Automatic Water Stress detection in wheat crop canopy using Chlorophyll fluorescence image dataset

Pre-processed Chlorophyll fluorescence dataset of wheat canopy images for automatic water stress detection using machine learning models. The dataset is comprised of file system-based information about the wheat variety Raj 3765. The analysis is done for the period of sixty days, twenty-four chlorophyll fluorescence images every day for each (Control and Drought) have been recorded. A total of (1440 x 2 (for both control and drought) =2880) images has been utilized. The size of the dataset was increased to (2880*20) for this study's endeavor using data argumentation techniques in order to develop the more generalized model.

The created dataset is subjected to different pre-processing pipelines comprising noise and contrast enhancement procedures. Pre-processed output images are subjected to novel segmentation algorithm called "Cfit k-means " to extract appropriate ROI which maximizes photosynthetic activity. The dataset and publications links are already shared in related links.

PSII (photosystem -II) colour features and GLCM correlation-based features are used for further analysis in the automation process.

The automation procedure involves the selection of an appropriate algorithm from the top 9 algorithms listed below for the said purpose.

1. Logistic Regression
2. Linear Discriminant Analysis (LDA)
3. K-nearest neighbours (KNN)
4. Decision Tree (CART)
5. Gaussian Naïve Bayes (GNB)
6. Support Vector Machine (SVM)
7. Extra Trees (ETC)
8. Gradient Boosting (GB)
9. Random forest (RF)