A camera focal length and pose unified calibration method based on multi-feature fusion

CN122156323APending Publication Date: 2026-06-05CHINESE PEOPLES LIBERATION ARMY UNIT 63660

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner: CHINESE PEOPLES LIBERATION ARMY UNIT 63660
Filing Date: 2026-02-14
Publication Date: 2026-06-05

AI Technical Summary

Technical Problem

Existing camera calibration methods fail in environments with weak textures and few features due to insufficient numbers of single features, and suffer from defects such as multiple solutions, unstable accuracy caused by nonlinear iteration, and low computational efficiency.

Method used

A multi-feature fusion method is adopted, which utilizes prior knowledge of camera position to transform the constraints of different types of features such as points, lines, and vanishing points into a unified vector model. The focal length is calibrated by the invariance of the vector angle between coordinate systems, and the attitude calibration is completed based on rigid body transformation, thus establishing a unified calibration framework.

Benefits of technology

It significantly reduces the requirement for the number of individual features. The focal length calibration process is a linear quadratic equation, which is simple and fast to solve. The pose calibration logic is rigorous and unique, which improves the numerical stability and computational efficiency of the calibration and enhances its applicability and universality in weak texture environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122156323A_ABST

Patent Text Reader

Abstract

The application provides a camera focal length and attitude unified calibration method based on multi-feature fusion, and belongs to the technical fields of computer vision, SLAM and photogrammetry. The application aims to solve the problem of calibration failure caused by insufficient number of single features in a weak texture environment. The method converts point, line and epipole features in an image into known vectors in a world coordinate system and vectors containing only unknown focal length f in a camera coordinate system by using a known camera position. At least two features are selected, and a linear equation about f is established by using the constraint that the angles of the vectors in the world system and the camera system are equal, so as to linearly solve the unique focal length. Then, the known vector pairs are constructed into virtual three-dimensional point pairs, the rotation matrix and the translation vector from the world system to the camera system are solved by establishing a new coordinate system and calculating the rigid body transformation therebetween, and the attitude calibration is completed. The application can realize efficient calibration without multiple solutions by using only two features, and has strong adaptability.

Need to check novelty before this filing date? Find Prior Art