Flossom Identification Of Lycium Barbarum Based On Convolutional Neural Network

Lycium barbarum is of great value in food and medicine,and has a large market in China.The flowering period of Lycium barbarum is an important node for the growth of Lycium barbarum,and the study of the flowering state of Lycium barbarum has an important role in the study of the growth state monitoring of Lycium barbarum.Ningxia wolfberry in China’s planting area is very wide,and at home and abroad have a very high visibility.In this paper,Lycium barbarum from Zhongning and Yinchuan of Ningxia Autonomous Region was taken as the research object,and the characteristics of different flowering stages of Lycium barbarum were extracted by using convolutional neural network method,so as to realize automatic recognition of the characteristics of flowering stage of Lycium barbarum based on surveillance video images.The deep learning technology is used to compare and analyze the artificial observation data and the images automatically observed by the real scene monitoring.The convolutional neural network based Lycium barbarum image automatic recognition algorithm is constructed.The accuracy of the recognition algorithm is verified and the continuous cycle iteration is carried out to complete the construction of imagebased Lycium barbarum Ningxia standard image database。The automatic identification of the flowering date information of Ningqi No.1 and Ningqi No.7 was realized,and the indoor observation of the growth of Lycium barbarum was explored,which was conducive to the discussion of what kind of conservation of Lycium barbarum should be conducted in the next step.In this paper,the Center Net model,a single-stage model based on key points and independent of anchor frame,was selected as the basic network to construct the growth model of Lycium barbarum in flowering period.Center Net model in the Hourglass-104 as testing works best when the backbone network,so we in the Center Net model is optimized when the Hourglass-104 as the backbone network,through the Hourglass-104 on the network by joining attention mechanism to improve the ability of feature extraction,so as to further improve the accuracy of detection,and the introduction of the depth of the separable convolution to reduce the online computation,so as to further improve the detection speed,The Flower Net model used to identify the flowering period of Lycium barbarum was obtained by improving the Center Net model.The final experimental results show that the Flower Net model can achieve good identification results in the detection and classification of the flowering period of Lycium barbarum.Compared with the Center Net model before optimization,Flower Net model not only improves the accuracy of the target identification of the flowering period of Lycium berry,but also improves the detection speed.The identification results were compared with Faster R-CNN model,YOLO model and Centernet model,and the results showed that the accuracy of the optimized Flowernet model in the identification of the small object of Lycium berry flowering period was higher than that of other comparison models,and the mean value Map of the average precision reached 78.9%.This indicates that the Flower Net model can be applied to the effective monitoring of the growth state of Lycium barbarum at the flowering stage.