基于随机配置网络的服装分类方法

Clothing Classification Method Based on Stochastic Configuration Networks

摘要: 为了提高服装图像分类的精度，提出了一种基于特征提取的正余弦算法优化SCN (Stochastic Con-figuration Networks)的服装图像分类预测模型。通过引入多通道并行的卷积、池化、批量标准化操作，提高了模型的特征提取能力。将正余弦算法引入到SCN的节点搜索过程，优化了SCN节点生成规则，进一步提高了SCN的准确性。结果显示，改进后的SCN算法能够寻找到更优的节点，图像分类准确率达到了94.05%，验证了本文模型的稳定性和可靠性。

Abstract: In order to improve the accuracy of apparel image classification, a predictive model for apparel im-age classification based on feature extraction with the sine-cosine algorithm optimized SCN (Sto-chastic Configuration Networks) is proposed. The feature extraction capability of the model is im-proved by introducing multi-channel parallel convolution, pooling, and batch normalization opera-tions. The sine cosine algorithm is introduced into the node search process of SCN to optimize the SCN node generation rule and further improve the accuracy of SCN. The results show that the im-proved SCN algorithm is able to search for better nodes, and the image classification accuracy reaches 94.05%, which verifies the stability and reliability of the model in this paper.