Research On Stereo Image Matching Method Based On Multivariate Features

Stereo image matching is an essential step in the fields of robotics,aerospace,remote sensing,industrial automation and other technical fields.Image matching refers to the process of identifying the same feature between two or more images by a certain matching algorithm,that is,using matching criteria to achieve the best search.Stereoscopic image matching is to match the image of the same point in the real world in the left and right images,that is,using one of the frames as the reference(usually the left image),and determining the position of a point and finding the closest match to the point in the right image.The point,in turn,uses the position difference between each pair of points to obtain the parallax of the left and right images,and realizes the three-dimensional reconstruction of the image according to the parallax information.However,how to improve the performance of the algorithm under the premise of ensuring the matching accuracy becomes the bottleneck restricting the application of the technology.The application principle and technology of two-dimensional multi-features based on multi-features based on Harris-SIFT(Harris-Scale Invariant Feature Transform)and ICA(Independent Component Correlation Algorithm)in stereo image matching technology.Based on the Harris-SIFT stereo matching algorithm,the multi-scale Harris sub-pixel corner points are extracted.The improved SIFT feature descriptor is used to generate 32-dimensional feature vectors.The hash table is introduced to match the feature points with Hamming distance,and then the improved Prosac algorithm is used.Make the final match.The experimental results show that the proposed Harris-SIFT stereo image matching enhances the matching accuracy and improves the matching efficiency.Based on the robustness of the SIFT algorithm and the high precision of the Harris algorithm,high precision and high efficiency stereo matching is realized.It provides a useful reference for the development of stereo image matching technology.Based on the independent component analysis(ICA)stereo image matching,the ICA features of the left and right images are extracted.The ICA features represent the independent component features of the left and right images to remove useless information and noise,and achieve the purpose of reducing the image feature dimension.The experimental results show that the proposed ICA-based stereo image matching algorithm can obtain 36-dimensional features,which effectively reduces the matching time and error rate.Using the FastICA algorithm,36 32*32 feature detectors are trained in theimage library,and the image blocks at different positions in the graph are randomly selected to calculate the respective 36-dimensional features,and the matching blocks with the same features are searched in the right image.A disparity map is obtained by using the relative positions of the matching blocks in the left and right graphs.The two-dimensional multi-feature stereo image matching method based on Harris-SIFT feature and ICA feature proposed in this thesis combines the accuracy of Harris feature,the robustness of SIFT feature,and the anti-noise performance of ICA feature,under the premise of ensuring the matching accuracy,the feature dimension is effectively reduced,which greatly improves the matching speed,and provides theoretical and experimental basis for the development and application of stereo image matching technology.