Time change and disordered multi-image-based four-dimensional modeling method

Abstract

The invention relates to a time change and disordered multi-image-based four-dimensional modeling method, which is technically characterized by comprising the following steps: automatically extracting characteristic points for each image from which thousands of characteristic points can be extracted generally according to different resolutions by a sift algorithm, partitioning a three-dimensional point set to acquire separate buildings, calculating the depth information of each building on the image, and performing shading judgment by using the size of the depth. By carefully observing the change rule of the city, the buildings are integrally built or detached, so the buildings are used as basic units. The method has the advantage of improving the accuracy of image time sequence deduction by using characteristic descriptors to automatically match the characteristic points among the images and further abstracting a primary three-dimensional point model. Meanwhile, thousands of three-dimensions are abstracted into a plurality of buildings so as to greatly improve the efficiency of local search algorithm.

Description

technical field [0001] The invention relates to a four-dimensional scene modeling method based on time-varying and disordered multi-images, relates to the cross field of computer vision and virtual reality, and comprehensively uses related technologies of multiple disciplines such as computer graphics, computer vision and photogrammetry . This technology can be used to reconstruct the evolution process of a 3D real scene that changes over time, and it can also be further used to reflect the changes in urban appearance and the development of cultural relics. Background technique [0002] The image-based 3D scene modeling process mainly includes two modules: image matching and camera calibration, which are also current research hotspots, but the processing schemes for a large number of time-varying and disordered images are rarely involved. [0003] Schaffalitzky and Zisserman proposed a scheme to deal with the relationship between multiple images in the literature Multi-view...

AI Technical Summary

Problems solved by technology

[0006] In order to avoid the deficiencies of the prior art, the present invention proposes a four-dimensional scene modeling method based on time-varying and disordered multi-images to solve the problem of four-dimensional scene modeling based on disordered images

Images