roadside landscape area, the multi-direction fitting region growth method combined with the HSV color space threshold is used to segment the pavement area, and the single channel pavement containing the complete crack information is extracted. Then the improved morphological filter is used to eliminate the large area of interference and the edge detection algorithm combined with salience analysis is used to identify the crack segment of the road surface to distinguish the complex crack from the road texture noise. The images with cracks are automatically screened, and the length of the cracks is calculated by combining with the human eye observation marks. The results show that the algorithm can effectively eliminate the noise and interference in the image, and better identify the cracks of asphalt pavement. The recognition accuracy of crack width can reach 9.7 mm, the accuracy of classification recognition is more than 80.0, and the accuracy of length measurement is more than 75.0.
【作者单位】：北京理工大学光电学院光电成像技术与系统教育部重点实验室;北京理工大学宇航学院;
【基金】：国家自然科学基金(61575023)
【分类号】：TP391.41;U418.6

【相似文献】