Identification Of Tea Sprouts And Their Picking Points Based On Deep Learning

Mechanically intelligent tea picking can ensure high efficiency and high bud and leaf integrity at the same time.The key to realizing mechanically intelligent tea picking is to enable computers to "recognize" tea buds and tea stems.In this paper,based on the target detection algorithm and image segmentation algorithm,the TB-Lai data set will be used to establish a tea sprout recognition model,a tea sprout classification model and a tea stalk recognition model.The main research contents of this paper are as follows:(1)Using the six algorithms of Yolov4,Yolov5,Yolox,Efficient Det,Faster r-cnn and Center Net to establish a tea sprout recognition model,the experimental results show that the performance of the tea sprout recognition model established by the six algorithms is Yolox,Yolov4,Efficient Det,Yolov5,Center Net,Faster r-cnn.The Precision,Recall,F1 score,and m AP of Yolox under the TB-Lai dataset are 89.34%,93.56%,0.91,and 95.47%,respectively.(2)Using the TB-Lai-Eq data set processed by color histogram equalization,the tea sprout recognition model was established with the same 6 algorithms.After comparing with the tea sprout recognition model established using the TB-Lai data set,it was found that Using color histogram equalization to process the tea sprout image can improve the accuracy of the model.(3)Using Yolox and Yolov4 algorithms,which have better performance in tea sprout recognition,combined with the TB-Lai data set,a tea sprout classification model was established.Through analyzing the experimental results,it was found that the tea sprout classification models established by the two algorithms were all wrong.In the case of classification,this may be caused by the similar color characteristics of the buds of some varieties.The m APs of the two models are 79.19% and 67.43%,respectively.The performance of Yolox in identifying tea buds is better than that of Yolov4.(4)Based on the DeeplabV3+ algorithm,TB-Lai is used to establish the identification model of the bud picking point.Through the experimental results,it is found that the recognition model of the bud picking point can identify the tea stalk,which proves that the mechanical intelligent tea picking is feasible.The m Io U,m PA,m Precision and m Recall of DeeplabV3+ are 78.43%,86.67%,85.89%,and 56.67%,respectively.The tea sprout identification model,classification model and tea stalk identification model established in this paper have achieved the identification,classification and picking point identification of the tender buds,which proves the feasibility of mechanical intelligent tea picking,and provides a basis for the application of mechanical intelligent tea picking in practice.provides some theoretical support.