Detection of Tile Surface Based on SSM-R-CNN Algorithm

A computer vision algorithm, SSM-R-CNN, based on feature extraction and adaptive weighting, is proposed to address the challenges of diverse tile types, complex surface areas, and varied defect morphologies, as well as low image contrast in tile surface defect detection. The algorithm extracts templates and defect images from the original images, aligns them using the SURF algorithm, and enhances the features through gray image differencing, mean, and variance normalization. By incorporating edge-preserving filters and bilateral filters, the algorithm achieves a flexible balance between image smoothing and detail preservation to meet the requirements of different application scenarios. Furthermore, the algorithm improves the Mask R-CNN model by integrating the SENet channel attention mechanism, Explore the interaction and importance of feature information between learning channels, thereby enhancing the network’s expressive power and accuracy. Experimental results demonstrate the outstanding performance of the proposed algorithm in tile surface defect detection, achieving excellent accuracy, recall rate, and F1-score of 98.72%, 95.65%, and 97.18%, respectively. The algorithm exhibits stable and reliable performance in handling complex textures, non-uniform lighting, and low contrast. With its promising prospects in the field of tile surface defect detection, this method provides a new approach to adaptive weighting.