The computer-vision-aided tool for object detection works on the basis of a machine-learning algorithm, which automatically detects and counts the trees.
As cities and towns embark on a rapid development drive at the expense of their precious green cover, researchers from Hyderabad’s Indian Institute of Information Technology (IIIT) have developed a method that would enable them to scientifically count trees and generate density maps. All that is required to generate a tree count and density map is to ride across the city with only a basic camera that will record videos and reveal the result within minutes.
The computer-vision-aided tool for object detection works on the basis of a machine-learning algorithm, which automatically detects and counts the trees. It then generates a colour-coded map of the city to highlight the extent of tree cover along routes. Arpit Bahety, a 23-year-old research fellow at the institute’s Center for Visual Information Technology lab, tested the tool using footage from a GoPro camera that was mounted on a helmet on the back seat of his scooter. A similar experiment was conducted in Gujarat’s Surat as well, this time using footage from a Samsung mobile phone. The system managed an 83 per cent accuracy during the trials in these two cities.
The researchers developed a machine learning model, which can detect trees at once. This platform also avoids duplication, a common error when many trees are manually counted. Speaking to Indian Express Online, Bahety said the camera can be mounted on any vehicle and it will detect tress. The only catch is that the camera has to be facing left. The camera will detect the tree based on the trunk. The video is then fed into a system that generates a map, enabling them to know the green cover of different areas. Over 50 trees a kilometre would be considered a good count, while anything lower than 20 would be low. The map denotes areas of good tree count with dark green those with a low count in black.
The team now hopes to employ the system in different cities to understand the model’s large-scale efficacy. It believes the tool could also help assess urban afforestation efforts.