Image Segmentation Based On Depth R-FCN In X-ray Images Segmentation And Recognition Of Prohibited Articles

X-ray equipment has been increasing in recent years,especially in terms of security,it is used to detect people’s luggage.This kind of equipment is widely used in airports to detect people’s luggage.With these devices,dangerous prohibited items can be found and terrorist acts can be prevented.The detection of prohibited objects requires a high degree of accuracy,and the manual operation is both cumbersome and requires constant attention and labor.Computer-based automated systems have many problems in the detection of prohibited items.In addition,the prohibited items overlap with other items in the luggage and their positions are concealed when they are detected,which makes the detection process difficult.In view of this,this article proposes a new method of feature extraction and target detection based on deep learning with better detection effect for X-ray images.First,the preprocessing method of image data is improved.Aiming at the problem that most of the image data in the database does not meet the input specifications of the selected deep neural network,a centralized cropping method will be used to achieve the target object in the cropped image.Then use methods such as image rotation to enhance the data set.After that,the ResNet152 residual network is used to extract the features of the processed image data,and the process uses transfer learning to train the model.This method not only breaks the limitation of the large amount of data required by the existing X-ray image detection deep learning,but also effectively avoids the problems of gradient disappearance,explosion and model degradation in deeper deep learning.Secondly,a regional full convolutional network(R-FCN)model is proposed to detect the features of the extracted X-ray image data.Due to the need to quickly detect baggage during the security inspection process,and R-FCN is the fastest among all target detection algorithms,and the accuracy is also very high,so the R-FCN is used for target detection.This article uses a fully-supervised classification detection model to achieve the highest security inspection accuracy.Finally,in order to evaluate the performance of the proposed segmentation recognition framework and feature extraction method,this paper conducts training and testing experiments on the data set.The experimental results show that,compared with traditional segmentation and recognition methods and some deep learning-based segmentation and recognition methods,the segmentation and recognition system proposed in this paper has better segmentation and recognition performance,especially for some categories of X-ray images with small sample sizes.