Research On Joint Ensemble Registration And Fusion Method For Muti-source Image

Joint ensemble registration and fusion,which aims to find the correspondence between multiple sets of source images and maximize the integration of relevant information,is a fundamental and critical problem in the field of image processing.Currently,it is widely used in medical science,remote sensing image processing and intelligent vehicle.Therefore,research on joint ensemble registration and fusion has high theoretic significance and practical values.To the best of our knowledge,there is a little literature dealing with joint ensemble image registration and fusion.In addition,most of the proposed methods focus on pairwise registration and fusion.Therefore,this thesis conducts research on joint ensemble registration and fusion of multi-source image.First,a joint ensemble registration and fusion method based on Student-t mixture model is proposed.Then,focuses on analyzing the true distribution range of image pixels,a bounded generalized Gaussian mixture model is introduced to characterize the joint intensity vector.Furthermore,in order to solve the problem of missed detections and low robustness in vehicle detection for intelligent vehicle,the joint ensemble registration and fusion method is introduced into the vehicle detection algorithm.The main contributions of this thesis are given as follows:1.The state-of-the-art algorithms and modeling methods for joint ensemble registration and fusion are given.The problem is summarized,and the motivation of this thesis is given.2.Propose a joint ensemble registration and fusion method for multi-source image based on Student-t mixed model.Considering the influence of noise and outliers,a joint ensemble registration and fusion method based on Student-t mixed model is developed to model the joint intensity vector,the state of the model is eatimated by expectation maximization algorithm.Computer simulations results on different datasets verify its effectiveness.3.A joint ensemble registration and fusion method based on bounded generalized Gaussian mixture model is proposed.The true distribution range of image pixels is considered comprehensively when describing the joint intensity distribution scatter plot,and a bounded generalized Gaussian mixture model is introduced to characterize the joint intensity vector.The fused image is obtained by estimating the model parameters.4.Experimental verification.To solve the problem of missed detections and low robustness in vehicle detection.The idea of joint ensemble registration and fusion is applied to vehicle detection algorithm in this thesis.Vehicle detection using multi-camera based on joint ensemble registration and fusion method is established,and the effectiveness of the proposed method is validated on different datasets.Consequently,the proposed methods have better performance than the state-of-the-art methods.The idea of joint ensemble registration and fusion is applied to improve vehicle detection algorithm.This thesis provides future directions for academic and engineering.