基于深度学习的沥青路面病害感知智能分析模型

Research on the Identification Method of Asphalt Pavement Disease Based on 3D Ground-penetrating Radar

近年来,我国高速公路网以每年数千公里的速度增长,高速公路的路面病害检测与养护工程已成为一大热点。然而,由于路面病害结构的复杂性与隐蔽性,现阶段的检测技术还存在速度慢、精度差、检测内容不全面、自动化程度低等问题,传统的检测技术难以适应当前道路建设的发展。三维探地雷达作为一种新型无损检测设备,兼顾无损、高效、精确、连续和数据处理简单的特点。基于此,为更加精准快速地检测沥青路面病害,掌握三维探地雷达的使用方法,本研究选取樟吉高速公路与昌金高速公路共约10km的路段作为试验路段,采用三维探地雷达进行检测,分析雷达检测图像与检测数据,并根据所获取的雷达数据建立YOLOX模型,提出一种基于YOLOX-ResNet50模型的路面病害识别方法。研究表明,根据三维探地雷达检测图像,可以很好地判别裂缝、修补、脱空、层间黏结不良和混合料离析五类病害。

In recent years,Chinese highway network has grown at the rate of thousands of kilometers per year,and the pavement disease detection and maintenance engineering of highways has become a major hot spot.However,the current detection technology still has disadvantages such as slow speed,poor accuracy,incomplete detection content and low automation due to the complexity and concealment of the pavement disease structure.The traditional detection technology is difficult to adapt to the current development of road construction.As a new type of nondestructive testing equipment,3D ground search radar takes into account the characteristics of nondestructive,efficient,accurate,continuous and simple data processing.Based on the above,in order to detect asphalt pavement diseases more accurately and quickly and to master the use of 3D ground-penetrating radar,a total of about 10 km of Zhanji Expressway and Changjin Expressway was selected as a test section,and 3D ground-penetrating radar was used for inspection.The radar inspection images and inspection data were analyzed,and a YOLOX model was established based on the acquired radar data.A pavement disease identification method based on the YOLOX-ResNet50 model was proposed.The study indicates that five types of diseases including cracks,repairs,debonding,poor interlayer bonding and mixture segregation can be well discerned based on the 3D ground-penetrating radar inspection images.