کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
6458642 | 1421108 | 2017 | 8 صفحه PDF | دانلود رایگان |
- Early water stress in maize can be detected with the use of digital images captured in the field.
- The detection model achieves a mean accuracy of 90.39% for early water stress.
- The presented method can be applied to the automatic detection of water stress in practical agricultural irrigation.
Early water stress detection is of great significance in precision plant breeding and agricultural production. In the field, outdoor cameras would be an applicable tool for early drought stress detection with high-resolution images. Based on image analysis, we presented a model to detect water stress of maize in the early stage. In the red-green-blue (RGB) color space, a simple linear classifier was proposed to extract green vegetation from maize images. After color image segmentation, fourteen-dimensional color and texture features were extracted from each image. Three water treatment levels (well-watered, reduced watered and drought stressed) were applied to maize plants. We adopted a two-stage detection model trained with different feature subsets to evaluate the water stress. The water stress detection model was based on a supervised learning algorithm, gradient boosting decision tree (GBDT). The recognition accuracy of three water treatments (ATWT) was 80.95% and the accuracy of water stress (AWS) reached 90.39%. Results showed that the proposed method had an effective detection performance between water suitability and water stress conditions in the maize fields.
Journal: Computers and Electronics in Agriculture - Volume 140, August 2017, Pages 461-468