Neutrosophic approach based intelligent system for automatic mango detection

Due to its deliciousness, flavoring, and nutritional value, mango is considered one of the most common fruits among the generations. Mango growth occurs numerous times a year. In India, mangoes are abundantly harvested in the summer months, and traders ship them to a variety of markets. In this scenario, the Indian markets need the automatic identification and recognition of mango species, which also ensures their quality. Moreover, the mango species identification also focuses on structure, geometry, and texture as well. Advanced technologies comprise its efficiency in automatic recognition, however often fail due to the intra-class heterogeneity among millions of mango species worldwide. Deep Convolutional Neural Networks (CNN) are employed in this study to more precisely identify the mango species. The collected input mango image is initially pre-processed using basic functions like noise reduction and scaling. The image is then converted to the neutrosophic domain, and the defined neutrosophic set is divided using the Improved Fuzzy C-Means Clustering (IFCM) algorithm. The Geometric Mean based Local Binary Pattern (GM-LBP) and Local Vector Pattern (LVP) features are extracted as part of the feature extraction process. To differentiate the mango species, a Convolutional Neural Network (CNN) model is applied at the end. In order to confirm the efficacy of various measures, a comparison of the suggested (Convolutional Neural Network with an Improved Texture feature set (CNN + ITF)) with previous works is made.