Abstract: In order to improve the accuracy and efficiency of image annotation system，an image annotation method based on a genetic feature selection and support vector machines is proposed.The method extracts the image visual features from the multimedia content description interface（MPEG-7） standard，uses a bi-coded chromosome genetic algorithm to select optimal weighted feature subset from MPEG-7 standard，and trains classifiers of the support vector machines using majority voting scheme for image annotation.The annotation results over 2 000 Corel images show that the method can select optimal weighted feature subset，and improves the accuracy and efficiency of image annotation system.

摘要： 为了提高图像标注系统的精度和效率，提出了基于遗传特征选择和支持向量机的图像标注方法。该方法从多媒体描述接口（MPEG-7）标准中抽取图像的视觉特征，采用双编码遗传算法从MPEG-7标准中选择最优的加权特征子集，并训练支持向量机分类器用于图像标注，支持向量机分类器采用多数投票机制。对2 000幅Corel图像的标注结果表明：该方法可以获得最优的加权特征子集，提高了图像标注系统的精度和效率。