AI

Near Real-time Visual Wildlife Tracking onboard UAVs

View the PDF file from the paper entitled Wildlive: Near the actual time tracking wildlife on board the plane, by Noguyen Nugco DAT and 9 other authors

PDF HTML (experimental) view

a summary:Direct wildlife tracking by processing high -resolution video directly on the unparalleled drone on a large scale, and most current solutions depend on broadcasting the video to the ground stations to support the mobility. However, both interacting air control of animal flying outside the visual sight and/or individual tasks and learn about the behavior of the mission to some extent depend on this ability. In response, we offer Wildlive- a frame near the detection of animals and follows it in the actual time of high-resolution images that work directly on the non-engraved air vehicles (drones). The system performs a multi -sided discovery and follows it at 17.81 frames per second for HD and 7.53 frames per second on 4K video flows suitable for operation during higher height trips to reduce animal disorder. Our Jetson Orin AGX system has been improved on the plane. It integrates the efficiency of tracking sporadic optical flow and taking the task samples with the techniques of detecting and retail the installed Yolo. Basically, the arithmetic supplier focuses on the spatial and temporal areas with high uncertainty to significantly improve drone processing speeds. Besides, we offer our Wildlive data collection, which includes animal cases consisting of 200 kilos thousand+ via 19K+ tires from the 4K Videos collected in OL Pejeta Conservance in Kenya. All tires contain the links to the ground truth, retail masks, as well as individual tracks and tracking points paths. We compare our system versus current-tracking methods including OC-Sort, Bytetrack and sorting. Multi -elegant tracking experiences with devices on the plane confirm that a high -resolution wild -time tracking is possible on drones while maintaining high accuracy levels as needed for future operating independence centered around animals and theme. Our materials are available on: URL https this address

The application date

From: Nogco Dat Nguyen [view email]
[v1]

Monday, April 14, 2025 12:21:16 UTC (46,010 km)
[v2]

Tuesday, 15 April 2025 12:06:09 UTC (46,010 km)
[v3]

Friday, 23 May 2025 12:10:33 UTC (46,854 KB)

Don’t miss more hot News like this! AI/" target="_blank" rel="noopener">Click here to discover the latest in AI news!

2025-05-26 04:00:00

Related Articles

Back to top button