Intelligent Detection of Construction Defects in Prefabricated Buildings Based on Improved Residual Networks

In recent years, with the widespread application of prefabricated buildings in residential and commercial buildings, defects in the construction process, such as component misalignment and loose connections, have gradually become important factors affecting their overall quality and structural safety. This article proposes an intelligent detection method based on an improved residual network to effectively identify and locate defects in prefabricated construction. Firstly, this article enhances the network’s ability to learn defect features of different scales by introducing a multi-scale feature extraction module; Next, the article employs an attention mechanism to enhance the feature selection process, boosting the model’s defect detection performance in complex environments; finally, the article reduces the vanishing gradient problem by improving the residual structure, making the training of deep networks more stable and efficient. In the experimental conclusion, the proposed method performs well in the defect detection task of prefabricated building construction, with an accuracy of 89.3% and a recall rate of 85.7%, significantly better than the unimproved model method, verifying the effectiveness and reliability of the improved residual network.