Naval target classification by fusion of IR and EO sensors

This paper describes the classification function of naval targets performed by an infrared camera (IR) and an electrooptical camera (EO) that operate in a more complex multisensor system for the surveillance of a coastal region. The following naval targets are considered: high speed dinghy, motor boat, fishing boat, oil tanker. Target classification is automatically performed by exploiting the knowledge of the sensor confusion matrix (CM). The CM is analytically computed as a function of the sensor noise features, the sensor resolution, and the dimension of the involved image database. For both the sensors, a database of images is generated exploiting a three-dimensional (3D) Computer Aided Design (CAD) of the target, for the four types of ship mentioned above. For the EO camera, the image generation is simply obtained by the projection of the 3D CAD on the camera focal plane. For the IR images simulation, firstly the surface temperatures are computed using an Open-source Software for Modelling and Simulation of Infrared Signatures (OSMOSIS) that efficiently integrates the dependence of the emissivity upon the surface temperature, the wavelength, and the elevation angle. The software is applicable to realistic ship geometries. Secondly, these temperatures and the environment features are used to predict realistic IR images. The local decisions on the class are made using the elements of the confusion matrix of each sensor and they are fused according to a maximum likelihood (ML) rule. The global performance of the classification process is measured in terms of the global confusion matrix of the integrated system. This analytical approach can effectively reduce the computational load of a Monte Carlo simulation, when the sensors described here are introduced in a more complex multisensor system for the maritime surveillance.