PATT: A performance analysis and training tool for the assessment and adaptive planning of Mine Counter Measure (MCM) operations

Abstract

Militaries are becoming increasingly aware of the need to quantitatively assess their Mine Counter Measures (MCM) capabilities. Recent developments in mine-hunting technology such as the use of AUV's, automated Computer Aided Detection / Computer Aided Classification (CAD/CAC) models and high resolution sonars must be evaluated to assess their abilities to meet the ever increasing demands of the MCM community. Efficient and cost-effective techniques for training MCM operators are also required. The Performance Analysis and Training Tool (PATT) module for SeeTrack Military assesses the MCM capabilities of a complete MCM system. The capability of an operator or a CAD/CAC algorithm to effectively clear a survey region is quantitatively measured (e.g. probability of detection) by PATT using an Augmented Reality approach. The key to this approach relies on accurately inserting simulated, ground-truth targets into real sensor data. Automated mission planning, risk analysis and Q-route planning are capabilities which derive the quantitative analysis output from the core PATT module. MCM capabilities are generally evaluated through sea trials which are both expensive and only use a small number of targets. A statistically robust measure of capability is therefore difficult to obtain. The PATT module deals with this by inserting multiple simulated mine targets into real sidescan sonar data allowing accurate quantitative estimates to be obtained. The topology of the seafloor is estimated through image segmentation algorithms and critical sonar parameters such as the range and resolution are determined during the process to ensure that the simulated targets are accurately integrated into the imagery. Evaluation results can therefore be obtained against a variety of controlled ground truth parameters such as sonar range, mine type and mine orientation. MCM capabilities are heavily impacted by the environment. The PATT modules uses SeeByte's Seafloor Classification module to provide information on the seabed characteristics within the survey region. A wavelet-based classification system initially classifies each individual sonar image after which a Markov Random Field (MRF) based fusion system merges these results to provide a large scale classification mosaic of the region. The impact of the seafloor on MCM capabilities can therefore also be measured. The output of the core PATT module is a series of ROC Curves providing a measure of Probability of Detection (PD) versus the Probability of False Alarm (PFA) with respect to the confidence level of the detector for any seafloor type, target type or range of analysis. The statistics can be used to provide a measure of risk of undiscovered mine targets being present in the survey region given the number of targets found using a binomial model. This capability is critical for strategic mission planning. This can later be used for planning Q-routes through the application of Fast Marching methods. PATT may also be used for AUV re-planning to maximize the use of MCM capabilities. Given the MCM capability evaluation provided by PATT, the system can use this information to provide an optimized mission plan. This mission will consider the survey region being inspected along with the impact this will have on the MCM system to maximize the probability of discovery. Results provided will demonstrate that more sophisticated mission plans are required for complex environments while the typical lawnmower trajectory usually employed in MCM operations is sufficient for benign regions. This paper will present the core technologies used within the PATT module. First an overview of the augmented reality module will be given. The paper will show how the CAD/CAC performance over different types of seafloor can be used for risk analysis and fast marching based route planning. Results from each of the different components of the tool will provided. Real and simulated data is presented.