Projects:2017s1-110 ‘Real-Time’ FPGA Based Object Recognition & Threat Detection in Hardware

Project sponsored by elmTEK Pty Ltd in concert with Defence Science Technology Group (DSTG)

Students: Aaron Panella & Stephen La Vista

Supervisors: Dr. Danny Gibbins & Dr. Braden Phillips

Abstract

Students, sponsored by elmTEK Pty Ltd and the Defence Science Technology Group (DSTG), investigated the development of a threat detection system for use in the next generation of Australian Army land vehicles, being developed as part of LAND 400. The proposed system will be similar in function to Missile Approach Warning (MAW) systems, already in use by the Navy and Air Force, and is required to have low Size, Weight, Power & Costs (SWaP-C) and be rugged enough to be deployed in mobile land vehicles.

The project investigated the development of an image processing system utilising a high framerate Infrared (IR) camera to capture fast-moving threats, and a Field Programmable Gate Array (FPGA) which enables the development of highly parallelised image processing algorithms and helps achieve ‘real-time’ operation. Threat detection algorithms were investigated by one student, while the other investigated image registration algorithms for stabilisation.

Background

Next generation platforms of the Australian Army future land force expect a level of threat warning and situational awareness comparable to that enjoyed by the Air and Sea domains, whilst demanding significantly reduced size, weight, power and expense. In collaboration with the Defence Science & Technology Group, elmTEK is investigating the trade-off in performance associated with low cost sensing for Electronic Warfare applications, whilst maintaining the sensor performance to detect and classify incoming threats and augment situational awareness to crew. This project will continue the work of elmTEK interns over the summer by building threat detection and object recognition modules onto the current Image Processing Framework.

FPGAs are an exciting technology that are becoming increasingly popular for applications with high volumes of data, due to their ability to compute and process in parallel. This is a distinct advantage over conventional processors. Students will perform research into real-time image processing with FPGAs, efficient hardware architectures for FPGAs, performing mathematical operations efficiently in hardware and proper memory and clock usage for hardware architectures

Work Breakdown

Both students will work on hardware/framework changes. Image Registration will primarily be done by one student, with the other assisting, as a means to achieve image stabilisation. Threat Detection will be done by the other student.