Method for real-time detection of length and width size distribution of crystal population in glass crystallizer by binocular telecentric cameras

A method for real-time detection of length and width size distribution of a crystal population in a glass crystallizer by binocular telecentric cameras, wherein

• the method is implemented based on a non-contact high-resolution backlit calibration pole to ensure that two telecentric cameras synchronously collect images of a crystal solution in a reactor, thereby measuring the three-dimensional size of crystals;
• the backlit calibration pole is composed of a light-emitting diode (LED), a ceramic calibration plate printed with a checkerboard pattern, a reflecting prism, an optical fiber and a stainless steel sheath;
• a light ray emitted by the LED at the top of the backlit calibration pole is guided to the reflecting prism at the bottom of the backlit calibration pole through the optical fiber, and the reflecting prism reflects the light ray vertically to a back surface of the ceramic calibration plate engraved with the checkerboard pattern;
• reflected light provides illumination for the shooting of two telecentric cameras;
• the backlit calibration pole is inserted into the glass crystallizer and rotated clockwise or counterclockwise along a pivot to perform in situ calibration;
• and the backlit calibration pole can be assembled into different sizes to apply to the in situ calibration of the crystallizers with different volumes;
• the images are synchronously collected in real time by placing a telecentric camera at two different angles respectively outside the reactor;
• the images are called binocular images and comprise a left view and a right view;
• for the poses of binocular cameras, a two-step in situ stereo imaging calibration model is established, comprising calculating the closed-loop solution of camera parameters to obtain the initial values of model parameters and subsequently using optimization algorithms to seek optimal camera model parameters;
• for the collected binocular images, an analysis method for crystal image matching is proposed, comprising image preprocessing, point of interest detection, feature matching and mismatch removal;
• and after key length and width corner points are detected by using the contour features of all crystals, the three-dimensional reconstruction of the length and width corner points is conducted for each matched crystal through a calibrated stereo imaging model, so as to quantitatively evaluate its three-dimensional size.