Studies On The Tea Tree Living Real-time Monitoring Based On Computer Vision

Picking time of fresh tea leaves has a direct impact on the quality and yield of finished tea. From the perspective of picking time of fresh tea leaves, this study chooses two "leaves and a bud" in the fresh tea leaves as the research object. Using related theories and researches in computer vision technology, this paper proposes a method based on computer vision for real-time monitoring about growing of living tea trees. The method provides technical support for improving the information management system and intelligent level of operation of tea garden. It has broad application prospects.Through in-depth analysis and studies of current research, this paper formulates the technology roadmap of real-time monitoring system for the growing of living tea trees based on computer vision; designs the operational schemes of real-time monitoring system; chooses the hardwares and softwares of machine vision which are suitable for the recognition of fresh tea leaves; investigates the captured image edge sharpness and blocking phenomenon of fresh tea leaves at different angles and determines the best shooting angle is 45° (the angle between the shooting equipment and horizontal surface); Through comprehensive comparisons of processing development environment of various image, it selects efficient and convenient Lab VIEW 2014, IMAQ Vision and Vision Assistant 2013 as development softwares in the research.According to the color characteristics of the fresh tea leaves and background, this paper analyses the living tea trees using several common color models such as RGB, HSI, YIQ, YCbCr. By using image histogram analysis, the paper selects the improved G-B color factor to give the the color image a grayscale processing. It also selects median filtering algorithm to give the grayscale image of the living tea trees a filtering processing. It chooses the OTSU Method in adaptive thresholding segmentation processing of the grayscale image of the living tea trees. Based on the Mathematical Morphology, it selects opening operation and closing operation to give the fresh tea leaves a binary image morphology processing.Using the real-time monitoring softwares for the living tea trees based on computer vision, it collects and recognizes the sample datas of the living tea trees in the tea gardens, and the recognition rate is 90.3%. By selecting some related parameters such as covering area index, covering average length index and recognition index of tea foliage, it applies the discrimination principles and methods of "Bayes" and uses SPSS to give the 50 demarcate sample datas in the field a discrimination training. And it builds a discriminant model to predict unknown samples. There are three discriminant types:"Not picking", "Suitable picking", "Best picking". The accuracy rate is 98% after cross-examination.