Barcode Detection Algorithm Based on Improved Swin Transformer
Barcode detection is a technology widely used in different industries to identify,verify and check the quality and accuracy of barcodes.However,the traditional detection method has poor performance in complex situations such as large changes in the processing scale.Detection speed and efficiency are low,and the application range is limited.The traditional method usually only focuses on the local information of the image,ignoring the characteristics of the barcode at different scales.In order to solve the above problems,a barcode detection algorithm based on improved Swin Transformer is proposed.Firstly,by introducing local perception and multi-scale feature extraction mechanisms,it has better robustness and can cope with barcodes of different sizes and shapes.Then,the innovative idea of FCOS-based detection framework is introduced.Finally,the model training and testing of the improved algorithm on the labeled barcode dataset show that the accuracy and recall of the improved model proposed in this paper are improved by 5.78％and 3.18％compared with the YOLOv4 algorithm,respectively,and the overall performance is better than other mainstream algorithms,which effectively improves the barcode detection ability and achieves high detection accuracy.
NETL
NSTL
万方数据
