Panaji: A team from Don Bosco College of Engineering, Fatorda, in a project sponsored by ISRO, has developed an algorithm that enables accurate identification of land features like forests, waterbodies, etc, using satellite images. Unlike applications like Google Earth, the machine-learning algorithm even helps identify details like the type of crops being cultivated in a field.
The tool is expected to be immensely helpful in town and country planning, and in carrying out environmental studies, among other uses.
Rahul Kotru, Musab Shaikh and Satyaswarup Banerjee of the electronics and telecommunication (ETC) branch have developed the deep learning algorithm, under the guidance of lead scientist, Varsha Turkar, who heads the department, and Shreyas Simu.
This data can be captured during day and night independent of weather and climatic conditions. Every object on earth can be identified with its unique characteristics with the help of this data. Traditionally, this kind of work requires rigorous field work with human centric time-consuming processes. However, the technique developed by Don Bosco College’s students helps identify and demarcate these objects without human intervention.
Turkar had received a total of Rs 26 lakh in funding from ISRO as part of a four-year-long project, and a part of this sum was utilised in the development of this algorithm.
“When you open Google Earth, you can see any part of the earth up close. But the problem with the satellite taking those images is that they require the Sun as a source of light. One cannot take images during rainfall or when the weather is cloudy. We work in the microwave domain.
These kind of satellites send microwave signals, which can penetrate through any kind of weather,” said Turkar.
She said an algorithm is required to analyse the data, a technology for which is available in countries like Japan, at a high price.
“We were able to receive data for analysis free of cost because of the association with ISRO and IIT-Bombay. Work on such algorithms is being carried out elsewhere in India as well, our algorithm uses semantic segmentation. The machine learning algorithm analyses such images and can tell even the kind of crop being grown in a particular field. Microwaves have the ability to categorise each and every characteristic on earth,” explained Turkar, who has been working in the area since 2007 and has a PhD from IIT-Bombay.
Her team of students began working on the project from March 2020, and the algorithm is the result of their work over the last one-and-half years.
The project is part of NISAR, which is the biggest collaboration between NASA and ISRO.
“For a developing state like Goa, the algorithm will be a boon in disguise as it helps to identify and demarcate various features like settlement, forest, waterbodies, and mangroves, etc, from images acquired by the satellite. It may also prove helpful for town planning, civil engineering, environmental studies and agriculture,” said Kotru.
“With its very high accuracy of 95 per 100 samples, this model can be implemented across the globe successfully,” said Turkar. It may be noted that this model has already been tested for areas in San Francisco, Mumbai and Delhi.
Source: timesofindia.indiatimes.com
