Development Of Intelligent Tea Picking System Based On Image Processing

With the continuous improvement of people’s quality of life,more and more people choose to buy famous and excellent tea,and the trading volume of famous and excellent tea in the market keeps increasing.At the present stage,the picking of famous and excellent tea mostly relies on manual labor,with heavy picking task and short suitable harvest period,which seriously restricts the large-scale production of famous and excellent tea.The development of machine vision technology and agricultural robot provides a new idea for mechanized tea picking.Aiming at the difficulty of manual picking of famous and excellent tea,this paper developed an intelligent picking system based on machine vision and agricultural robot.The main tasks are as follows:(1)Complete the mechanical properties analysis of tea extraction.The texture instrument was used to pull the tea leaves from the tea branches,so as to simulate the way of manual tea picking,and measure the stress range required when the tea leaves of three kinds of tea tree sections were pulled,which provided theoretical basis for the load size of the intelligent picking device of famous and excellent tea in the later design.(2)Complete image collection and achieve segmentation of famous and excellent tea,and carry out two-dimensional labeling of famous and excellent tea.A tea image acquisition device was set up to collect tea group images,and the optimal segmentation threshold of famous tea and background was obtained through image enhancement,denoising,sharpening,color space conversion and other operations.The support vector machine classification algorithm was applied to the recognition of famous and excellent tea.By extracting H,S and V values of famous and excellent tea and background in HSV color space,the kernel function was used to raise the dimension of the data,and the famous and excellent tea and background were successfully separated.A minimum rectangular frame is used to wrap the segmented famous tea,and the two-dimensional coordinate of famous tea is replaced by the coordinate of the center point of the rectangular frame.(3)Establish the identification model of famous and excellent tea.Deep learning was applied to the recognition process of famous and excellent tea.Two convolutional neural networks,Fast R-CNN and YOLOV5,Used for famous and excellent tea identification.By comprehensively comparing the advantages and disadvantages of the two networks in the recognition task of famous and excellent tea,YOLOV5 convolutional neural network was finally selected as the recognition model of famous and excellent tea.(4)Obtaining three-dimensional coordinates of famous and excellent tea.The binocular stereo vision system was established,the binocular camera was calibrated,and the 3d coordinates of famous tea were successfully obtained by locating and measuring the target area predicted by YOLOV5 famous tea recognition model.(5)Build the hardware platform of famous and excellent tea intelligent picking and complete the writing of the control program.The hardware part includes the whole support of the famous tea picking device,the installation of SCARA manipulator,the connection of pneumatic circuit and the optimization design of the terminal collection device,and uses the obtained three-dimensional coordinates to pick tea leaves.(6)Carry out the picking experiment of famous and excellent tea.The identification and positioning experiment of famous and excellent tea was carried out on the intelligent picking platform of famous and excellent tea.The experimental results show that the recognition accuracy of famous tea is above 82%,and the positioning error is less than6 mm.It is feasible to pick famous tea with flexible-fingers.