Image Classification via Intermediate Discriminative Representation

摘要：

motivated in part by the PLSA (probabilistic latent semantic analysis) model from the statistical text literature, in this issue, we explored a kernel based assignment method with squared pooling strategy to implement the application of bag-of-words model in commodity images classification. Instead of searching nearest visual word to represent local patch, in this paper, we present a kernel based assignment method to assign visual words and company with squared pooling function in pyramid match method to intermediate representation for commodity images analysis, and explored a general cluster algorithm to create automatically so-called discriminative visual words and SVM classifier with histogram intersection kernel to classify commodity image. Experimental show that this method does a good and promising performance in commodity image classification.

关键词： commodity image classification visual words pyramid match histogram intersection

作者: Aiping Feng Honggang Zhang Lixia Liu Jun Guo

作者单位: School of information and Communication Engineering,Beijing University of Posts and Telecommunications Beijing, China

会议类型: 国际会议

会议名称: 2010 International Conference on Information Security and Artificial Intelligence(2010年信息安全与人工智能国际会议 ISAI 2010)

会议地点: 成都

会议语种:英文

页码: 433-437

在线出版日期: 2010-12-17（万方平台首次上网日期，不代表论文的发表时间）