Research On Railway Wagon Number Recognition And Brake Shoe Positioning In Complex Background

High-quality and efficient railway wagon inspection work is a key factor to improve the efficiency of railway wagon transportation.Railway wagon number recognition and brake shoe positioning are important aspects of the inspection of railway wagons.At present,some automation equipments have been applied to some railway wagon inspection stations to assist the completion of railway wagon inspection work.However,the inspection of railway wagons mainly relies on manual work,which is time-consuming and labor-intensive,and the work efficiency and accuracy cannot be guaranteed.It makes railway wagons consume a lot of time in the inspection of railway wagons,which greatly reduces the efficiency of railway wagon transportation.In recent years,with the rapid development of artificial intelligence,computer vision technology and robot technology have been widely applied.Using intelligent robots to replace manual work,applying intelligent machine recognition technology to railway wagon inspection,and improving the efficiency and quality of railway wagon inspection have become an inevitable requirement for the development of the times.This paper simulates intelligent robots for railway wagon number recognition and brake shoe positioning,and proposes a method based on deep learning and swarm intelligence optimization machine learning,using computer vision technology to recognize the number of the railway wagon images,and to detect the location of the brake shoe parts on the wagon bogie images.The main work of the paper is as follows:(1)The first is to solve the problem of the number area positioning of railway wagons.On the basis of studying the positioning effect of traditional image processing technology,the railway wagon number area positioning method based on the deep learning model Faster RCNN is proposed,and a better positioning effect is achieved.Two classic deep convolutional neural networks,Alexnet and VGG19,are respectively selected as the backbone feature extraction network of the Faster RCNN model to position and extract the image of the railway wagon number area.For the position and extraction of the railway wagon number area image,the YOLOV2 network model with a lightweight convolutional neural network as the backbone network is used to further perform sequential positioning and segmentation of the railway wagon number characters.(2)Secondly,for the requirement of railway wagon number characters recognition,the HOG(Histogram of Oriented Gradient)feature is extracted from the railway wagon number characters after positioning and segmentation.Based on the comparison with SVM and Sparrow Search Algorithm-SVM(SSA-SVM),the Grey Wolf Optimizer Support Vector Machine(GWO-SVM)algorithm is selected to perform multi-classification recognition on the characters of the railway wagon number in sequence.Aiming at the problem of individual character classification and recognition errors in the overall recognition process of railway wagon number,combined with the rules of railway wagon number encoding,a multi-classification recognition method: Combined Grey Wolf Optimizer Support Vector Machine(C-GWO-SVM)is designed,which improves the overall recognition accuracy of the railway wagon number.The superiority of C-GWO-SVM in character recognition is verified by the experimental comparison with Combined Support Vector Machine(C-SVM)and Combined Sparrow Search Algorithm optimizer Support Vector Machine(C-SSA-SVM)algorithm.(3)Finally,for the brake shoe positioning operation,by training two sets of cascaded Faster RCNN models,combined with the position characteristics and environmental characteristics of the brake shoe in the triangular hole of the bogie,the operation process for step-by-step positioning of the brake shoe is proposed.A set of Faster RCNN models is used to position and segment the triangular hole area of the railway wagon bogie image;the second set of Faster RCNN models is used to position and detect the brake shoe target in the segmented triangular hole area.Two convolutional neural networks,Alexnet and VGG19,are respectively selected as the backbone feature extraction network of the Faster RCNN model to realize the accurate positioning of the brake shoe.