A multi-camera global calibration device and method without a common view field based on a camera group

Abstract

The invention discloses a multi-camera global calibration device and method without a common field of view based on a camera group, and the method comprises the steps: employing the camera group to replace a long connecting rod between two plane feature point targets in a traditional method, and effectively solving a problem that the common space constraint is limited due to the connecting rod. Firstly, a method based on image ambiguity and noise intensity is adopted to carry out calibration image screening; Secondly, solving to-be-calibrated external parameters by using a hand-eye equation AX= YB; and finally, the internal parameters, the lens distortion parameters and the external parameters between the cameras are optimized at the same time, the global setting precision is effectivelyimproved, and a 3D target optimization function is formed by introducing polar line constraint and back projection errors, so that the problem of instability of multi-parameter joint optimization is effectively solved. The method is wide in application range, stable in mechanical structure, flexible and convenient to operate and particularly suitable for complex multi-camera space distribution such as long working distance and asymmetric working included angles.

Description

technical field [0001] The invention relates to visual measurement technology, in particular to a multi-camera global calibration device and method based on a camera group without a common field of view. Background technique [0002] Multi-camera visual measurement systems have a wide range of applications in many fields such as visual inspection, 3D reconstruction, panoramic photography, video surveillance, robot navigation, motion estimation, and augmented reality. All these applications often require the measurement of all cameras to be accurately unified in the same coordinate system to complete the inspection. However, an inherent problem is that the cameras may have non-overlapping fields of view, which makes it difficult to establish common spatial constraints, especially in some complex spatial distributions of cameras, such as long working distances (several meters to tens of meters), asymmetric working angle, or narrow working space, etc. Therefore, accurate exte...

AI Technical Summary

Problems solved by technology

[0022] From the above analysis, it can be seen that the existing multi-camera global calibration method without a common field of view, based on the method of high-precision three-dimensional coordinate equipment, can be suitable for large field of view cameras, long working distances, asymmetric working angles, etc., but It is difficult to operate when the working space is limited, and the cost is high, and the work efficiency is low; the method based on self-calibration is difficult to apply in industrial measurement due to the low calibration accuracy; the method based on auxiliary targets often relies on auxiliary Targets, flat mirrors, lasers, etc. This calibration method can solve some specific situations, but the operation is cumbersome and the accuracy is moderate; the method based on hand-eye calibration, connecting two cameras to be calibrated, is only suitable for situations where the cameras can move freely. In industrial online measurement, the camera often cannot move freely, and at this time it can only connect two targets, but when the camera has a large field of view or a small depth of field, or complex multi-camera spatial distribution such as long working distance, asymmetric clip In situations such as corners, it is difficult to operate fixed targets, and it is difficult to obtain effective and clear images, thereby reducing the accuracy of global calibration.

Images