基于Tabnet的蘑菇毒性识别与应用

打开文本图片集

Application TabNet in Toxicity Identification of Mushrooms

LIU Ning123， LI Shu14， SHAO Changsheng1³， WANG Qi1，4 ， HUANG Qing³*

（1.Hefei Mushroom Vally Innovation Research Institute，Hefei 23ooo， China; 2.CollgeofMaterials and Chemistry，AnhuiAgriculturalUniversity，Hefei23o36，China;3.InstituteofInteligentMachines，Hefei InstitutesofyicalScience，ChineseAcademyofSciences，Hefei230o31，China; 4.EngineeringResearch CenterofChineseMinistryof Education for Edible and Medicinal Fungi， Jilin Agricultural University， Changchun 130l18， China）

Abstract： Mushroom poisoning is one of the significant challenges in the field of food safety， and traditional identification methods rely on expert experience or complex chemical analysis，limiting their eficiency and accuracy. In this study， the TabNet model was introduced and used to take advantage of its adaptive feature selection and end-to-end training，and the TabNet model wasconstructed and trained based on the mushroom dataset publishedby the UniversityofCalifornia.At the same time，thecharacteristic analysis further revealed hat odor was a reference value with higher discrimination，and the accuracy of toxicity discrimination was improved by combiningthe characteristics such as spore printing color.In adition，an online toxicity detection system based on TabNet is developed，which alows users to realize seamless connection of toxicity discrimination by inputting features，and realizes real-time toxicity discrimination.A variety of mushrooms were verified with anaccuracyrate of over 98% ，demonstrating the wide applicability of the model to the toxicity of mushrooms in different regions. The system can quickly and accurately analyze and report the toxicitystatus of mushrooms， which further enhances the practical application value of the model.

Keywords： artificial intelligence; classification system; TabNet; poisonous mushrooms; smel; online system

蘑菇是一种广泛存在于自然界中的真菌，在食品与医药领域具有重要价值，但部分毒蘑菇与可食用蘑菇形态相似，易引发误采误食，对人们生命安全构成严重威胁。（剩余5388字）