Research And Development Of Parallel Automatic Tea Picking Machine Control System Based On Machine Vision

Tea has its own unique properties and is deeply loved by human beings.Whether it is the Silk Road in ancient China or the new shipping routes opened up in modern times,tea has always been an important cargo,circulating around the world,and precious tea is favored by the world.The harvesting of tea leaves consumes a great deal of human and material resources as well as time,and these factors have limited the momentum of China’s tea industry to move even further forward.Tea picking technology needs to be constantly iterated and updated,and it is even more important to differentiate the traditional"one standard"tea harvesting machinery and design new types of intelligent tea pickers.In the research of this dissertation,a complete set of work operation system needs to be designed to realize the picking of tea sprouts more efficiently and harmoniously.The tea picking robot has been designed and implemented with a"hand-eye"co-operative operating system combining machine vision and Delta parallel mechanics.Adaptive calibration-free visual servoing through neural network machine learning and efficient and stable research on the picking path in the"hand"system.The main research elements are as follows:First of all,according to the demand of tea bud picking and the important basis of robot design,this dissertation designs and develops a"hand-eye"cooperative control system for a parallel automatic tea picking machine."Hand"system is designed,to achieve repetitive bud collection work and a quadruple-of-freedom Delta parallel mechanism.The actuator is designed for tea particularity,and the robot can work in the field through the crawler.In the"eye"system,a"global-local"camera arrangement scheme is designed,which realizes the data processing of tea images and provides prerequisites for tea picking.Secondly,according to the current research status of hand-eye relationship,the research basis of"hand-eye"cooperative operation is expounded,and a hand-eye calibration technology suitable for Delta parallel mechanism is proposed.In this dissertation,the"hand"unit has"3+1"degrees of freedom,and the main motion is translational motion.Therefore,based on this,a"hand-eye"connection with translation as the main element and a combination of translation and rotation is established.In the tea robot,the hand-eye unit of the"hand"unit and the"eye"unit work together to avoid the situation that the rank of the coefficient matrix is not satisfied,which makes it difficult to calculate.Then,according to the particularity of tea picking and comparing some shortcomings of visual servoing,a non-calibration visual servo system strategy for tea bud picking was studied,which constructed a new task function based on the image of tea buds.,which has better performance in noise processing of tea images.The workflow of the system is established,and a neural network-based convergence unit is designed to act as a vision controller to replace the inverse matrix of the interaction matrix to avoid the singularity of the solution.A fuzzy logic（FL）based motion controller is proposed to allow the gain to change with the situation,increasing the convergence speed of the system based on the L₂parametres of the task function and its derivatives as inputs during each tea picking cycle.Furthermore,according to the jerk optimization principle of trajectory planning in tea picking,an optimal velocity curve model based on time and jerk is proposed.The advantages and disadvantages of various speed curves are expounded,and the adaptive improvement of the harmonic jerk model is proposed,which is improved to a jerk curve with 15 stages,which avoids the shortcomings of other curves.A particle swarm optimization algorithm suitable for the particularity of tea picking was designed,and the optimal time interval parameters in the model were solved by using it,and the trajectory period of tea picking was optimized.Following the kinematic constraints,the time and jerk of the picking trajectory are comprehensively optimized,and a good trajectory smoothness is maintained.Finally,according to the needs of the research,an experimental prototype platform of robot upper and lower computers with"hand-eye"collaborative work was built,and a host computer control software suitable for the picking system of the prototype platform was developed.According to the particularity of picking,reflecting the principle of"hand-eye"coordination,the hardware of the platform is designed and selected.The upper PC is used as the main body of the"brain"unit,and the motion control card is used as the motion control center of the"hand"unit.Implement the"global-local"camera installation scheme.The software of the picking platform system is studied,its functions are developed,the cooperative work of the"hand-eye"unit,the work of uncalibrated visual servo,and the jerk optimization experiment of trajectory planning are verified.