Design Of Object Detection And Segmentation Module For Vehicle Recognition

In recent years,computer information interaction has gradually changed from simple text information interaction to image information interaction through recognition.In the era of rapid development of computer vision,more and more companies are turning the achievements of image recognition technology into practical products and applying them to daily life,industrial applications and military activities.Vehicle target is an important military or civilian target.Recognition of vehicle target is conducive to improving the comprehensive ability of ITM and national defense strength.Vehicle recognition methods can be divided into image texture,image color,image contour and other recognition methods according to the characteristics of vehicle images.Cognitive psychologists advocate that human visual system is more accurate,more stable and more robust in recognizing target contour features than color and texture features.Therefore,object recognition based on image contour is a popular method in the field of image recognition.In this paper,the pre-processing method of vehicle contour extraction and denoising and the simplification method of vehicle contour feature are studied based on the contour of vehicle target image.Firstly,the smoothing filtering pretreatment of natural image and its application in edge preserving and denoising of image contour are studied.Firstly,the intensity and gradient function of image pixels are defined to control the structural difference between the filtered image and the original image,and the filter objective function is constructed.The objective function is optimized by variable splitting method and the final smoothing model is obtained.Finally,the altermating minimization method is used to solve the model,obtain the smooth image,and then enhance the color image.For smoothed natural images,better image edges can be obtained by edge extraction,which verifies the pre-processing effect of the algorithm in contour extraction.Then,the image segmentation method and edge detection operator algorithm are studied.Firstly,watershed segmentation algorithm,Graph Cut algorithm and Grab cut algorithm are introduced.Then,Canny operator,Sobel operator,Prewitt operator,Roberts operator,Kirsch operator,Laplace operator and Log operator are introduced.Sobel operator is improved by comparing them in many aspects.The edge lines of Sobel operator are refined with the accuracy.Finally,the gray level transformation,morphological processing and small area filling are described to get the edges of vehicle targets.Finally,an improved contour simplification algorithm for discrete curve evolution is studied.In order to reduce the complexity of contour description and matching,enhance the robustness of contour features to boundary noise,and consider the contour recognition features,a contour simplification algorithm based on discrete curve evolution is proposed.First,a threshold function which controls the evolution of discrete curves is defined to improve the discrete curve evolution algorithl.The improved discrete curve evolutionary algorithm is used to extract the features of contours,and the contours with important visual components are obtained,which simplifies the complexity of contour description.