Research On 3D Reconstruction Technology Based On Multi-view Images

With the advancement of science and technology and the improvement of living standards,teachers and students often use different media tools to assist teaching or learning in the process of education and teaching.Usually media devices are the carrier of technology,the most critical of which is the technology itself.The media is the object that carries the information.The technology is the system knowledge of manufacturing a product,a process or a service provided.It can be seen that different functions and purposes can be realized by using different technologies,and then used by the teachers and students through the media.In addition,in the field of educational technology,students-oriented and teacher-assisted learning models such as flipping classrooms?problem-based learning,and game-based learning with highly value.When students are learning the complex spatial structure of the graph and the 3D model,the teacher is required to present the corresponding teaching content,otherwise the student may not be able to complete the learning task of the part of the content.At the same time,the advancement of technology has promoted the widespread application of digital models.In recent years,the demand for various 3D models has shown a sharp increase.How to build a more accurate 3D model as a key part of which also attracted the attention of more and more attention.The 3D reconstruction technology can automatically reproduce the 3D point cloud model of objects or scenes in the real world,and it is widely used in architectural design,medical rehabilitation,cultural relic protection,digital modeling and other fields.Image-based 3D reconstruction Using a series of multi-view images to generate 3D models is both convenient and low-cost.Image-based 3D reconstruction is easier to generalize than interactive parametric modeling and laser-based scanning.Aiming at the work of image-based 3D reconstruction in the past,by analyzing and comparing the basic ideas of existing 3D reconstruction technology and the advantages and disadvantages of related algorithms,a 3D reconstruction algorithm based on multi-view images is proposed,and the algorithm is optimized.Get the sparse 3D point cloud of the target with good accuracy and realism.In order to achieve the above objectives,this study first extracts feature points from a given multi-view image and matches the feature points of different images to obtain image grouping;then calculates the projective reconstruction for each group,and according to the increment The method is used to solve the global projective reconstruction.Finally,the projective reconstruction is restored to the metric reconstruction according to the camera self-calibration result,and the result is optimized by the minimization method to obtain the final sparse 3D point cloud.The 3D reconstruction has a large number of research methods and techniques in recent decades.Different experts and scholars have proposed various solutions to different research directions.However,so far there is no best,universal 3D reconstruction algorithm for all practical applications.It can be seen that how to design a purposeful and targeted algorithm still has great challenges.In order to robustly reconstruct the 3D model from multi-view images and improve the accuracy of the results,we propose an algorithm for reconstructing 3D point clouds from multi-view images.In this thesis,we design a 3D reconstruction method,which can reconstruct the 3D point cloud model of real-world objects or scenes more accurately for the generation of virtual objects or scenes.It also shows how to get the final result based on several steps.The experimental results show that the result of the algorithm has high accuracy and realism.