1. Algorithm: Waste Detection and Area Estimation Pipeline

To make the exact step by step process clear, here is how a frame actually moves through our processing loop:

1. Raw video was captured using the mounted camera on drone

2. A preprocessing algorithm was run adjusting the contract and sharpening the edges to that debris are easier to spot

3. The cleaned up image is bed into the YOLO v26 model

4. The model flags all the trash it sees and gives us the boundary coordinates for each item.

5. Instead of analyzing every item individually, we find the absolute outermost edges of all those smaller boxes to

6. The total are of the garbage patch is calculated

7. To visualize this in reality, we calculate how much physical space the camera actually sees, using its height and viewing angle

8. By comparing that physical ground space to the image dimensions, we figure out our exact scaling ratio.

9. We apply that ratio to convert the digital pixel count into a real world footprint measured in square meters.

10. We overlay that final physical area and the master bounding box directly onto the video feed.

11. The data is transmitted via MQTT to the ground station to update the live dashboard.