Automatic Identification of Nanfeng Mandarin by Using a Stacked Autoencoder-Based Deep Learning Algorithm with Smartphone

EasyChair Preprint 667

Abstract

With the increasing deep learning applications in agriculture, identification of agriculture product of geographical indication by images had received remarkable attention from both academic and engineering fields. In order to facilitate the identification of Nanfeng mandarin, an identification method based on smartphone image and deep learning was proposed. In this paper, the research team proposed a classification scheme by using a stacked autoencoder based deep learning algorithm with three views of Nanfeng mandarin. As smartphone photography becomes more sophisticated, and the speed of a smartphone's internet connection fully supports real-time transmission of images, the image data can be easily recorded and quickly uploaded by the smart phone than other professional image equipment. A stacked autoencoder based deep learning algorithm was employed here for mandarin image classification so as to precisely recognize four type mandarins that were Nanfeng mandarin, Shaowu mandarin, Liucheng mandarin and Guangchang mandarin respectively. Experimental results indicated that the method based smartphone image and deep learning algorithm achieved a high accuracy of 89.45% for Nanfeng mandarin recognition than traditional identification methods. The method proposed in this paper may provide a convenient, fast and relatively accurate way for people to automatic identify Nanfeng mandarin. 

Keyphrases: Nanfeng mandarin, Smartphone, Stacked Autoencoder, automatic identification, image

@booklet{EasyChair:667,
  author    = {Zhiyong Xiao and Ya-Ru Liang},
  title     = {Automatic Identification of Nanfeng Mandarin by Using a Stacked Autoencoder-Based Deep Learning Algorithm with Smartphone},
  howpublished = {EasyChair Preprint 667},
  year      = {EasyChair, 2018}}