Intelligent Detection Model for Surface Cracks in Masonry Walls

Too many cracks on the surface of masonry wall will easily lead to the instability of building structure and cause safety problems. The traditional detection method is to determine whether the wall meets the building code through manual observations which lead to increased costs for a lot of labor and time. To solve this problem, an intelligent detection model was designed to detect and locate surface cracks of masonry walls. Based on YOLOX, we added convolutional attention module CBAM into the backbone network to improve the detection accuracy by integrates the attention mechanism of channel and space. The masonry wall data set adopted in this study was acquired from the construction project of Shipai Town, Anqing City, Anhui Province, which met the actual engineering requirements. The experiment results show that the proposed model obtained 75.10% AP value on the masonry wall crack data set, which was 4.63 points higher than the original Yolox respectively, and the recognition accuracy was greatly improved.