| Identification of different wheat cultivars using two deep convolutional neural networks |
| Paper ID : 1163-NICAME1402 |
| Authors: |
|
Hossein Bagherpour *, Siavash _Shamohammadi Department of Biosystems Engineering, Faculty of Agricalture, Bu-Ali Sina University, Hamedan, Iran |
| Abstract: |
| Due to the existence of diversity in wheat cultivars, Choosing the right variety for cultivation in different soil and water conditions is very important. In order to have optimal performance in the cultivation of wheat, cultivars must be identified first and then cultivated in suitable environmental conditions recommended by experts. The similarity of appearance characteristics in wheat cultivars sometimes causes problems for experts in identifying wheat cultivars. On the other hand, the consequences of improper wheat cultivation will cause low yield and the loss of wheat basic varieties. Therefore, the main object of this study is to investigate the ability of Deep Convolutional Neural Network (DCNN) models based on image processing for the classification of four wheat cultivars in order to achieve high accuracy and speed for this task. Therefore, 400 RGB images of four Iranian wheat varieties (Hashtrod, Heydari, Mihan and Zarine) were prepared with the initial size of 2322*4128. From these images, the images of individual seeds were cropped and the total number of images reached 38120. Two deep layer networks Inception_Resnet_V2 and Inception_V3 were used to classify wheat cultivars. The results showed that the Inception_Resnet_V2 network performed better than the Inception_V3 network with an accuracy of 93.5%. |
| Keywords: |
| Machine vision, Classification, Wheat varieties, Deep layer network |
| Status : Paper Accepted (Poster Presentation) |
