Research And Implementation Of Recognition Algorithm For X-ray Specific Object Images

This issue comes from a research project proposed by a logistics enterprise in Inner Mongolia.With the X-ray security inspection machines being widely used at numerous security sites,using X-ray inspection machines to detect suspicious items in packages has become a significant part of logistics safety.Due to the fact that the research on unattended X-ray image recognition technology is still in the initial stage,the current security personnel should be on duty all the time while using X-rays security inspection machine.This reduces the work efficiency of security personnel.Therefore,how to realize real-time detection of the specific object image under the X-ray security inspection machine correctly has become an urgent research issue in logistics industry.This paper combines the requirements of engineering projects put forward by a logistics company,designs an image recognition algorithm to detect specific objects on an unattended X-ray security inspection machine without affecting the normal operation of the security inspection machine during logistics inspection.Specific objects are supposed to be accurately and quickly identified through machine vision technology to meet the timeliness of logistic industry.This paper focuses on three algorithms implemented on specific objects recognition on X-ray images.Firstly,the basic digital image process is verified by SIFT algorithm.Secondly,the paper comes from practical problems and needs of a logistic industry,so in order to combine the algorithm with the hardware system,an image recognition algorithm based on the histogram was implemented and successfully applied on X-ray images with simple background.Finally,in order to make up for the shortcomings of the histogram method,a convolutional neural network is built successfully,which improves the accuracy rate and test speed.