基于注意力机制的艾德莱斯绸纹饰图案分割研究

Segmentation of Adelaide Silk Decorative Patterns Based on Attention Mechanism

目的由于艾德莱斯绸具有丰富的色彩和复杂的纹饰图案,在对其进行图案分割时难度较大,容易出现错分割和漏分割的情况。为此,提出了基于注意力机制的艾德莱斯绸纹饰图案分割算法。方法采用FCN模型对艾德莱斯绸纹饰图像进行卷积训练,突出图像的语义特征信息。利用通道注意力模块和位置注意力模块,分别对艾德莱斯绸纹饰图像展开学习,得到维度完全相同的特征图。将两个模块特征图融合后与FCN模型输出图像再次融合,得到艾德莱斯绸纹饰图像的特征提取结果,选取图像中的感兴趣区域,完成对艾德莱斯绸纹饰图案的分割。结论实验结果表明,所提方法取得了精准度较高的分割结果,分割图像边缘清晰,没有出现错分割和漏分割的情况,分割结果总体上较为理想。

Due to the rich colors and complex decorative patterns of Adelaide silk,it is difficult to segment its pat-terns,and it is prone to errors and omissions in segmentation.To this end,the work aims to propose an attention mecha-nism based segmentation algorithm for Adelaide silk decorative patterns.The FCN model was used for convolutional training of Adelaide silk decorative images,highlighting the semantic feature information of the images.By using the channel attention module and the position attention module,the Adelaide silk pattern images were respectively learned to obtain feature maps with identical dimensions.After fusing the feature maps of the two modules with the output image of the FCN model,the feature extraction results of the Adelaide silk pattern image were obtained.The regions of interest in the image were selected to complete the segmentation of the Adelaide silk pattern.The experimental results show that the proposed method has achieved high accuracy in segmentation results,with clear edges of the segmented image and no occurrence of wrong or missed segmentation.The overall segmentation results are relatively ideal.