Semantic Segmentation of Road Cracks Based on Improved DeepLabV3+

As the scale of China's modern transportation road network jumps to the forefront of the world, once the road cracks will affect traffic safety. However, the intelligent detection of road cracks is still facing the problems of inefficient manual inspection, fuzzy segmentation of edge features and high leakage rate of small cracks. To address the above problems, this paper proposes an improved road crack segmentation method RSDM-DeepLab based on DeepLabV3+ algorithm, which adopts ResNet-50 as the backbone network to strengthen the morphological feature extraction ability of cracks, improves the complete identification effect of crack region by integrating the multi-scale ASPP module with dense connection mechanism and bar pooling layer, and constructs a multi-scale feature fusion decoder to optimize the edge localization of cracks. decoder to optimize the crack edge positioning accuracy. The experimental results show that the mPA, mIou and F1 Score of the model proposed in this paper are improved by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$3.08 \%, 2.65 \%$</tex> and 2.67 %, respectively. It provides an efficient and accurate intelligent detection solution for road preventive maintenance.