Rice Growth Phase Detection in Rice Field Plots Using Drone Imagery

Abstract

This research aims to detect the growth phase of rice in paddy fields using drone images and the Mask R-CNN method. This algorithm excels in object detection and segmentation at the pixel level, thus enabling more efficient monitoring of rice growth compared to conventional methods. The growth phases detected include water, vegetative, reproductive, ripening, and bera phases. High-resolution drone imagery was used to identify growth phases in rice paddy plots. The evaluation results showed high performance in the vegetative phase with a Precision of 1.00, Recall of 0.77, and F1Score 0.87 and bera phase by achieving a Precision of 0.73, Recall of 0.89, and F1-Score 0.80, which was used to evaluate the accuracy of the model. The average mAP for the bounding box was 62.17, and the average mAP for segmentation was 65.84. This research shows the great potential of using drone imagery and deep learning in the automatic monitoring of rice growth phases.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint