Drones Can Automatically Track Targets in The Forest

Austrian researchers have built an AI-powered drone that can follow moving things through thick forest. So much for retreating to the woods and surviving off the land if the robots ever attack.

Research paper:

While detecting and tracking moving targets through foliage is difficult (and often even impossible) in regular aerial images or videos, it becomes practically feasible with image integration …This finding together with the implementation of an initial drone-operated camera array for parallel synthetic aperture aerial imaging allows presenting first results on tracking moving people through dense forest. Besides people, other targets (e.g., vehicles or animals) can be detected and tracked in the same way.This might impact many application domains, such as search and rescue, surveillance, border control, and wildlife observation.

There are several methods for tracking moving objects through thick vegetation, including FLIR and other thermal optics devices, but this adds a novel AI-powered twist: the ability to follow color changes due to occlusion.

One of the most difficult difficulties in the AI industry is teaching an AI to figure out what’s going on in a picture when a large quantity of crucial information is obscured. According to the Journal of Remote Sensing, the researchers created a system capable of tracking opaque objects in motion in real-time using a method known as “color anomaly detection.”

Background: The team developed a lightweight 1D camera array that takes aerial photos of densely forested regions using a drone. Color anomaly detection is used in traditional aerial photography methods to scan photos for groupings of pixels that do not match the real surroundings.

The team’s contribution, on the other hand, was to construct a hardware and machine learning combo that integrated old methodologies with their novel aerial camera system to produce something capable of detecting color abnormalities and following them across blocked regions.

Quick take: This has the potential to have immediate and significant advantages for search and rescue missions. Because of its minimal computation needs and power overhead, this is a system that may be sent to remote places of the globe rapidly and placed into operation. I can imagine this saving the lives of lost hikers or aircraft accident survivors stranded in isolated locations.

It might potentially have a significant impact on conservation efforts if combined with GPS tagging and other tracking activities.

However, it is clear that something like this might transform a Predator drone, for example, into something capable of following targets across obstructed environments (such as packed city blocks or jungles) even if it loses contact with its human operators or has its communications blocked.

Do you want to learn more about this topic?

Comments

Popular posts from this blog

PyTorch and Albumentations for Image Segmentation

US$1 Billion Market Cap! Facial Recognition DeepGlint

Unlock the Power to Create Wealth Through ChatGPT Learning