Publication: IOT-based plant health monitoring using yolo model
| datacite.subject.fos | oecd::Engineering and technology::Electrical engineering, Electronic engineering, Information engineering::Electrical and electronic engineering | |
| dc.contributor.author | Cheong, Kuang Yen | |
| dc.date.accessioned | 2025-02-28T08:16:12Z | |
| dc.date.available | 2025-02-28T08:16:12Z | |
| dc.date.issued | 2023-07 | |
| dc.description.abstract | One of the important aspects in agriculture is plant health monitoring to prevent decrease in crop yield. However, manual monitoring is labour and time intensive. Although many automated plant health monitoring systems had been proposed and developed in literature, there is still room for improvement. This project aims to build an automated IoT system for monitoring plant health from leaves and monitoring its environment. In this system, the task of monitoring plant health was automated by using YOLOv4 and YOLOv5 object detection algorithm to detect plant health. The plant used in this study is eggplant plants. A total of 3 datasets is used to train custom YOLO object detection model. The classifications of plants health are based on the physical condition of the leaf. The classifications of plant health in this study are healthy, early stress and severe stress for first dataset. The classifications for second and third dataset are healthy and stress. The YOLO object detection was performed on Jetson Nano Board connected to webcam. Besides, the environmental parameters that affect the growing of eggplant such as soil moisture, temperature and humidity was monitored. Soil moisture sensor and DHT11 sensor was used to monitor these parameters. The YOLO plant health detection results, soil moisture value, temperature value and humidity value was sent to Adafruit IO IoT Cloud Platform for remote monitoring purpose. From training results and results of running YOLO model on Jetson Nano, YOLOv5 showed a better quantitative and qualitative performance than YOLOv4. Every YOLOv5 model achieve 100% detection accuracy on Jetson Nano. The results also showed that lesser number of classes and higher number of training images in dataset can increase the performance of YOLO model. Besides, the results also showed that the acceleration by TensorRT on YOLO model can improve the fps performance by around 1.68 – 2.05 times. | |
| dc.identifier.uri | https://erepo.usm.my/handle/123456789/21202 | |
| dc.language.iso | en | |
| dc.title | IOT-based plant health monitoring using yolo model | |
| dc.type | Resource Types::text::report::technical report | |
| dspace.entity.type | Publication | |
| oairecerif.author.affiliation | Universiti Sains Malaysia |