Richard O. Lane

Maritime anomaly detection and threat assessment

Ships involved in commercial activities tend to follow set patterns of behaviour depending on the business in which they are engaged. If a ship exhibits anomalous behaviour, this could indicate it is being used for illicit activities. With the wide availability of automatic identification system (AIS) data it is now possible to detect some of these patterns of behaviour. Monitoring the possible threat posed by the worldwide movement of ships, however, requires efficient and robust automatic data processing to create a priority list for further investigation. This paper outlines five anomalous ship behaviours: deviation from standard routes, unexpected AIS activity, unexpected port arrival, close approach, and zone entry. For each behaviour, a process is described for determining the probability that it is anomalous. Individual probabilities are combined using a Bayesian network to calculate the overall probability that a specific threat is present. Examples of how the algorithms work are given using simulated and real data.

The effects of Doppler and pulse eclipsing on sidelobe reduction techniques

Radar sidelobe reduction techniques based on deconvolution generally rely on an accurate estimate of the system point spread function (PSF). Targets traveling at nonzero velocity induce a Doppler shift and have an altered PSF, which reduces sidelobe reduction performance. Also, in situations where a target is close to the radar, pulse eclipsing occurs- reflected energy arrives at the receiver while it is switched off during transmission. Eclipsing has the effect of a range-varying PSF, which also reduces sidelobe reduction performance. This paper describes a method to account for both Doppler and pulse eclipsing using the thresholded minimum mean square error (MMSE-T) sidelobe reduction algorithm. A new procedure for estimating noise power, which is required by the algorithm, is presented. Simulation results show the modified algorithm is able to reduce sidelobes such that a weak target obscured by the sidelobes of a 40 dB stronger target is clearly revealed, assuming the weak target would be detectable alone. These results hold when the true target velocities are not known as long as a reasonable estimate is obtained through tracking or Doppler processing of the strongest targets. A qualitative comparison of MMSE-T with the iterative re-weighted least squares (IRLS) algorithm shows it to be the better of the two.

A Bayesian approach to simultaneous autofocus and superresolution

This paper presents a numerical Bayesian approach to the autofocus and super-resolution of targets in radar imagery. An ill-posed inverse problem is studied in which the known linear imaging operator is subject to an unknown degree of distortion (defocusing). The goal is simultaneously to reconstruct a high-resolution representation of a target based on noisy lower resolution image measurements and to estimate the degree of defocus. We present a Markov chain Monte Carlo algorithm for parameter estimation, illustrate the approach on an explanatory example and compare our technique with a maximum likelihood approach. Given a model for the sensor measurement process, this technique may be applied to any type of radar image such as those produced by a synthetic aperture radar (SAR), inverse SAR (ISAR) or a real beam imaging radar. The proposed approach fits into a larger set of procedures aiming to exploit targeting information from different radar sensors.

Vehicle Detection in Infrared Imagery Using Neural Networks with Synthetic Training Data.

We present a new approach to the detection, localization, and recognition of vehicles in infrared imagery using a deep Convolutional Neural Network that completely avoids the need for manually-labelled training data by using synthetic imagery and a transfer learning strategy. Synthetic imagery is generated from CAD models using a rendering tool, allowing the network to be trained against a complete set of vehicle aspects and with automatically generated meta-data encoding the position of the vehicle in the image. The proposed approach is fast since a single network is used to compute class probabilities for individual pixels in the image. Results are presented illustrating the robust recognition and localization performance achievable with the novel approach for vehicle detection in real high-resolution infrared imagery.