With the continuous development of power system construction and operation, the acceptance of the quality of distribution line engineering has gradually become an important link to ensure the safe and stable operation of the system. In this paper, a defect detection method for distribution line engineering quality based on improved YOLOv5s is proposed. Aiming at the problems of low detection efficiency and large error of the traditional method, the study optimizes the detection performance by introducing a lightweight YOLOv5s model and combining the BiFPN network, the CBAM attention mechanism, and the improved loss function. The experimental results show that the improved model has significant improvement in several performance metrics. The average detection accuracy (mAP) reaches 94.11% when using the model, which is only 0.56% lower than the original YOLOv5s model. In addition, the computational and parametric quantities of the model were reduced by 84.49% and 85.00%, respectively, showing excellent lightweight characteristics. The model is also optimized in terms of detection speed and is able to reach 106.27 frames per second. The conclusion shows that the improved YOLOv5s model performs well in the detection of quality defects in distribution line engineering, which can effectively improve the detection accuracy and shorten the detection time to meet the real-time detection needs of industrial sites.
