Global spatial scale measurement method based on two vanishing points and a local object scale

Abstract

The invention discloses a global spatial scale measurement method based on two vanishing points and a local object scale, which comprises the following steps of: 1) selecting two groups of mutually orthogonal parallel lines from a building with a to-be-estimated scale in an image, and calculating coordinates of two vanishing points formed by the two groups of parallel lines; 2) calculating a rotation matrix and a translation vector calibrated by a monocular image camera through the two vanishing point coordinates and a real scale of a known line segment; and 3) converting a pixel coordinate into a world coordinate by using the rotation matrix, the translation vector and focal length information implied in the vanishing point coordinates, and calculating the estimated scale of a building object through an Euclidean distance formula according to a world coordinate and a topological relation of each point of the building object. According to the invention, the problem that it is difficult to find three groups of mutually orthogonal parallel lines meeting the requirement through calibration of three vanishing point cameras is avoided, the limitation on input images and required geometric constraints are less, and the method can be applied to an ancient building garden image, is suitable for more scenes and is high in generalization.

Description

technical field [0001] The invention relates to the technical field of scale estimation of building objects in monocular images, in particular to a global space scale measurement method based on two vanishing points and local object scales. Background technique [0002] Restoring the three-dimensional geometric structure of building objects in single or multiple two-dimensional images, that is, three-dimensional reconstruction technology, is the basis of computer vision, and it is also a research difficulty and hotspot in many frontier fields such as artificial intelligence and virtual reality. Among them, the 3D reconstruction based on a single image has become a key research direction of 3D reconstruction because it avoids the difficulty of feature matching in reconstruction of multiple images. Estimating the scale of buildings from a single image is an important task in 3D reconstruction. For the scale estimation of a building object in a single monocular image, the mono...

AI Technical Summary

Problems solved by technology

However, this calibration method based on the vanishing point has the condition that there are three sets of parallel straight lines that are orthogonal to each other in the image. This condition is difficult to meet in practical applications, and the application range is restricted.

Images