The Optimal Combination Strategy Of Image Multi-Feature And Its Application

Image features play a key role in image recognition,target tracking,image registration and other applications,and are an integral part of the image processing field.Common image features include color features,texture features and local features.In the actual application process,a single feature may perform well in some specific scenarios,and when the scene becomes complex,it is very likely that using only a single feature will not meet the actual requirements.At this time,by combining various features effectively,combined with the characteristics of different features,even in complex scenes,it can also perform well.In the context of image registration,this thesis studies the optimal combination strategy of multiple features.At present,amount all the image registration methods,the feature-based image registration has the best effect and its application is the most extensive.Firstly,we extract the features in the two images,then match the obtained features of the two images,and finally calculate the coordinate transformation model,and obtain the image registration result by image interpolation.In this thesis,the image registration experiment is carried out by using multiple features,and the matching results obtained by combining different feature are compared and analyzed to find the optimal feature combination strategy.The main research contents and results of this thesis are as follows:(1)A variety of different image feature extraction algorithms are studied.The principle of each algorithm is introduced in detail.It mainly includes feature point detection algorithm and feature description algorithm,and analyzes the characteristics of different features and common application scenarios.(2)The theoretical basis of image registration algorithm is studied,the principle and flow of image registration algorithm are analyzed,the evaluation methods of image registration algorithm are studied,and the precision and accuracy calculation methods of image registration are determined as the evaluation standard of the experimental results of this thesis.(3)The image registration method based on multi-feature combination is studied,and a variety of different feature combination methods are analyzed.A double feature registration method and a triple feature registration method are proposed.They are used to find an optimal feature combination strategy by combining multiple features.(4)The multi-feature combination registration algorithm module and the registration result analysis module are designed and implemented.Under the Visual Studio software platform,the experimental platform is built by C++ language.The experimental data set is generated through simulation.A large number of tests are carried out using the experimental platform,and the experimental results are compared and analyzed.