Information-code recognition technology promotes societal progress and provides convenience.Owing to the effect of the photography environment,the information-code recognition effect must be improved,and the information-code angle tilt affects the decoding accuracy.Using information code-based power-transformer error test wiring judgment as the background,this paper proposes an information-code correction algorithm based on the improved PPYOLOE-R.First,based on the PPYOLOE-R detection algorithm,a lightweight network,ESNet,is integrated to improve accuracy and reduce the number of model parameters.Second,dynamic convolution is introduced to further enhance feature extraction,reduce information loss in the model due to subsampling,and enhance the feature-extraction capability of the model channel.Finally,to satisfy the real-time requirements of Artificial Intelligence(AI)edge devices,model fusion technology is applied to fuse the inference model to improve the model detection speed without changing the accuracy of the model.To enrich the dataset,two-step rotation data-enhancement and Mosaic+Mixup data-enhancement methods are used to fully utilize existing information in the dataset and improve the learning ability of the model.Experimental results show that the accuracy of the improved algorithm is 89.46％,which is 1.95％higher than that of the original model,and that the detection speed increases from 154 to 50 ms per photograph.Compared with other algorithms,the improved algorithm offers the advantages of small size and high speed,and the decoding efficiency and accuracy can be improved significantly using the corrected information code.
information code correctionArtificial Intelligence(AI)edge computingPPYOLOE-R algorithmdynamic convolutionmodel fusion
维普
万方数据
