Against the backdrop of road connectivity between China and Laos,research on license plate recognition in Laos is crucial for cross-border vehicle management in China.However,existing single license plate recognition methods cannot be directly applied to dual license plate recognition tasks in Laos.This paper proposes an end-to-end Lao car license plate recognition method that integrates a two-layer attention network,in response to the problems of tight arrangement of characters in the upstream provinces of the Lao car license plate,difficulty in segmentation,and high similarity and difficulty in recognizing the downstream consonant characters.Extracting and enhancing the representation of upstream province features and downstream character features through channel and spatial attention.Applying the idea of classification to provincial information acquisition,effectively addressing the problem of single character recognition not being possible due to character adhesion.Using sequence annotation to alleviate the difficulty of similar character recognition and improve the accuracy of character recognition.The experimental results show that the proposed method improves the accuracy of the baseline model by 0.8 percentage points,reaching 92.7%.
deep learningLaos dual license plate recognitiondouble layer attention networkchannel and spatial attentionend to end
维普
万方数据
