Abstract: The methods that based on color or color-space features for image semantic classification are simple and efficiently，but no consider the shape feature of the object in the image，so its classification result always not ideally.The paper uses cloth image as date，proposes a new image classification method which base on multi-features including the color feature and the shape feature.The relationship between high-level emotional and level features is strongly.The paper brings forward the relationship between high-level emotional representation and perceptual level features based on the clothing image.Classification is performed by PNN（Probabilistic Neural Networks）.The experiment indicates that this method is great effective in image semantic classification and improves the image semantic classification precision.

Key words: image of clothing, multi-features, Probabilistic Neural Networks（PNN）, semantic classification

摘要： 基于颜色或颜色-空间信息的图像分类方法，由于没有考虑图像中所含目标对象的形状特征，分类效果不够理想，以服装图像作为数据源，提出并设计了颜色-边缘方向角二维直方图，将图像的颜色特征与形状特征融合起来进行图像分类。图像中的低阶可视化特征与高阶情感概念之间有着密切的关联，分析了服装图像的颜色和形状的融合特征与情感之间的相关性，采用概率神经网络作为分类算法来完成情感语义分类，实验结果表明，该方法的分类精度有了明显的提高。

关键词: 服装图像, 特征融合, 概率神经网络, 语义分类