| China is the birthplace of tea,and it is the first country in the world to grow and drink tea.In view of the low efficiency and high cost of traditional manual tea collection,as well as the development and application of machine vision and image recognition technology,tea collection is moving towards mechanization and intelligence.At present,the traditional tea picking machines used in the market adopt the "differential picking" method,which still needs to be manually selected after picking.Therefore,it is necessary to study an intelligent tea picking robot that can be picked selectively,and use vision to guide the robot to realize intelligent picking.This paper proposes a tea bud recognition and positioning system based on RGB-D camera.The system uses Intel’s Realsense d415 camera to obtain color images and depth images,and performs image processing through NI’s LabVIEW vision software.The main research contents are as follows:Recognition of tea buds depends on the difference in color and size of buds and background(old leaves,stems,soil)to be recognized on color images.Perform gray-scale processing on the color image in RGB,HSL,HSV,HSI color space to find out the color components suitable for segmentation;use linear filtering method or non-linear filtering method based on convolution calculation on the gray image to remove noise and find out Appropriate noise removal method;according to the gray image histogram,compare the segmentation effect of the image global threshold segmentation method and the image local threshold segmentation method on the detection target,find the most suitable threshold segmentation method;base on the segmented binary map Morphology and advanced morphology operations and particle feature analysis,remove areas that are too large and too small,leaving the right target particles,and finally achieve the identification of buds.In terms of tea bud positioning,the two-dimensional coordinates of the bud center of mass are determined according to the analysis of particle characteristics,and the two-dimensional feature points are brought into the depth image.The depth data obtained from the depth image realizes the three-dimensional positioning of the bud;camera calibration is performed to eliminate The camera is distorted and the camera parameters are obtained,and the threedimensional coordinates can be converted into coordinates in the world coordinate system to prepare for the subsequent grabbing experiment.The research results in this paper show that the RGB-D-based tea sprout identification and positioning system can more accurately identify tea sprouts and locate the sprout positions,with a recognition accuracy of 92%,which meets the design requirements of intelligent tea picking robots. |
