The Technology Research For Online Detection Of Silicon Rod’s Parameters Based On Machine Vision

Parameters detection is closely related to the means of detection. To solve the problems of online detection of the parameters for large size silicon rod, a non-contact and full size detection technology based on machine vision was presented in the paper. This technology overcomes the shortcomings in manual detection effectively, and also has the advantages of fast speed, high accuracy and non-contact. The main study includes the following in the paper:The composition of size visual detection system for silicon rod has been described and the principle of binocular vision has been analyzed. A size detection system based on the principle of orthogonal binocular has been designed and realized for silicon rod. Also the hardware’s parameters and selection for visual system are elaborated in detail. A principle prototype based on the principle of orthogonal binocular vision is set up. Meanwhile algorithm of the dimension detection has been designed and verification experiment has been completed for the size detection visual system.In order to inspect scratches, holes and missing angle defects in the surface of silicon rod, a defect visual detection system based on triangulation principle for silicon rod has been designed and the system selection has been finished. A monocular visual prototype for defect detection based on triangulation method has been set up. Also the verification experiment of defects detection visual system and the design of the algorithm for defect detection have been finished.The thinking of scanning detection has been proposed to the demand of full size detection. A stepper motor control system using DSP TMS320F2812 as microcontroller chip has been designed. The purpose of motor control system is to achieve the horizontal and vertical moving. The architecture of the motion control system, the hardware circuit and software algorithm of its sub-module all have been designed in detail. The algorithm correctness for sub-modules of driving and controlling system has been tested and verified.Finally, the experimental verification has been completed for multi-visual size detection and defect visual detection. Compared the detection results of the binocular vision system with the direct measuring results by vernier caliper, the two results are basically coincided with each other. And scratches and holes’ parameters information have been extracted in a piece of "smooth" silicon wafer by the triangle monocular vision system with the algorithm, while these defects can not be clearly identified through our eyes. The results show that the monocular vision system can realize the recognition of scratches and holes, and its recognition magnitude can reach to the level of μm. Through the two experiments above, it is verified that the design for two kinds of visual system is feasible.The visual system in the paper can realize the size and defect detection for silicon rod. Its size measuring accuracy is better than manual detection obviously. And the scratches and holes with the recognition level of μm can be detected without damaging silicon rod again. The designed motion control system can realize the full range of large size parameters detection when it is coupled with precision guide rail. The research results can also be used on other workpieces detection in the industrial field.