Joint Shape and Texture Based X-Ray Cargo Image Classification

Jian Zhang, Li Zhang, Ziran Zhao, Yaohong Liu, Jianping Gu, Qiang Li, Duokun Zhang; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2014, pp. 266-273

Security & Inspection X-Ray Systems is widely used by custom to accomplish some security missions by inspecting import-export cargo. Due to the specificity of cargo X-Ray image, such as overlap, viewpoint dependence, and variants of cargo categories, it couldn't be understood easily like natural ones by human. Even for experienced screeners, it's very difficult to judge cargo category and contraband. In this paper, cargo X-Ray image is described by joint shape and texture feature, which could reflect both cargo stacking mode and interior details. Classification performance is compared with the benchmark method by top hit 1, 3, 5 ratio, and it's demonstrated that good performance is achieved here. In addition, we also discuss X-Ray image property and explore some reasons why cargo classification under X-Ray is very difficult.