Design And Implementation Of Intelligent Mine Vehicle Management System Based On License Plate Recognition

All coal companies have always paid attention to the transportation.However,the management of transport vehicles still has problems.Firstly,manual registration of frequently entrance and exit transport vehicles previously caused many mistakes and affected transportation efficiency.When trucks are weighed during loading and unloading coal,the record may be lost during the operation in the transportation.Weighing data,entrance data and enterprise vehicle database require a lot of manpower to manage and maintain.Secondly,the license plate is the main information of the vehicle.For the convenience of transport vehicles management,the automatic recognition of license plate is a key point.The process of traditional license plate recognition has at least three steps,including plate detection,plate localization,character recognition.The recognition error in each step will affect the result of license plate recognition.And because of the coal slag and the dust,dirt such as coal sludge and coal ash will easily appear on the license plate,and the plate will also be worn out.It causes the low recognition performance of license plate characters.For the management of the coal enterprise,this paper has carried out a detailed demand analysis and architecture design.The intelligent mine management system based on license plate recognition is developed by the Spring Boot framework.Via the technologies of reading video streams to obtain video surveillance,vehicle recognition,license plate recognition identify vehicle license plate in the mining area.Integration of data(such as entrance and weighing data)can reduce manual operations and independence between systems.In view of the particular license plates of coal enterprises,this paper collects and annotates 5,000 license plates in the production environment and 5000 public license plates.Through the use of different object detection methods and backbone networks training and testing on the data set.Finally,selecting a better YOLOv3 object detection model with the Mobile Netv3 backbone.On the other hand,the route is not fixed,and it is not possible to control vehicle access with access bar.And the vehicle speed of entrance and exit is changed,so the recognition of the license plate needs to be real-time.This paper attempts to pruning and compressing the trained model,and the time of predicting an image only using the CPU is 616.8ms and the m AP is reduced from 96.16% to 96.01%.Through the function testing and performance testing of the system,it can be used for vehicle management in coal enterprises and applied in enterprises.