Researches On High-quality Tea Flushes Identification For Mechanical-plucking

Nowadays, the plucking of high-quality tea flushes still relies on manual work,which conveys the problem of labor shortages and high labor costs. Therefore, thetea flushes plucking robot with the ability to pluck the flushes automatically andselectively urged to be studied. Besides, identifying the tea successfully is thepremise of intelligent tea plucking.This paper mainly focuses on the research of spring tea, which is part of theresearch on tea flushes plucking robot. It provides a strategy of identifying the teaflushes automatically, and the main research is organized as follows:During the research of tea flushes identification, video camera is utilized tocollect tea images under natural conditions. After a continuous exploring and study,choose B,I,b,S and Cb to gray the orign images. Image binary segmentation algorithmis discussed and compared to find out the proper way for gray-scale imagesegmentation. The median filter is used to get rid of the noise. The area of connectedregions is obtained by using mathematical morphology, in order to remove the regionwhich is large or smaller than the requirements of the tea flushes, and then realizethe goal of the tea flushes identification. Finally, the centric method is presented todetermine the center position of the tea flushes.During the process of tea flushes identification based on spectral feature,FieldSpec HandHeld is used to measure the spectral data of old tea leaves and teaflushes in both laboratory and outdoor. Experiments showed that698-712nm can beused to distinguish the characteristic bands of the flushes and old leaves,700nm canbe used as the characteristic wavelength point to distinguish between tea flushes andold leaves.The innovation points of this paper are as follows:(1) It is the first time tofocus on the algorithms of identifying the tea flushes based on color image analysis.Experiments showed that R-B, I, b, S and Cb indices were particularly effective fortea flushes identification, it is feasible to segment the flushes from the backgroundusing these five indices.(2) The characteristic wavelength point, which is able todistinguish the old leaves and flushes is searched through spectroscopy.Based on related basic knowledge such as digital image processing technologyand spectroscopy, this paper has achieved the identification of flushes under thenatural environment. Furthermore, it offers support on technology for the design ofintelligent tea plucking machines.