Design And Realization Of Binocular Machine Vision System For Tissue Culture Transplant Robot

With the development and wide application of plant tissue culture technology,there have been a lot of related studies on the automatic production equipment or robot of plant tissue culture technology.In the automatic production process of tissue culture seedling transplantation,a set of visual system is needed to identify and locate the tissue culture plant and the corresponding parts of tissue culture seedling in order to enable the machine to complete the cutting and transplantation of tissue culture seedlings.For this challenging problem,there is no universal and fully portable solution at present.Therefore,it is of great practical significance to study the tissue culture plantlet transplantation robot machine vision system and provide necessary technical conditions for the development of robots.This paper for laboratory research and development of Potato Tissue Culture Plantlets Transplantation Robot(PTCPTR)platform,using the principle of binocular vision,complete PTCPTR machine vision system was studied.Through the analysis of PTCPTR workflow and the actual demand for the vision system,a complete vision system scheme is proposed.In order to make the vision system cover the working field of view and meet the accuracy requirements of vision positioning,a follow-up vision solution is proposed.In order to meet the real-time requirements of the system and have better data processing capacity,the AM5728 multi-core heterogeneous processor of TI Corporation is adopted.In order to solve the problem of poor imaging quality of tissue culture seedlings in the test tube due to the reflection of light in the test tube,a light distribution scheme combining polarizing light source and polarizing mirror was proposed.In this paper,the camera calibration method and tissue culture node recognition algorithm are studied,and a complete vision software design is realized.The calibration method of binocular camera is analyzed,the internal parameters of the camera are calibrated by Zhang Zhengyou calibration method,and the attitude matrix of the camera is solved by BA(Bundle Adjustment)algorithm.The defects of the traditional algorithm of tissue culture seedling node recognition were analyzed and a new algorithm based on structure was proposed.The software of the vision system was designed,and the software system was modeled by UML(Unified Modeling Language).The key algorithms were optimized by multi-core parallel operation.The visual system GUI(Graphical User Interface),binocular vision follow-up control,intersystem communication,visual recognition and 3D reconstruction,camera calibration,ROI(Region of Interest)configuration,working point configuration and other functions were realized.In this paper,the prototype of the PTCPTR vision system was tested separately,and the combined test with the tissue culture plantlet transplantation robot platform was completed.The results show that the PTCPTR vision system can accurately identify tissue culture plantlets and their nodes,and get the accurate location coordinates of nodes.The obtained coordinates are transmitted to PTCPTR,and PTCPTR controls the corresponding actuator to complete the transplant operation and reaches the expected target.