Research on Key Techniques of Text Recognition under Strong Light Noise

Text recognition technology based on deep learning has achieved great success in recent years. However, these recognition models have poor robustness under the influence of strong light noise. The existing solution to this problem leverages preprocessing methods to remove uneven lighting, which shows just passable effects on noise reduction for strong light. In this paper, we propose a text image recognition algorithm with attention-based strong light region perception module to solve this problem. The model is trained to detect strong light regions and guide the generated feature maps to focus on the clear part of the input image. Experimental results show that this model mitigates the impact of strong light noise pollution and improves the robustness of recognition technology on multiple datasets.