Abstract: Relevance feedback technology based on machine learning becomes focus in image retrieval.In relevance feedback based on SVM，sample is lack and unbalance，precise feedback is low.For these problem，in this paper，a new relevance feedback method based on machine learning is presented，which combines on Boosting and SVM.It improves image retrieval speed and accuracy.A new feedback method based on long machine learning is presented.Experiments show that the proposed system is not only efficient but also effective.

Key words: relevance feedback, machine learning, Boosting, image retrieval

摘要： 基于机器学习的相关反馈技术是基于内容的图像检索研究的热点。由于基于SVM的相关反馈技术存在样本数量少，样本正负比例不平衡，反馈准确率低等问题，文中先对Boosting方法进行改进，提出了用先验知识的Boosting方法与SVM结合的短期机器学习相关反馈方法（BSVM）；在此基础上为进一步提高系统反馈速度与准确率，通过保存训练好的分类器和它对应的样本，提出了基于长期机器学习的相关反馈方法（LBSVM）。文中提出的两种方法与其它方法进行了比较实验，结果表明，该方法优于其它方法。

关键词: 相关反馈, 机器学习, Boosting方法, 图像检索