Unrestraint multi-camera picture matching method based on space-time consistency

Abstract

The invention discloses an unrestraint multi-camera picture matching method based on space-time consistency. The method comprises the following steps of reading pictures captured by a reference camera and an auxiliary camera; extracting characteristic points of a current picture; calculating the characteristic description vector of each characteristic point; performing similarity comparison of the characteristic description vectors and establishing matching relations of corresponding characteristic points; querying stable matching relations in the matching relations, and according to the number of the stable matching relations, screening characteristic point sets used for calculating a transformation matrix; according to the matching relation of the screened characteristic point sets, calculating the transformation matrix to obtain the matching relation between pictures of the two cameras; and reading a next picture and repeating the abovementioned steps till the end. During the screening of the matching relations, the space-time consistency is fully considered. The effective prior information is obtained by retrieving the stable matching relations and is used for screening the current matching relation, so that the stability of image matching between adjacent frames of the unrestraint camera can be effectively enhanced, and the matching precision and efficiency is improved.

Description

technical field [0001] The invention belongs to the technical field of image matching, and more specifically relates to an unconstrained camera picture matching method based on spatio-temporal consistency. Background technique [0002] In many applications such as multi-camera joint object detection, tracking, scene stitching and reconstruction, multi-camera image registration is required. The idea of ​​image registration in the prior art is to extract the images acquired by each camera at the same time, extract the feature points of each image, and then use the feature point matching method to obtain the optimal matching point set between images, and then use the point set to calculate the camera The projective transformation relationship matrix between them, and finally realize the registration between multi-camera images. However, the effect of this processing mode depends on a stable set of optimal matching points. When faced with applications such as unconstrained rela...

AI Technical Summary

Problems solved by technology

However, the effect of this processing mode depends on a stable set of optimal matching points. When faced with applications such as unconstrained relative spatial relationships between cameras, since the spatial positions of each camera change independently and freely, the registration relationship between cameras varies with time. When the time changes rapidly, the optimal matching point set between the images obtained by the camera will change greatly at the same time. The detection and positioning of the scene produces severe jitter, which is not conducive to the subsequent processing of multi-camera video images

Images