Research on Text Detection Algorithm for Work Ticket Images Based on Feature Enhancement and Multi Feature Fusion

The images of construction work receipts are filled with various noises, such as stains, scratches, handwritten handwriting, and seal images, which result in poor performance of traditional text detection methods. This article proposes a text detection method based on efficient feature enhancement and multi-scale feature fusion to address this issue. By integrating an efficient segmentation module and a learnable post-processing method, the performance of the detection method is enhanced by using the visual language pre training model oCLIP. The segmentation module adopts image segmentation method for text detection of construction work ticket images, including an efficient feature enhancement module and a multi-scale feature fusion, which together constitute the key components of the segmentation module. Experimental results have shown that the model achieves good text detection performance in low resolution and high noise work order images.