Parameter optimization-based non-common field-of-view camera calibration method

Abstract

The invention discloses a parameter optimization-based non-common field-of-view camera calibration method, which comprises the following steps of: respectively acquiring a target I and a target II atthe positions of a camera I and a camera II, and changing the poses of the target I and the target II for multiple times to obtain initial data, wherein the initial data comprises image coordinates ofa calibration point in a target I acquired by a camera I and image coordinates of a calibration point on a target II acquired by a camera II; calculating an initial external parameter between the camera I and the camera II, an initial conversion relationship between targets, an initial rotation translation relationship between the target I and the camera I, and an initial rotation translation relationship between the target II and the camera II; and introducing calibration point deformation on the target I and the target II into a projection model of a camera, constructing a target function,and solving an optimal result by utilizing initial data and an optimization method to obtain an optimal conversion relationship. According to the method, the calibration point deformation quantity isused as the parameter to be optimized, the iterative objective function is solved, and the calibration precision loss caused by deformation is effectively improved.

Description

technical field [0001] The invention relates to the field of camera calibration, in particular to a method for calibrating a camera without a common field of view based on parameter optimization. Background technique [0002] When performing visual measurement of large-scale objects, it is often necessary to use multiple camera fields of view to cover the measured object. There is no common field of view between multiple cameras. At this time, it is necessary to obtain high-precision pose relationships between multiple cameras ( The main methods of global calibration of multi-vision sensors are: 1. Global calibration based on coordinate measuring equipment: this method uses electronic theodolite, laser tracker and other equipment and two-dimensional or three-dimensional targets as an intermediary for multi-sensor global calibration. The process is complicated, the need to introduce external standard equipment, the cost is high, and the calibration points do not correspond to...

AI Technical Summary

Problems solved by technology

[0002] When performing visual measurement of large-scale objects, it is often necessary to use multiple camera fields of view to cover the measured object. There is no common field of view between multiple cameras. At this time, it is necessary to obtain high-precision pose relationships between multiple cameras ( The main methods of global calibration of multi-vision sensors are: 1. Global calibration based on coordinate measuring equipment: this method uses electronic theodolite, laser tracker and other equipment and two-dimensional or three-dimensional targets as an intermediary for multi-sensor global calibration. The process is complicated, the need to introduce external standard equipment, the cost is high, and the calibration points do not correspond to the actual measurement points; 2) Global calibration based on dual targets: This method uses two targets (two-dimensional or three-dimensional) and performs rigid Connection, select a camera as the global camera, use the invariance of structural parameters between the two-plane targets, and calibrate one camera and the global camera each time, and finally obtain the structural parameters of each camera relative to the global camera to achieve global calibration. The method has the advantages of high flexibility and convenient implementation. In principle, the calibration points on the double targets are required to have no deformation. The calibration point above is deformed, that is, there is a deviation between the theoretical three-dimensional coordinates of the calibration point and the actual three-dimensional coordinates. This deformation error also makes the actual external parameters of the final camera extrinsic participation deviate, which affects the calibration accuracy.

Images