Illegal Manned Identification Algorithm Of Agricaltual Vehical Based On Deep Learning

With the development of economy,agricultural vehicles like tractors and tricycles are playing more and more important roles in the countryside.However,because of the phenomenon of illegally carrying people,the traffic accident of agricultural vehicles often leads to heavy casualties.Therefore,this paper proposes an illegal manned identification algorithm of agricultural vehicle based on deep learning.Through the recognition of agricultural vehicles illegally carrying people and license plates of these vehicles,it makes a contribution to reduce the phenomenon of vehicles illegally carrying people.The specific research contents of this paper are as follows:1.This paper proposes a YOLOX-based recognition algorithm for us to locate agricultural vehicles and agricultural vehicle license plates.Then,this paper proposes two methods to improve YOLOX for the special scene of illegal manned identification on agricultural vehicles.First,the position parameter loss function used by YOLOX is the IOU loss function,which is changed by the generalized IOU loss function in this paper,so that the network can converge faster and better.Then,YOLOX uses non-maximum suppression,which is changed by soft non-maximum suppression,so that the network can do not filter some boundary boxes that should not be filtered,so as to reduce false detection.Finally,experiments verify that YOLOX can be used to recognize agricultural vehicles and license plates of agricultural vehicles,and the improved YOLOX has fewer missed detection compared with the unimproved one.2.In view of the inaccurate position detection of YOLOX in recognizing license plates of agricultural vehicles,this paper proposes a license plate preprocessing algorithm based on CNN to eliminate the noise in license plate images.Then,based on Conv Ne Xt network,the Conv Ne Xt-ST network is proposed and applied to license plate recognition of agricultural vehicles.Finally,experiments verify that compared with other end-to-end license plate recognition algorithm without pretreatment,the agricultural vehicle license plate preprocessing algorithm plus agricultural vehicle license plate recognition algorithm in this paper can better recognize the characters of agricultural vehicle license plate.