Vision-Based Wilted Plant Detection

Plants wilt in most situations, whether outdoors or an office due to water shortage or poor exposure to sunlight. Overwatering, climate changes, pests, disorder, moisture, and fertilizer issues are all common causes. Our proposed model uses various image processing, feature extraction, and classification techniques. The Model makes use of ORB and SIFT for extracting features from a custom-made dataset of the wilted plant. The feature vector dimension optimization is performed using K-means and PCA. These features were provided to a machine learning model. The proposed system can play a major role in providing information about the plant's health status. This solution will help homemakers, offices, and hotels with indoor plant watering requirements, by recognizing the wilted plants. Four classifiers namely random forest, AdaBoost, decision tree, and logistic regression are compared. The trained model accurately classifies and detects wilted and healthy plants. The combination of SIFT with a Random Forest classifier gave the highest accuracy of 85.41%, comparable with the existing models.