High-Precision Image Recognition Technology in Pattern Analysis of Intangible Cultural Heritage Textiles

Xue Bai

Article
2026 / Volume 9 / Pages 1162-1190
Received 10 August 2025; Accepted 14 October 2025; Published 28 April 2026
https://doi.org/10.31881/TLR.2026.1162

Abstract
Aiming at the problems such as high consumption of computing resources and insufficient feature extraction when processing complex pattern images in the existing methods, based on VGGNet, ResNet-50, EfficientNetV2 and soft attention mechanism, a new type of recognition and classification method for patterns of intangible cultural heritage textiles is proposed. This method effectively enhances the multi-scale feature extraction ability of the network for complex intangible cultural heritage textile patterns by introducing the improved ReLU activation function and the optimised convolutional block structure, and improves the recognition accuracy and computational efficiency. The findings demonstrated that this method had the highest classification and recognition accuracy rate for Shuizumaweixiu, Xiqinciciu, Hamicixiu, Suxiu, Xiangxiu, Shuxiu and other embroiderings in the textile pattern image classification dataset, while the classification accuracy rate for the types of Yuexiu was relatively the lowest. Overall, the research method achieves an average prediction accuracy rate of over 88% for the eight types of pattern images in the textile pattern image classification dataset. Not only does it outperform other advanced models numerically, but it also performs more evenly across various categories. It is demonstrated that the research method has significant adaptability and generalisation capabilities in the field of pattern recognition of intangible cultural heritage textiles. It can effectively address the limitations of conventional deep learning models in complex pattern classification, thereby providing a scientific foundation and technological support for the digital preservation and inheritance of intangible cultural heritage textiles.

Keywords
high precision, image recognition, convolutional neural network, efficientnet, textiles