Research on text recognition method for construction work order images

The problem of diversified text identification in work order images and the need for faster response time in certain scenarios is addressed in this paper. Firstly, the lightweight convolutional neural network PP HGNet is improved, and then the SVTR network is effectively fused with the improved lightweight convolutional neural network PP HGNet to enhance the adaptability of the model to diversified text identification. At the same time, by using attention modules to guide CTC training strategies, a guiding attention mechanism is provided for the model, enabling it to recognize text more accurately. Through experimental verification, the improved method proposed in this article significantly improves the response speed of text recognition in construction work ticket images while maintaining a high recognition success rate for diversified text identification.