Research on Enhanced Concrete Crack Recognition Algorithm Based on YOLOv11

To address the challenges of low recognition accuracy and high false-negative rates caused by the intertwined interference of mold,water stains,and construction textures,as well as the complex multi-branch morphology of concrete cracks,this study proposes an enhanced perception algorithm,YOLOv11n-FLD.First,a concrete crack dataset oriented toward complex engineering conditions was constructed based on a structural inspection project in Guangzhou and relevant resources.Through standardized cropping and data augmentation,the dataset was expanded to 2730 images and partitioned into training,validation,and test sets at a ratio of 8:1:1., In terms of algorithmic improvements:a Frequency-Spatial Convolution module is introduced into the backbone network, utilizing the Discrete Wavelet Transform to effectively decouple high-frequency crack signals from low-frequency background information. Furthermore, a Large-Kernel Separable Convolutional Attention mechanism is integrated into the SPPF module to expand the effective receptive field and enhance the network's ability to capture the linear topological features of slender and branching cracks. To mitigate semantic loss during downsampling, a Deep Robust Feature Downsampling module replaces traditional stride convolutions, ensuring the fidelity of critical details in deep layers. Experimental results demonstrate that on the self-built concrete crack dataset, the YOLOv11n-FLD model achieves a mean Average Precision ( mAP_50 ) of 97.8% and a Recall of 95.9%, representing significant improvements of 7.5% and 8.2% over the baseline YOLOv11n, respectively, with the localization accuracy mAP_50:95 increasing by 10.6%. Compared with mainstream algorithms ranging from YOLOv5n to YOLOv10n, the proposed model ranks first in all precision metrics ( mAP_50 、 mAP_50:95 ). These findings validate the significant advantages of the proposed mechanisms in handling complex background interference and fine-feature recognition, providing a reliable solution for the accurate recognition of concrete cracks.