Camera extrinsic calibration method, apparatus, computer equipment and readable storage medium
By extracting and matching road surface image feature points in the BEV coordinate system, and combining the constraints of parallelism and equal length, the camera extrinsic parameters are iteratively calculated, which solves the problem of low calibration accuracy of traditional camera extrinsic parameters and achieves high-precision calibration in natural road environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- FOSS (HANGZHOU) INTELLIGENT TECH CO LTD
- Filing Date
- 2025-12-18
- Publication Date
- 2026-05-26
AI Technical Summary
Traditional camera extrinsic calibration methods are difficult to establish an accurate calibration environment in after-sales calibration settings, resulting in low calibration accuracy. This is especially true in the dynamic calibration of side-facing cameras, where the vanishing point error is large and the roll angle cannot be accurately calculated.
By obtaining the set of BEV feature start and end points in the BEV coordinate system, and using the image feature points extracted from the road images continuously collected by the vehicle camera, combined with the constraints of parallelism and equal length satisfied by the set of BEV feature start and end points in the BEV coordinate system, the angular extrinsic parameters of the camera are iteratively calculated to avoid vanishing point errors caused by small angles between image viewpoints.
It improves the accuracy of camera extrinsic parameter calibration, enabling calibration in natural road environments, avoiding vanishing point errors caused by small image viewing angles, and accurately calculating pitch, yaw, and roll angles.
Smart Images

Figure CN122089846A_ABST