Research On 3D Reconstruction Matching Method

Image matching is a key step in 3D reconstruction.Although the algorithm of 3D reconstruction matching is improving and optimizing,its matching performance and efficiency are difficult to meet the requirements.Therefore,the research of 3D reconstruction matching method has been one of the hotspots in computer vision research,especially the feature-based matching method has important practical significance.One of the difficulties of feature matching is the feature extraction method.In 3D reconstruction,whether it is used in simple image or complex image feature extraction,how to find an excellent feature extraction method to extract special image features is an urgent need to solve the problem of image matching.On the one hand,this paper firstly compares Harris,Forstner and FAST operators for feature extraction of simple images,and gives their performance in feature extraction through detailed analysis and experimental test of each method.The experimental results show that Harris has good performance in extraction performance and efficiency.On the other hand,for the feature extraction of complex images,due to the large computational complexity and low matching performance of the SIFT algorithm,this paper proposes an improved method to extend the subdimension of the ke y point description by using a fine molecular region.Improve the performance of matching,then reduce the key point vector dimension through PCA,and improve the efficiency of matching.This method is called MYSIFT method.The experimental results show that the MYSIFT method can overcome the effects of scale transformation,rotation transformation and illumination transformation on image matching better than SIFT algorithm,and it can reduce the storage space and matching time.It provides a new method for the practical application of 3D reconstruction of complex images.