Vehicle Classification Based On The Vehicle Chassis

With the increasing of terrorist activities,a wide range of car bomb attacks occur frequently.Social security has become the focus of public attention around the world.So the security checks of accessing vehicle are very important.As carrying using car chassis is convenient,the car chassis becomes the focus in car security systems.The long-standing human detection,video surveillance and other measures can not meet the requirements now.Therefore,the development of real-time intelligent detecting foreign object system of vehicle chassis is imminent.Vehicle chassis identification is an important part of the real-time intelligent detecting foreign object system of vehicle chassis.This paper focuses on the general steps of the image recognition:Image segmentation,Image feature extraction,classification and recognition.And this paper focuses on the research of vehicle chassis image region segmentation,vehicle chassis image feature extraction,and the tactics of vehicle chassis design and so on.From the works real shooting imaging of vehicle chassis based linear array CCD,and combined the different characteristics of the images shot by buried camera,which have vehicle chassis parts and the non-chassis partially,and according to the signal changing with the chassis image content difference,this paper find images' corresponding boundary according the characteristic peak hopping of chassis image horizontal,vertical projection curve to completing the dividing of chassis image area.After that,against the problem of left and right offset of the cars in process engineering,this paper conducts image registration according to the maximum of their correlation coefficient between projection curve.For the local image content tensile and compression because of uneven vehicle speed and linear CCD imaging,this paper removed row of image data which has high similarity with their adj acent rows to suppress the distortion of.Experiments show that the registration of the chassis image,the suppression of distortion of travel,improves degree of similarity within the classes and is very beneficial for subsequent identification and classification.Finally,this paper extract the vertical projection of the curve of the edge of vehicle chassis image as classification and identification's features.Standard image features of the standard library is also stored like this.Then according to the degree of similarity between vectors and the degree of dissimilarity mathematical theory,for example correlation coefficient,Euclidean distance,etc,this paper strikes the correlation parameters of characteristic curves between the chassis which is needed to be measured and the standard image.And in order to complete the test image with the standard library features all the images this papers conducts eleven comparison.Through a lot of experiments this paper identifies the policy rules of the chassis image classification and find a vehicle chassis models' category or unusual circumstances.Experimental results demonstrate the effectiveness of the proposed algorithm in this paper.