Fisheye camera internal reference calibration method based on single picture in vehicle-mounted surround view

Abstract

The embodiment of the invention provides a fisheye camera internal reference calibration method based on a single picture in vehicle-mounted surround view. The fisheye camera internal reference calibration method comprises the following steps: 1, splicing a plurality of two-dimensional checkerboard calibration plates to form a three-dimensional calibration object; 2, acquiring a single calibration graph of the three-dimensional calibration object through a camera; 3, acquiring pixel coordinates and world physical coordinates corresponding to checkerboards on each two-dimensional checkerboard calibration plate in the calibration graph; 4, solving an internal reference initial value of the camera through a linear model; and 5, obtaining the final internal reference of the camera through nonlinear model optimization. The fisheye camera internal reference calibration method has the advantages that a plurality of two-dimensional checkerboards are spliced to form the three-dimensional calibration object so as to achieve fast internal reference calibration of a single picture of the fisheye camera, and compared with a calibration method based on a plurality of calibration pictures of a two-dimensional calibration object in the prior art, the fisheye camera internal reference calibration method needs less manpower consumption on the premise of obtaining the same precision.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for calibrating internal parameters of a fisheye camera based on a single image in a vehicle-mounted surround view. Background technique [0002] With the development of automobile intelligence, more and more cars are equipped with a surround view monitoring system. The surround view monitoring system uses multiple fisheye lenses with a large field of view to collect information around the car and splicing it into a 360-degree surround view around the car body. To assist the driver to drive safely. In order to obtain a better surround view stitching effect, the internal parameters of each camera need to be accurately calibrated. The existing calibration methods are not suitable for mass production of products because they are labor-intensive and time-consuming. How to quickly and accurately calibrate the internal parameters of the camera is very important for m...

AI Technical Summary

Problems solved by technology

[0007] Calibration of camera intrinsic parameters based on two-dimensional calibration objects requires manual adjustment of the positions of the calibration board and the fisheye camera multiple times, and the collection of calibration board images from different angles of view. Need as many calibration board postures as possible to ensure the calibration accuracy, which requires a large labor cost, which is not conducive to mass production and shipment of products, and after-sales maintenance

[0008] Camera calibration based on three-dimensional calibration objects requires the production of high-precision three-dimensional calibration objects, which are usually difficult to process, high in cost, and not easy to reuse

In addition, the three-dimensional calibration object still has the disadvantage of self-occlusion, and it is not easy to obtain high-quality calibration images.

Images