基于FCCA的多特征融合的检索方法

Multi-feature fusion retrieval algorithm based on Faster-RCNN-Canonical correlation analysis

为了加强对民航机场游客行李包的有效管理,减少人工检索相关旅行包的工作量,提出了一种基于FCCA的多特征融合的检索方法。通过卷积神经网络获取图像位置和类别,提取颜色和纹理特征,通过典型相关分析方法将两个特征进行融合然后作为检索的依据,用欧氏距离进行匹配。基于典型相关分析方法的多特征融合检索算法,可应用于机场游客行李包管理领域,具有重要的理论研究价值与实践意义。

This paper proposes a retrieval method of multi- feature fusion,in order to strength the effective management of the tourists’luggage as well as to reduce the workload of manually retrieving travel luggage. The image location and category are obtained by convolution neural network and the features of color and texture are extracted. The two features are fused by canonical correlation analysis and then it is used as the basis of retrieval. Euclidean distance is used to match the image. The multifeature fusion retrieval algorithm is based on canonical correlation analysis. And it can be applied to the field of luggage management for airport tourists,which has important theoretical research value and practical significance.