Scientists develop AI-powered drone for automatic target identification

In the first innovation of its kind in the country, scientists have developed a drone-mounted software framework to enable automatic detection of objects like humans, bunkers and tanks from optical images and pinpoint their exact location.

The project has been undertaken by the Central Scientific Instruments Organisation (CSIO), a constituent laboratory of the Council for Scientific and Industrial Research, and took about two years to complete. The framework is meant to enhance UAV capabilities in surveillance and automation.

This is unlike the video or still feed from an airborne drone, which is viewed and analysed by controllers to locate objects or targets. Here the system itself identifies the desired objects based on the task at hand.

CSIO scientists said that besides defence and law enforcement applications, the system can also be used for civilian purposes such as disaster management as well as geological and cartographic surveys.

Referred to as ‘Vision-based Target Detection and Localisation’, the technology was transferred for commercial production and marketing to L&T, Mumbai, two weeks ago. Five companies had initially responded to CSIO’s proposal to transfer the technology.

The detection and identification of ground objects from images captured by aerial cameras is through sophisticated algorithms. Vision-based software determines the position of the object on the ground through the use of GPS.

Scientists claim that the positioning of objects is highly accurate, with the error of probability being less than a metre from a height of 250 metres.

The software module can be installed on any type or size of drone, depending upon the mission profile and endurance required. It can also be adapted for use in the dark through the use of infrared or night vision cameras.