We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Utilizing satellite and UAV data for crop yield prediction and monitoring through deep learning.
- Authors
Mathivanan, Sandeep Kumar; Jayagopal, Prabhu
- Abstract
Agriculture is sighted more use cases of drones, and with the expanding population, food yields are becoming more well-organized. Drones are used in examining crops and exploiting data to determine what requires greater attention. This research study focuses on how deep learning (DL) has been used with drone technology to create solutions for detecting crop fields within a certain regions of interest (ROI). Extracting images from a drone and analysing them with a DL system to identify crop fields and yields for less-developed nations are solution to a prevalent challenge that land use–land cover (LULC) encounters. The limitations of drone spot-checking in the context of agricultural fields and the constraints of utilizing DL to detect yields. Also, a novel method is offered for detecting and tracking crop fields using a single camera on our UAV. The estimated background movements using a perspective transformation model given a sequence of video frames and then locate distinct locations in the background removed picture to detect moving objects. The optical flow matching is used to determine the spatiotemporal features of each moving item and then categorize our targets, which have considerably different motions than the backdrop. Kalman filter tracking has used to ensure that our detections are consistent across time. The hybrid crop field detection model is to evaluate on real uncrewed aerial vehicle (UAV) recordings. And the findings suggest that hybrid crop field detection successfully detects and tracks crop fields through tiny UAV's with low computational resources. A crop field module, which aids in reconstruction quality evaluation by cropping specific ROIs from the whole field, and a reversing module, which projects ROIs-Vellore to relative raw pictures, are included in the proposed method. The results exhibit faster identification of cropping and reversing modules, impacting ROI height selection and reverse extraction of ROI location from raw pictures.
- Subjects
DEEP learning; CROP yields; FIELD crops; OPTICAL flow; CROP quality; KALMAN filtering
- Publication
Acta Geophysica, 2022, Vol 70, Issue 6, p2991
- ISSN
1895-6572
- Publication type
Article
- DOI
10.1007/s11600-022-00911-7