The Improvement Of ICP Algorithm Research Based On Feature Points Guidance

Image registration is a key subject of graphics,point set registration technology is an important branch of image registration,the point set registration technology in many industries is showing strong vitality,it is one of the key technologies of robot navigation,satellite remote sensing,medical and other fields.As in the field of robot navigation,point set registration technology is a prerequisite for robot navigation,it is necessary to ensure the robot to determine the route in an unknown environment;in the field of medicine,point set registration technology is the auxiliary means of medical treatment,it can improve the accuracy of lesion localization in the treatment process.Because of the importance of point set registration technology,this paper puts forward the corresponding solution algorithm for clipping image and noise on the basis of studying the traditional iterative closest point algorithm(ICP).The main contents of this paper are as follows:ICP algorithm is an excellent algorithm for point set registration.It achieves good results in registration of rigid body points.It is a widely used point set registration algorithm.However,the traditional ICP algorithm on the selection of corresponding points is simply the two most points for the corresponding Euclidean distance of the nearest point,so there will be a corresponding point error,for cutting the image registration results poor performance.Feature point is an important research direction in image registration technology.This paper introduces feature points into ICP algorithm to improve the ICP registration algorithm,and studies the feature points in the ICP algorithm to achieve accurate registration of cropped images.In the process of shooting and imaging,images often suffer from noise due to various influences.In this paper,an improved ICP registration algorithm based on feature points is introduced to realize the accurate registration of trimmed images.Good results can be obtained in noisy images,but the presence of noise will have a great impact on point set registration and feature point registration.In order to solve the influence of noise on the matching results,the bidirectional registration method is used to weaken or even solve the noise.Bidirectional registration,as its name implies,is the simultaneous registration of two point sets.In the experiment,two sets of experiments are carried out simultaneously.In the second experiments,the location of the data set and the model point set in the first group are interchanged and registered.In this way,the noise can be extracted as the feature points and the possibility of being chosen as the final matching point is greatly reduced.The influence of noise on the experimental results is weakened,and the registration accuracy under noisy conditions is improved.