China has developed AI technology capable of intercepting advanced stealth fighters such as the F-22 and F-35.
Beijing Institute of Technology, in collaboration with the China Academy of Air-to-Air Missile Research, recently created a lightweight AI model applicable to missile-mounted infrared imaging systems, according to the South China Morning Post on August 14. The research findings were published in the Chinese journal 'Journal of Electronic Measurement and Instrumentation.'
Infrared-guided missiles track targets by detecting heat emitted from aircraft engines and surfaces. Stealth fighters also emit heat, making them targets for infrared missiles. To counter this, aircraft release high-temperature flares to disrupt missile tracking. The research team developed an AI model that distinguishes between the infrared characteristics of heat signals from aircraft and flares.
The team tested the AI model using 3,245 infrared images collected from three types of aerial targets, including aircraft mimicking the F-22 and F-35, as well as loitering munitions. Simulation results showed that the AI model achieved a target identification accuracy of 97.1%. In actual hardware tests, the accuracy was recorded at 96.4%. The average inference time was just 1.5 milliseconds, with total power consumption at approximately 2.2 watts.
Notably, the researchers significantly reduced the size and computational requirements of the AI model for missile integration. They lowered the number of parameters by 16.1% compared to existing methods and reduced the required computational load by 19.2%. This allowed for high accuracy and fast processing speeds while adhering to strict size, weight, and power consumption constraints for missiles. The researchers stated, 'The lightweight AI model can provide rapid identification while maintaining high identification capability, making it widely applicable for future air-to-air missiles.'
However, the team noted that specific data used for identifying fifth-generation fighters cannot be disclosed due to confidentiality issues. They plan to continue research to further enhance identification accuracy and processing speed.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.