Autonomous driving of UGV video and Yolo crack detection results
In original paper, a method for automatic detection of cracks in interior wall surfaces using an unmanned ground vehicle (UGV) is proposed.The method consists of three main steps: (1) acquiring interior wall surface images using an autonomous UGV, (2) generating orthoimages to capture an overall view of the inspection area, and (3) detecting crack locations using the YOLO-v7 model. For creating the orthoimages used by Structure from Motion (SfM), an indoor navigation algorithm combining AR markers and LiDAR data is proposed. The algorithm detects AR markers using OpenCV and uses LiDAR to measure the distance and angle between the UGV and the wall. The UGV is guided to maintain a constant distance and align parallel to the wall. Experimental verification is conducted on mortar-finished interior wall surfaces in a building, demonstrating the effectiveness of the proposed method. The UGV captures images while in motion, creating orthoimages and detecting cracks using YOLO-v7. The results show that orthoimages allow for the detection of significant cracks, but direct utilization of UGV-captured images is necessary to detect all cracks. The proposed method offers a convenient way to acquire wall surface images and enables automatic crack detection using orthoimages and YOLO-v7. The navigation algorithm facilitates UGV traversal at a constant distance from the wall. Overall, the research presented the potential of using UGVs for automatic crack detection from images of indoor building environments. Here the data of autonomous driving of UGV video and Yolo crack detection results is uploaded here.