Considering the high incidence of accidents at tunnel construction sites, using robots to replace humans in hazardous tasks can effectively safeguard their lives. However, most robots currently used in this field require manual control and lack autonomous obstacle avoidance capability. To address these issues, we propose a lightweight model based on an improved version of YOLOv5 for obstacle detection. Firstly, to enhance detection speed and reduce computational load, we modify the backbone network to the lightweight Shufflenet v2. Secondly, we introduce a coordinate attention mechanism to enhance the network’s ability to learn feature representations. Subsequently, we replace the neck convolution block with GSConv to improve the model’s efficiency. Finally, we modify the model’s upsampling method to further enhance detection accuracy. Through comparative experiments on the model, the results demonstrate that our approach achieves an approximately 37% increase in detection speed with a minimal accuracy reduction of 1.5%. The frame rate has improved by about 54%, the parameter count has decreased by approximately 74%, and the model size has decreased by 2.5 MB. The experimental results indicate that our method can reduce hardware requirements for the model, striking a balance between detection speed and accuracy.

In underground engineering, the detection of structural cracks on tunnel surfaces stands as a pivotal task in ensuring the health and reliability of tunnel structures. However, the dim and dusty environment inherent to underground engineering poses considerable challenges to crack segmentation. This paper proposes a crack segmentation algorithm termed as Focused Detection for Subsurface Cracks YOLOv8 (FDSC-YOLOv8) specifically designed for underground engineering structural surfaces. Firstly, to improve the extraction of multi-layer convolutional features, the fixed convolutional module is replaced with a deformable convolutional module. Secondly, the model’s receptive field is enhanced by introducing a multi-branch convolutional module, improving the extraction of shallow features for small targets. Next, the Dynamic Snake Convolution module is incorporated to enhance the extraction capability for slender and weak cracks. Finally, the Convolutional Block Attention Module (CBAM) module is employed to achieve better target determination. The FDSC-YOLOv8s algorithm’s mAP50 and mAP50-95 reach 96.5% and 66.4%, according to the testing data.