Research On Prohibited Item Detection Technology Based On X-ray Image Of Security Inspection Machine

The flow of domestic personnel is increasing day by day,and security is particularly important.The security inspection field faces the problem of strong repetitiveness and heavy tasks.At present,the security inspection work mainly relies on manual inspection,which is not only inefficient but also prone to false and missed detection.In order to solve this problem and alleviate the work pressure of security personnel,it is urgent to design a prohibited items assisted detection system to help complete the security inspection task efficiently and accurately.This thesis first establishes the prohibited items dataset in X-ray images,and then proposes prohibited items detection algorithm based on FM-RP-Net for the characteristics of X-ray images clutter and prohibited items scale difference.Finally,according to the designed prohibited items detection algorithm in X-ray images,the assisted detection system is realized.The work and research results of this thesis are reflected in the following aspects:(1)Establish the prohibited items dataset in X-ray images.This thesis collects 6 classes of 5,148 common prohibited items in X-ray images and merges them with the public prohibited items dataset in X-ray images.Image de-noising and enhancement algorithms are used to improve image quality for noise problems in X-ray images.Finally,use the software to label the class and location of the prohibited items in the X-ray images.(2)Design a prohibited items detection algorithm in X-ray images based on FM-RP-Net.Firstly,for the characteristics of prohibited items scale difference,a feature pyramid idea is introduced in the prohibited items feature extraction module,and a multi-scale feature fusion structure is designed to enhance the FM-RP-Net model’s ability to perceive and express prohibited items in different sizes.Then,based on the clutter of X-ray images,the prohibited items region proposal module proposes the position prediction algorithm to predict regions that may contain prohibited items.Setting threshold to eliminate the useless background areas,and precisely locate the prohibited items.Finally,for the task objectives of different stages of prohibited items detection algorithm in X-ray images,the multi-task loss function is introduced as the target optimization function of the whole model.The design comparison experiment shows that the prohibited items detection algorithm in X-ray images proposed in this thesis can detect prohibited items accurately and quickly.(3)Implement prohibited items assisted detection system in X-ray images.This thesis uses the QT graphical interface application and the C++programming language to implement the prohibited items assisted detection system in X-ray images.The system includes a user login module,an image preprocessing module,and a prohibited items detection module.Verify the modules of the system through software testing.The result shows that the prohibited items assisted detection system in X-ray images can automatically detect a variety of prohibited items,which has important practical value in the security field.