中文摘要：

针对基于图像底层特征的泡沫图像分类识别正确率不高、存在语义鸿沟问题，提出一种基于向量空间模型（VSM）的浮选泡沫图像分类方法。该方法借鉴文本分类方法，对工业摄像机获取的大量泡沫图像通过分块、底层特征提取和聚类，构造泡沫状态词汇表；在此基础上，经词汇相似度和词频计算，用词袋向量描述泡沫图像；最后，采用VSM实现实时泡沫图像的有监督分类识别。用某金属浮选过程工业现场泡沫图像数据对该方法进行了实验验证，实验结果表明，该方法的工况识别平均准确率近90％，明显优于基于底层特征的分类方法，并在一定程度上解决了语义鸿沟问题，具有很好的应用价值。

英文摘要：

A flotation froth image classification method based on vector space model（VSM） is presented to solve the problem that existing flotation froth image classification methods based on froth bottom characteristics have the shortcomings of low accuracy and semantic gap. Referring to the method of text classification, the new method divides froth images into blocks, extracts these blocks＇ bottom characteristics and clusters them to build the table of the froth status words. Based on the table, the similarity between the words and the word frequency are calculated, and then the froth images is described with a bag-of-word vector. Finally, the classification and recognition of real time froth image is realized based on the VSM method. The results of the experiment performed by using the real plant data show that the proposed method can increase the classification accuracy to nearly 90% ,which is higher than the accuracy of classification method based on bottom characteristics. The proposed method can solve the problem of semantic gap to a certain extent, showing its important value in industrial application.