Global calibration method with apparatus based on rigid bar for multi-sensor vision

摘要

The present disclosure provides a global calibration method based on a rigid bar for a multi-sensor vision measurement system, comprising: step 1, executing the following procedure for at least nine times: placing, in front of two vision sensors to be calibrated, a rigid bar fasten with two targets respectively corresponding to the vision sensors; capturing images of the respective targets by their corresponding vision sensors; extracting coordinates of feature points of the respective targets in their corresponding images; and computing 3D coordinates of each feature points of the respective targets under their corresponding vision sensor coordinate frames; and Step 2, computing the transformation matrix between the two vision sensors, with the constraint of the fixed position relationship between the two targets. The present disclosure also provides a global calibration apparatus based on a rigid bar for a multi-sensor vision measurement system. Putting the present disclosure to use can increase the accuracy of the global calibration, and also be suitable for the calibration of the multi-sensor vision system in the extremely large working space, which enlarges the application range of the present disclosure.

主权项

1. A global calibration method based on a rigid bar for a multi-sensor vision measurement system, comprising:
step 1, executing the following for at least nine times:
placing, in front of two vision sensors to be calibrated, a rigid bar fasten with two targets respectively corresponding to the vision sensors;capturing images of the respective targets by their corresponding vision sensors;extracting coordinates of feature points of the respective targets in their corresponding images; andcomputing 3D coordinates of each feature points of the respective targets under their corresponding vision sensor coordinate frames; and step 2, computing the transformation matrix between the two vision sensors, with the constraint of the fixed position relationship between the two targets,
wherein computing the transformation matrix between the two vision sensors includes:computing a rotation matrix and translation vectors;establishing an objective function on the basis of minimizing re-projection error;solving an non-linear optimal solution for the rotation matrix and translation vectors to acquire the transformation matrix between the two vision sensors, by using a non-linear optimization approach.