一种基于SVM的路面影像损伤跨尺度识别方法  

A Road Damage Identification Method Based on Scale-span Image and SVM

作　　者：沈照庆[1] 彭余华[1] 舒宁[2] 

机构地区：[1]长安大学公路学院,西安市南二环路中段710064 [2]武汉大学遥感信息工程学院,武汉市珞喻路129号430079

出　　处：《武汉大学学报（信息科学版）》2013年第8期993-997,共5页Geomatics and Information Science of Wuhan University

基　　金：国家973计划资助项目(2006CB701303);长江学者和创新团队发展计划资助项目(IRT1050);中央高校基本科研业务费专项资金资助项目(CHD2011JC011;CHD2011JC085);广西科技攻关与新产品试制项目(桂科攻10124006-10)

摘　　要：针对路面裂缝识别精度和效率较低的问题,提出了一种精确快速检测路面破损图像的跨尺度识别方法。在影像的空间域中,将路面影像根据一定的阈值分为若干个不同尺度的特征影像,根据不同路面损伤影像特性选择若干最佳损伤识别尺度影像有效叠加为跨尺度影像模型,设计了基于支持向量机(support vectormachine,SVM)的"1Vm"道路损伤影像裂缝模式识别器,可以有效提取细裂缝和粗裂缝。实验结果表明,该算法提高了路面破损图像识别的精度和效率,能快速准确地识别出路面损伤的程度和范围。In order to improve the accuracy and efficiency of the pavement crack recognition, a precise fast pavement damage detection method is proposed based on scale-span images. In the image space domain, a pavement image is transformed into differently scaled images ac- cording to different thresholds firstly. Several damage identification scale images are selected to set up scale-span image model by superimposing according to different pavement damage image characteristics. At the same time, the road image cracks pattern identifier is designed based on support vector machine （SVM） ＂1 V m＂ to extract the fine cracks and thick cracks effectively. The test results show that the algorithm has greatly increased the accuracy and efficiency of the pavement cracks image recognition, The pavement damage degree and range can be also identified quickly and accurately.

关 键 词：路面损伤 影像识别 SVM 跨尺度 道路工程 

分 类 号：P237.4[天文地球—摄影测量与遥感] P237.9[天文地球—测绘科学与技术]