High-Precision Measurement of Binocular Telecentric Vision System With Novel Calibration and Matching Methods

Abstract

In stereo vision-based three-dimensional measurements, calibration, and stereo matching are the most challenging tasks for accurate three-dimensional reconstruction. The traditional binocular vision algorithm has low precision, and we propose a binocular vision system using a telecentric lens. In this paper, we propose a calibration and matching algorithm for the telecentric binocular vision system, which collects only two pictures of the calibration plate to complete the system calibration. The algorithm is a special application of 3D reconstruction. In this paper, if the measurement points are selected as points on the grid, the appropriate extension can be used for high-precision 3D reconstruction, which will be explained in detail in the experiments. In order to reduce the calibration error and obtain high maneuverability, the binocular vision system is only calibrated once with a two-dimensional calibration board, and the depth information is obtained through the parallax of the detection process, so as to effectively simplify the calibration process and reduce the errors introduced in the calibration process. For the matching of images in the right and left cameras, the polar theory of telecentric binocular vision system is constructed according to the characteristics of telecentric lens imaging model based on the traditional binocular vision system theory. The matching algorithm only needs to apply the two-dimensional calibration information and then calculates the results according to the parallax combined with the polar line matching algorithm, which simplifies the algorithm flow and reduces the error. In this paper, experiments show that the algorithm proposed is simple and has high precision, good stability, and high-practical value.