Analysis of binocular vision technology

The basic principle of binocular vision is based on the principle of triangulation. We place two industrial cameras side by side. Using the projection optical system, the field of view of the two industrial cameras can overlap at the required object distance. Through the pictures taken by each industrial camera, we You can capture scenes from different perspectives, as shown below:

CCAS binocular stereo vision model

Before implementing 3D measurement, we first need to determine where the surface points of the left industrial camera image are displayed on the right industrial camera image; second, we must know the associated pixel intersections of the left and right camera images. This involves two techniques: stereo matching and dual targeting.

1.Dual-targeting

Binary targeting is the process of initializing the parameters of the binocular 3D measurement system, that is, the parameters of the binocular system are measured using a known world coordinate system, which can be described by the following relationship:

Binocular vision measurement and calibration relationship

Camera calibration

Through the above relationship, it is easy to find that "image coordinates" are the basis of each calculation. In order to accurately calculate the coordinates, we need the camera to be an "ideal model", but in fact any imaging system has its distortion. The camera's distortion model is as follows. We can determine the camera's distortion coefficient through single-camera calibration.

CCAS binocular imaging distortion model

3. Stereo matching

Because the image obtained by a single camera can only calculate two-dimensional coordinates, because we use two sets of cameras, and the relationship between the two sets of cameras is also known, then if we can image a point in the three-dimensional space on the left and right cameras Both 2D coordinates are calculated, and you know that this is the same point, so you can calculate 3D coordinates. The technique for confirming the point with the same name is stereo matching. There are many algorithms for stereo matching, of which the local matching method is the most commonly used, but as far as the existing algorithms exist, no one algorithm can achieve 100% matching. Generally speaking, the more points to be matched, the lower the matching accuracy.

CCAS binocular stereo vision calibration results

4.Three-dimensional measurement

No matter what kind of matching method is used, the ultimate goal is to calculate the 3D coordinates of the points of interest to achieve our purpose of obtaining 3D information. At present, there are three common applications: three-dimensional restoration, three-dimensional navigation, and spatial tracking.