YOLOv7 Deep Learning Model for Pavement Crack Detection Using Close Range Photogrammetry Dataset.

Abstract

Developing an effective system for detecting and classifying pavement cracks is crucial for ensuring traffic safety. However, the procedure of manual inspection for identifying these cracks can be hazardous and time-consuming. Thus, it's essential to implement an automated approach to make the detection process more efficient. Overcoming challenges like varying intensity levels, inconsistent data availability, and ineffective traditional methods make this task complicated. This research's aim is to contribute to the development of an efficient system for detecting pavement cracks. Pavement crack detection using close range photogrammetry is a process for identifying, characterizing and evaluating pavement surface cracks that is revolutionizing the speed, accuracy and cost of assessing the structural integrity of pavements. These images are used by analysis software to generate detailed digital maps of the pavement surface. These digital maps can then be used to identify and measure pavement cracking. The use of close-range photogrammetry for pavement crack detection offers numerous advantages over traditional pavement inspection methods, including improved accuracy and flexibility in the analysis of pavement cracks and the ability to analyze large areas of pavement quickly. The quality of the images captured depends on the type of camera used, but most cameras offer high-resolution imaging at close range. The customized YOLOv7 model, which is a state-of-the-art deep learning algorithm, was used in this study. The precision of the outcome reports is 0.854 and recall from the custom dataset is 0.755. The results of the suggested system were satisfactory compared to the results of reference studies.

Keywords

• deep learning yolov7
• pavement crack detection
• close range photogrammetry