WiFi everywhere with Google Loon
Helium balloons are carrying WiFi connectivity to developing regions lacking network infrastructure
New software improves thermal camera identification capabilities.
Lost mountaineers and hikers are a common occurrence. A swift rescue is often a matter of life and death, especially if they have suffered a mishap or there is a danger of extreme temperatures.
In these cases, helicopters equipped with thermal cameras that detect people's thermal footprint are often used. At least in theory. In reality, on hot days, treetops and other areas can reach a human body's temperature, making stray people indistinguishable. One way to optimize tracking is to apply artificial intelligence and machine learning to interpret the images obtained. That is the proposal of three researchers at Johannes Kepler University in Austria. The results have just been published in the scientific journal Nature Machine Intelligence.
The system developers drew inspiration from clustered radio telescopes, whereby several images are merged into a single image. The first step of the artificial intelligence software used is to generate a single image obtained from the photos of a thermal camera installed on a drone or helicopter. The image is then processed to achieve a greater depth of field to distinguish the ground from the treetops. This is achieved by computing different focal lengths. In addition, the software has used a library of images to train the machine learning functionalities. To demonstrate its effectiveness in the real world, they used a group of volunteers who went into the forest. In tests, they have shown their system to be between 87% and 95% effective, in contrast to the 25% success rate of conventional thermal cameras.
Besides locating lost people, drones can also have applications in forest monitoring. One of the current trends is installing autonomous sensors in forested areas, some of them powered by triboelectricity. The problem is that they need to be installed manually, often in inaccessible areas. Now, at Imperial College London's Aerial Robotics Laboratory, they have turned to one of the new applications of drones to speed up the process.
In this new system, a drone can launch wireless sensors in the form of a dart into a tree's trunk or branches. The drones are equipped with cameras to identify the right spot. Also, the sensors incorporate a smart material that changes shape when heated and then adheres to the surface. Once installed, they provide valuable information about the ecosystems. On the one hand, they provide data on temperature, light, humidity, or animal movement. They are also useful tools for fire prevention, either by detecting risk situations due to low humidity and high temperatures or the first fire signs.
For now, the drones used require an operator, but the medium-term goal is to use autonomous drones that can go deep into the forest and locate the ideal spots to place the sensors. In the long term, it will be possible to create sensor networks in areas such as the Amazon jungle, where the density of vegetation makes operations very difficult.
These environmental protection technologies join other initiatives of great interest, such as hot air balloons operated with artificial intelligence or satellite systems to monitor the health of the oceans.