Birds getting hit by airplanes taking off or landing often makes headlines but there are other ways the birds succumb to death untimely. They get injured or succumb to injuries after getting hit by communication towers, windows, and different types of objects made by humans. However, not much is known about solar panels leading to the death of the winged creatures.
The solar panel setups cause the death of as many as 138,000 birds in the USA every year. When the birds collide with solar panels, the panels do not get much damaged. However, the majority of birds that hit the solar panels succumb to their death. To tackle this menace, the Department of Energy has given an amount of $1.3 million to the Argonne National Laboratory to set up an automatic bird activity monitoring system.
A software engineer working for Argonne, Adam Szymanski talked at length about the hurdles the team is coping with. The team is planning to develop a model for tracking drones that are already being used to monitor bird activities.
The best way to figure out the number of birds dying from a collision with solar panels is by counting the corpses. However, this means the researchers do not get accurate data. So, Szymanski and his team are working on a system that can make the process automated and more accurate. Szymanski says they have developed an advanced camera setup that can detect the moving objects automatically. Then it can categorize it as a bird or other flying object. Thereafter, the system can also classify the activity of identified birds. So, if the bird detected actually collided with anything or not can be detected by it. Advanced machine learning is deployed in it.
The study aims to come up with a sophisticated edge computing model. It will work within the imaging sensor rather than relying on offsite processing afterward. The researchers will make use of high-performance computing to tweak the efficacy of this advanced bird-spotting system. Szymanski says Argonne has a large HPC facility.
As of now, the team is using footage worth 10 hours a day from 8 cameras. This will be done for a month. Then, this footage will be segregated into five-minute clips, to be processed parallel. The data obtained by this method will serve as a machine learning model. All these are meant for setting up an edge computing model that will work within the camera. Using older offsite video processing methods will not produce expected results, say the team members. Doing the work in real-time using the camera is important.
However, the researchers are not only drawing on Argonne Blues. The bird detection system was actually developed earlier to trace flying drones. For a computer model, detecting a flying drone in the sky is not very different from catching a flying bird. In both scenarios, the computerized model is able to catch a random flying object.
There exist other systems for drone detection but they are not very accurate or efficient. This is owing to the fact such systems rely on drone tracing by using radio signals. However, not all drones use radio signals at all times. Szymanksi, therefore, created an advanced computing model that relies on upward-facing cameras to catch flying drones. He however points out that these 2 systems have their differences.
The drones and birds do not fly using the same mechanisms. The drones are laden with some type of propeller and that rotates when the machine is in the sky. So, it is easier to identify the drones using cameras. With birds, it can be a different story. Not all birds are the same size. Then, they flap their wings when flying or making a change in direction which is easy to spot. However, at times, the birds glide towards the ground, and then their wings are closed. At that time, it is hard to figure out it is a bird.
The new model for detecting birds is also different in other areas. The earlier drone model mostly was used to observe the empty sky. The new bird detection model will have cameras at solar plants aimed at the large solar panels and not so much at the sky. What this means is now the camera sensor has more objects to process and the detection process is harder. Szymanksi however is upbeat about it.
As a matter of fact, the number of birds dying after hitting solar panels may not be as high as other environmental calamities. However, there is no denying the fact that it is a project that strikes a balance between producing eco-friendly energy and saving wildlife.