Automated Detection for Concrete Surface Cracks Based on Deeplabv3+ BDF

Concrete cracks have always been the focus of research because serious damage they cause to structures. With updating hardware and algorithms, detection concrete structure surface based on computer vision has received extensive attention. This paper proposes an improved algorithm open-source model Deeplabv3+ names it BDF according optimization strategy used. first replaces original backbone Xception with MobileNetv2 further all standard convolutions depthwise separable (DSC) achieve a light weight. The feature map shallow convolution layer is additionally fused improve detail segmentation effect. A new proposed, which different from two-stage training. training carried out in order transfer learning, coarse-annotation fine-annotation comparative test results show that showed good performance validation set achieved highest mIoU efficiency, reaching real-time accurate detection.