Research On Image Stitching Method For The Complicated Scene Of Coalmine Tunnel

The tunnel system is an important part of the mine system.Currently,under the background of the intelligent coalmine,the position of the tunnel in the coalmine is becoming more and more important.The image of coalmine is used for image stitching to obtain a panoramic view of the coalmine,which can do some basic research on the three-dimensional reconstruction of coalmine tunnel,remote visualization and the establishment of digital models.It can even guide production tasks such as coalmine excavation and comprehensive mining,which has a certain positive significance for the development of coal industry.In view of the complex scenes of coalmine tunnels,this article aims to improve the existing image stitching methods to overcome the problems in the scenes,mainly by constructing feature point directed line segment models to eliminate mismatched points to enhance the robustness of feature points.On this basis,two improvements are made: one is to improve image fusion method by seam-driven,for speeding up the speed of seamlines optimization and eliminating the ghosting problem that is easy to occur in the image stitching process;the second is to improve the grid in the image registration The method of meshing is based on the basic idea of different grid densities in different areas,which can find an optimal area density value to enhance the accuracy of image registration.The detailed research work is as follows:(1)A mismatch elimination algorithm based on the feature point directed line segment model is proposedIn order to better solve the problem of mismatching of feature points in the complicated scene images of coalmine tunnel and the projection distortion that usually occurs in the process of image stitching,a coalmine tunnel image stitching method based on directional line segment mismatch elimination is proposed.In the first,the SIFT(Scale Invariant Feature Transform)algorithm is employed for image feature extraction and matching to obtain a rough matching point pair.Then construct a directed line segment model of coarse matching point pairs of adjacent images,and use the direction and length attributes of the line segment to eliminate mismatched point pairs once.After that establish the directed line segment model and its direction label of the characteristic points in the respective images,and then perform direction matching on the directed line segments corresponding to the adjacent images,and use the probability statistical model to remove the mismatched point pairs twice to obtain the final fine matching point pair.Finally,establish an image grid model,use the AANAP(Adaptive As-Natural-As-Possible)algorithm to align and stitch the images,and use the weighted average method to fuse the images to complete the image stitching.Perform feature matching and image stitching experiments on coalmine tunnel images and four sets of public data sets.The proposed mismatch elimination algorithm has better real-time performance and higher matching point accuracy than the RANSAC(Random Sample Consensus)algorithm;In addition,the registration accuracy of the corresponding coalmine tunnel image stitching is higher,and the panoramic stitching image obtained is more natural.Experimental results show that the proposed algorithm is an image stitching algorithm with high accuracy and good stitching effect for the complicated scene of coalmine tunnel,and it has better robustness and availability.(2)An image fusion method by seam-driven is improvedAiming at the ghosting problem in the process of image stitching of coalmine tunnel and the low efficiency caused by the excessively large image pixels,based on the idea of dynamic programming,an image fusion method by seam-driven is proposed.In the first,the energy function of the pixel node is constructed using the color difference and gradient difference in the overlapping area of the reference image and the image to be registered.Then construct the mathematical model of the pixel node by analyzing the expression of the pixel node.After that propose a set of suture search methods on the basis of this mathematical model.And SSIM(Structural Similarity)was introduced to determine the energy function,and experiments were conducted on coal mine tunnel images and four sets of public data sets.Finally,experiments show that this method can successfully eliminate the ghosting or geometric dislocation in the image stitching process,and improve the image stitching result.(3)An image registration method with improved mesh division is proposed.In order to improve the accuracy of image registration and the effect of image stitching,based on the idea of region segmentation,an image registration method with improved grid division is proposed.In the first,obtain the panoramic image under the global homography matrix,and divide the different regions in the panoramic image;Then,set different mesh densities for different regions,and use the proposed meshing algorithm to divide the mesh;After that set the minimum grid size of different regions,and find the best combination of minimum grid size through comparative experiments;And then,under the best grid density value,a comparative experiment of image stitching is carried out to verify the effectiveness of the method in this paper.Finally,experiments were carried out on coal mine tunnel images and four sets of public data sets,and compared with the method in Chapter 3 of this article,it was found that the accuracy of the method has been significantly improved.In this paper,we have done some researches on the image stitching of complex scenes of coal mine roadways,such as removing mismatched points,optimizing seamlines,and meshing,which improve the accuracy of image registration and the effect of image stitching.Basic research work has been done for the application of coalmine tunnel images,which has certain positive significance for the construction of digital mines.The thesis includes 77 figures,15 tables and 88 references.