Application of Deep Learning Algorithm in Crack Detection of Green Building Materials
Feng Wang · 2022
Aiming at the problem of low detection precision of traditional crack detection, Faster R-CNN was taken as the basic network, the features of block3 and Block5 in its backbone network VGG-16 network were fused, and two cascaded R-CNN networks with IoU thresholds of 0.5 and 0.6 respectively were used as the detection network to detect the cracks of green concrete. Simulation results showed that the proposed method reduces the loss of structural information by fusing the features of Block3 and Block5. By cascading the two R-CNN networks with different IoU thresholds, the classification and location of regional suggestion boxes are more accurate, and the precision of crack detection is improved. The average detection precision is 98.29%, and the detection speed is 12. 26fps. Compared with traditional detection algorithms based on SSD and YOLO-V3, the average detection precision and detection speed are improved to varying degrees, which has certain effectiveness and superiority.