This project demonstrates a proof of concept towards solving the challenge of proper plant monitoring and adequate data collection on large farms. By using a drone programmed to fly along a specified path, the drone transmits the video stream to a laptop for processing via wifi and the following steps are carried out;
-
Object detection is performed on each frame of the video stream to locate the crops by drawing bounding boxes.
-
Next is Object Tracking which associates a unique id to each crop and also keeps track of the number of crops on its path i.e crop counting.
-
and lastly the system takes a snapshot of the detected crops for data collection and further examination by farmer
- Flight Performance.
- Max Flight Distance: 100m
- Max Speed: 8m/s
- Max Flight Height: 30m (without wifi repeater)
- Camera.
- Photo: 5MP (2592x1936).
- Format: JPG(Photo); MP4(Video)
- FOV: 82.6°
To get enough images for this task, we converted frames from a 5 minutes video taken by the drone to get about 150 images, which was then used for training and validation. Data Annotations was done using LabelImg.
🏽Custom Object Detection using Yolov4.
The object detection algorithm used was the YoloV4 which offers great training speed and impressive performance.
🏽Multiple Object Tracking Using DeepSort.
To prevent taking multiple snapshots of the same crops, a unique id has to be assigned to each detected crop. DeepSort is the state of the art for object tracking.
Obafemi Awolowo University Teaching and Research Farm.
The AI Guy's Object tracking implementation using Tensorflow