Object Detection, Location and Identification at Radio Frequencies in the Near Field

Recent events in the UK (eg the 2019 London Bridge Attack, in which 2 people were killed, and the Terror Related Streatham Incident, where 2 people
were stabbed) have highlighted the need for improved early stand-off detection of threats, which include knives, guns and improvised explosive devices. To be able to characterise and identify these small objects at stand-off distances in the order of 10s metres from the sensor using electromagnetic field measurements requires frequencies in the 300MHz to 12GHz range, where wave propagation effects are important. Frequencies in this range have also been traditionally been used in radar (radio detection and ranging) for large objects (eg ships, aircraft and air borne threats) over much larger distances from the sensor using far field scattering pattens. However, while radar is traditionally associated with the positioning and detection of objects in the far field, radar can also used be for the classification of objects in the near field (such as in autonomous vehicles, parking sensors, and ground penetrating radar (GPR) for finding landmines and unexploded ordnance, archaeological searches and the location of utilities for the construction industry). Furthermore, there is also considerable interest in improved object positioning given the development of autonomous vehicles by Google, Tesla, Uber and many others as well as related applications in autonomous manufacturing. In all these applications there is also considerable demand to improve the characterisation and identification of small objects that are not impeded by boundaries that can be penetrated by electromagnetic fields (eg walls, ground, clothing, smoke, fog or clouds).

This proposal is aimed at improving the characterisation, classification and identification of small objects in the near field using electromagnetic frequencies in the range 300MHz to 12GHz leading to new mathematical results, statistical computing tools for object identification and design recommendations for electromagnetic sensors. Our hypothesis is that a higher tensor description of an object combined with a probabilistic classification approach provides an effective means of identifying small objects using electromagnetic field measurements positioned away from the target, but in the near field, at wave propagation frequencies.

To test our hypothesis, we will derive new asymptotic expansions, which lead to new object characterisations in terms of new tensor descriptions. We will investigate new minimal contracted representations of objects using these tensors and understand the information about an object that can be obtained from these minimal representations. We will develop new computational tools for computing these characterisations and classifiers that build on a library of tensor coefficients to make object predictions from practical measurements.