Tea Bud Segmentation And Recognition Based On Machine Vision

Tea is a necessity in people's daily life.At the same time,tea also occupies a place in export commodities.Among them,famous tea is widely sought after by people for its lean processing quality and good tea quality.But the picking of famous and excellent tea is still in the stage of manual picking,and the low-efficiency picking of famous and excellent tea can not expand the production all the time,so a set of full-automatic equipment and theory about the picking of famous and excellent tea is still to be studied,and the most critical and difficult problem in the picking process is the identification and positioning of tea buds.In this paper,on the basis of synthesizing the specific environment of famous tea picking and the related progress of target recognition in recent years,based on the theory of machine vision,the problem of recognition and location of tea leaves is mainly studied.On the basis of a series of image preprocessing and color space transformation processing,the shape feature subset of tea leaves is established;the method of feature mask matrix is used to distinguish the young and old leaves,and Based on the calibration method of Zhang Zhengyou camera and the improved RANSAC algorithm,the 3D coordinates of the buds are extracted and located.The specific research work includes:First of all,because of the complex environment of the tea picking scene and the reasons of the camera itself,the collected tea image has low contrast and much noise.In order to solve the problem of low image contrast,this paper first preprocessed the tea image,combined with digital image processing technology,smoothed the image,and enhanced the edge of the tea buds.The bilateral filtering technology is used to process the image and filter out the noise of the image.The binary image is obtained by thresholding the image with Ostu method.Edge detection is used to extract the edge of the image.We get the information of tea shoots and optimize the image.Secondly,the color characteristics of tea shoots can bring a lot of information,so we first extract and separate the color information.The statistical method of gray histogram based on RGB channel shows that the color of each channel is relatively smooth,and direct segmentation is relatively difficult.Therefore,HSI and HSV color space are used to transform the original image,and color transformation and corresponding threshold range judgment scheme are used to obtain the position and basic contour of buds.After the analysis of a large number of experimental data,different separation thresholds are used in the transformed channels,and then the final result is the fusion of the data of each channel,which achieves a better segmentation of the bud area.Finally,on the basis of a series of image preprocessing and color space transformation processing,the shape feature subsets of tea shoots are established in the stage of leaf identification and positioning;the identification of buds and oldleaves is realized by using mask feature matrix method,and automatic segmentation is realized.Integrating the depth data of realsense depth camera,using Zhang Zhengyou camera calibration combined with the improved RANSAC algorithm to calibrate,obtain the three-dimensional spatial coordinates of the identified bud leaves,and then through the camera coordinate system to the mechanical arm coordinate system transformation,finally upload to the mechanical arm control system through the Ethernet interface.Thus,the automatic segmentation,recognition and location of tea buds are completed.The preliminary field test shows that the method has a good effect on tea buds recognition and location.