Pedestrian And Non-motor Vehicle Violation Detection System Based On Deep Learning

As the smart city concepts and technologies developing,smart transportation serves as an extremely important part of smart cities.Traffic congestion has become an urgent problem in the urban transportation network.Due to the popularization of the national new energy policy and the popularization of the concept of green travel,urban residents tend to travel more towards non-motorized vehicles.However,the increase in the use rate of non-motor vehicles and a series of related traffic violations caused by imperfect management regulations have exacerbated traffic congestion.In this paper,we adopt deep learning in the process of solving practical problems.Specifically,this paper proposes a non-motor vehicle license plate recognition algorithm based on convolutional neural network and innovatively proposes a front and back feature matching algorithm based on color features.This paper first preprocesses and structure the face data in the actual traffic images to facilitate subsequent specific work in terms of face recognition.Besides,this paper puts forward a distributed solution to improve the security and speed of obtaining data for the storage and acquisition of the actual face database.Secondly,this paper proposes a non-motor vehicle license plate recognition algorithm based on convolutional neural networks.Finally,this paper proposes a front and back feature matching algorithm based on the color characteristics of the clothes on the front and back of the driver.Each module verifies the superiority of its algorithm through experimental data.Finally,this paper includes a distributed pedestrian and non-motor vehicle violation detection system,because the data and system back-end deployment adopts a distributed way,which is more conducive to the follow-up of each module The system provides users with a simple and quick page management module.Traffic management personnel can issue tickets,identify violations,and display intersection screens.It also provides system users with rank management functions according to different departments.In terms of violation statistics,specific statistical information is found based on multiple query conditions.In terms of traffic visualization,the number of violations at each intersection is combined with the intersection map.And each module of the violation detection system has been specifically developed.