Analysis of three-dimensional positioning principle of binocular vision

Binocular stereo vision is a subject with wide application prospect. According to the idea and composition principle provided by CCAs, with the development of optics, electronics and computer technology, it will make continuous progress and become more practical. It will not only become the key technology in industrial detection, biomedicine, virtual reality and other fields, but also be used in aerospace telemetry and military reconnaissance And other fields. At present, binocular stereo vision technology has been widely used in production and life abroad.

            Three dimensional positioning principle of binocular vision

            The three-dimensional measurement of binocular stereo vision is based on the principle of parallax.

            Where base line distance b = distance between projection centers of two cameras; camera focal length is f. Let two cameras watch the same feature point (XC, YC, ZC) of space object at the same time, and obtain the image of point P on the "left eye" and "right eye" respectively. Their image coordinates are respectively pleft = (xleft, yeleft), right = (xright, yright).

            Now that the images of the two cameras are in the same plane, the image coordinate y coordinate of the feature point P is the same, that is, yeleft = yright = y, which is obtained from the triangle geometry relationship:

            Then the parallax is: leftrightdisparityxx  - 1. The 3D coordinates of the feature point P in the camera coordinate system can be calculated as follows:

            Therefore, as long as any point on the image plane of the left camera can find the corresponding matching point on the image plane of the right camera, the three-dimensional coordinates of the point can be determined. This method is a complete point-to-point operation. As long as there are corresponding matching points for all points on the image surface, they can participate in the above operations, so as to obtain the corresponding 3D coordinates.

            Three dimensional measurement of binocular stereo vision 

            The characteristics of binocular stereo measurement are similar to "binocular navigation and positioning", generally speaking: first, there are not many points to be extracted, just find the points of the part to be measured; second, the speed requirements are high, generally requiring multiple products to be measured in one second; third, the lighting environment requirements are high. Because there must be relative displacement between the camera and the measured object when measuring, the captured image may have different angles. The general implementation steps are as follows:

            First, camera parameter calibration. This part of the previous article has said that we will not go into details any more. If you are interested, you can go to see the relevant materials of CCAs binocular stereo vision. The main purpose of this part is to obtain the internal parameters and lens distortion coefficient of the camera, and obtain the stereo parameters of the dual camera at the current angle.

            Secondly, the subject is photographed and the feature points of the tested part are acquired. This is the most important part of binocular stereo vision measurement. The main difficulty lies in which algorithm should be used to obtain the feature points, and in this way of obtaining the feature points, we can't use the way of laser points or manually pasting the feature points to "make a trick", we must respect the original image. According to the experimental results of some image preprocessing algorithms provided by CCAs, it is found that more than dozens of preprocessing algorithms are needed to extract feature points, and the parameters of these algorithms need to be tested constantly. Because CCAs provides secondary development library, users can add some other algorithms in some extreme cases. The ultimate goal is to find the feature points accurately.

            Third, 3D coordinate acquisition. After the above steps are completed, stereo matching can be carried out and 3D coordinates can be calculated. This part is relatively simple, as long as the mathematical model is given directly into the formula. CCAs provides the corresponding algorithm model and routine.

            Fourth, specify the measurement range and output the results.