A Novel Deep Learning Method for the Identification and Categorization of Footpath Defects based on Thermography

Automated detection represents one of the most important criteria for assessing the condition of the roadway, and it represents a challenging problem for the creation of advanced practices and methods. A large data set that includes complicated surface conditions is necessary for autonomous damage identification utilizing deep convolutional neural network. This study offers a thermal-RGB fusion picture-based sidewalk defect identification method, in which the merged RGB-thermal image is created by different sensor data to accomplish quick and precise defect identification considering complicated pavement circumstances. The suggested technique achieves high sidewalk damage recognition accuracy by using EfficientNet as the foundation structure and creating an argument database. The efficiency of various input information (RGB, thermal, MSX, and fused picture) is tested individually in this article to examine the effects of input information and system on the identification outcomes. The findings demonstrated that the merged image’s degradation detection performance could reach 98.33 percent, and by applying the information after supplementation, the classification method is deemed to be significantly stable to attain 98.33 percent.