A Study On Authentic AI Recognition Of Wolfberry Fruit Based On Appearance Image

As an important traditional Chinese medicine resource,Lycium barbarum has been planted in Ningxia,Inner Mongolia,Gansu,Xinjiang and other places in China.Affected by geographical factors,there are significant differences in the quality of Lycium barbarum from different places.However,there is no efficient and accurate way to identify the origin and quality of Lycium barbarum.Therefore,it is of great significance to establish a convenient,rapid and practical method for the identification and quality analysis of Lycium barbarum.In this paper,we collected the digital images of wolfberry samples from different areas and varieties,used convolution neural network deep learning algorithm to extract and select the appearance image features of wolfberry from different areas and varieties,and established a model through feature training,which was based on the appearance features of wolfberry from different areas,and established a rapid,simple,high accuracy and authentic online detection method,Finally,the authentic online recognition of Lycium barbarum was realized.(1)In this paper,we collected Lycium barbarum from Baiyin of Gansu,Jiuquan of Gansu,Yumen of Gansu,Qinhuangdao of Hebei,Xingtai of Hebei,Baishan of Jilin,Hetao of Inner Mongolia,Nuomuhong of Qinghai and Jinghe of Xinjiang.The main varieties of Lycium barbarum are zhongningqi No.1,zhongningqi No.5,zhongningqi No.7,zhongningqi No.9,golmuningqi No.1,zhongningqi No.5,zhongningqi No.7,golmuningqi No.9,golmuningqi No.1,golmuningqi No.1,golmuningqi No.1 Qinghai golmuningqi No.5,Qinghai golmuningqi No.7,Qinghai golmuningqi No.9 and other varieties of Lycium barbarum were photographed.In order to improve the recognition ability of Lycium barbarum and non Lycium barbarum,this paper also collected images except Lycium barbarum,and used convolution neural network to recognize Lycium barbarum.(2)Before training the convolutional neural network model,the image database is preprocessed,including image segmentation and data enhancement.Then the image database is randomly divided into training set and verification set to train and verify the model.By changing epoch,batch size and learn rate parameters,the optimal prediction model is found,and the algorithm is further improved,After repeated training,an optimal training model is obtained.(3)In order to develop a simple,fast,cheap,safe and suitable online identification and detection method of Lycium barbarum,an image-based online identification system of Lycium barbarum is developed,which realizes online identification of Lycium barbarum through web page or app program.The specific operation process is as follows: taking photos,uploading pictures,displaying the results of Lycium barbarum barbarum identification,The processing time from submission of Lycium barbarum pictures to authentic recognition was3.3145 ± 6606 s,the results can be obtained.Through the browser or app,the appearance pictures of Lycium barbarum to be detected can be submitted to the website for authentic online identification,and the origin results can be timely fed back to the client for display.This prediction method overcomes the defects of the previous analysis steps,which are complex,time-consuming,expensive instruments,high cost,and not suitable for on-site rapid detection of Lycium barbarum samples in the market.The effective and accurate differentiation and identification of the origin and varieties of Lycium barbarum products on the market can be achieved,and the accuracy rate of prediction can reach 96.00%.The accuracy of on-line identification verification can also reach 96.00%,and the results are consistent.For the quality control of wolfberry,regulate the circulation of the market,better promote the development of wolfberry.