Real-time method for computing focal length of camera from image sequence

Abstract

The invention provides a real-time method for computing the focal length of a camera from an image sequence, comprising the following steps of: selecting two adjacent images from the image sequence; matching characteristic points of the two images to obtain matching point pairs; solving any seven groups of matching points of the matching point pairs to obtain multiple possible solutions of the focal length; evaluating the accuracy of each possible solution, and using an evaluating value as the weight of each possible solution; computing the weighted sum of the possible solutions according to the weights for evaluation to obtain the probability density function of the focal length; sampling on a probability density function camber of the focal length for multiple times to obtain multiple sampling points; and fitting the sampling points by applying a Gaussian function, and using the peak value of the Gaussian function as a final focal length value. The real-time method for computing the focal length of the camera can be used for overcoming the defects of the self-calibration of the camera in practical application, thereby fast and accurately evaluating the focal length parameters of the camera.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a realization method for recovering a three-dimensional model of a scene from an image sequence. Background technique [0002] In the field of computer vision, people have long studied how to obtain camera motion and 3D models of scenes from image sequences. The initial application is reflected in machine vision. Researchers judge the spatial relationship of the scene through the geometric information extracted from the image sequence, and further guide the movement of the robot. This type of research usually requires the use of special and expensive acquisition equipment, and the application Scenes are also limited. With the development and popularization of hardware equipment, researchers hope to be able to calculate the 3D model of the scene from the sequence of images obtained by common acquisition equipment, and the key technology involved is the calibration of the camera. ...

AI Technical Summary

Problems solved by technology

Compared with the Kruppa equation, this method has stronger constraints, and can be solved by linear and nonlinear methods, which has improved operability and accuracy. The disadvantage is that the constraints requiring positive semi-definiteness in the solution process are often difficult to meet.

Images