Research On Mrthods Of Close-range Image Segmentation And Step-by-step Image Pair Matching

In recent years,compared to other 3D modeling methods,3D reconstruction based on 2D images has the advantages of easy acquisition of color texture,low cost,good flexibility,and strong environmental protection,it can reconstruct a 3D model quickly and realistically,therefore,it has a better prospect of application.Although the 3D reconstruction method based on the calibration image can establish a high-precision 3D model,it is difficult to popularize the calibration material in the practical application scenario with the help of fixed geometric features and precise calibration,which is difficult to generalize,and the application of uncalibrated images to 3D reconstruction based on feature matching is simple and feasible,and has better application prospects.Image segmentation and image pair matching are the most important part in the process of 3D reconstruction based on uncalibrated images,the accuracy of matching the same name points has great influence on the quality of 3D modeling,therefore,the research of image segmentation and image matching methods has important theoretical and application values.In this paper,based on the effect of the quality of the original image on the effect of image segmentation,the image is enhanced when the original image is too bright or too low.By comparing the histogram equalization algorithm,gamma conversion algorithm,logarithmic transformation algorithm and Laplace operator method,it is found that low-luminance images are more suitable for strengthening with histogram equalization algorithm,and the high-brightness images are most suitable to use gamma transform algorithm for illumination compensation processing,the picture is more in line with human visual recognition system,which lays a solid foundation for image segmentation.On the basis of the mean shift segmentation algorithm,this paper uses the layered structure of Pyramid to segment the image after the enhanced processing,which makes the image segmentation effect better and improves the accuracy.In the segmentation process,the changes of three parameters in the function can have different effects on the result of image segmentation.By comparing the experimental results of image segmentation to select the most suitable parameters for the experiment,the problem of single parameter affecting image segmentation is solved.Aiming at the problem of image matching,this paper proposes a method of step-by-step to matching of homonymy points of image pairs.Firstly,the polygon obtained by image segmentation is used for overall matching to obtain the matching result of the overall frame in the image pair.Based on this,the image edges are applied.The extracted vector lines are partially matched,which reduces the search range and improves the matching accuracy.The experimental results show that by selecting the best image segmentation and edge extraction algorithm and combining the image pair matching method can improve the accuracy of the image matching,it can lay a good foundation for improving the quality of 3D modeling.