融合多尺度CNN和结构特征的傣文字符识别OA

As an ethnic script with a long history,the Dai language features complex character structures and high glyph similarity,making it difficult for single-feature methods to fully capture character details.To address this,this paper proposes a Dai character recognition model that integrates multi-scale CNN and structural features.First,structural attributes such as HOG feature vectors and projection histograms are extracted from character images to preserve more detailed information.Then,a multi-scale convolutional module is used to extract deep features at different levels.The structural and deep features are fused,and a global average pooling(GAP)layer is introduced at the end of the network to replace traditional fully connected layers for classification.This simplifies the computational process while significantly reducing the number of parameters.Experimental results show that the model achieves a recognition accuracy of 98.71%,outperforming other methods and yielding the best recognition performance.