LOCATION TRACKING SYSTEM OF DUCK FLOCK IN RICE FIELD USING COMPUTER VISION METHOD BASED ON VIDEO AND GPS FROM DRONE
Integrated agriculture (crop cultivation integrated with animal cultivation) is one of the solutions to the future food crisis. Unfortunately, agriculture in Indonesia still uses traditional farming technology. For this reason, a technological breakthrough is needed that can trigger the entry of var...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Subjects: | |
| Online Access: | https://digilib.itb.ac.id/gdl/view/77806 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Integrated agriculture (crop cultivation integrated with animal cultivation) is one of the solutions to the future food crisis. Unfortunately, agriculture in Indonesia still uses traditional farming technology. For this reason, a technological breakthrough is needed that can trigger the entry of various latest technologies. One of the problems in integrated farming that needs to be solved is the constraints of duck herders in finding flocks of ducks when the rice is tall and dense. Moreover, the size of the rice field area is no longer in the order of hundreds of square meters, but 10 hectares or more. Another obstacle also arises when the duration of herding that starts from morning to evening the ducks are generally only left once during lunch break. For this reason, the development of technology was initiated so that ducks can be found accurately and measurably. The technology developed is expected to be the basis for technology to monitor ducks remotely. In this research, the author designed a software prototype of a duck flock location tracking system used to detect the presence of duck flocks in rice fields from previously stored drone flight videos. In this research, TensorFlow Lite is used as a computer vision framework and DJI Mini 2 drone as an example of drone type. The computer vision model developed in the system has a mAP value of 24.92%. From the results of testing the system on two test videos, the system can find a flock of ducks that have been validated by checking three detection validation threshold parameters with measurable and accurate detection results according to the exact GPS coordinates and the difference between the position of the drone and the duck is the farthest by 13.81 m.
|
|---|