Detectability of Objects at the Sea Surface in Visible Light and Thermal Camera Images

In a number of ocean surveillance and remote sensing applications, visible light and thermal cameras are used to detect and identify objects at the sea surface. Knowing beforehand what the camera can detect or not can be important, yet highly difficult to determine. Optical models such as Modulation Transfer Functions can help in evaluating a camera system, but requires a deeper knowledge in optics, and detailed specifications of each component. The models also does not handle noise coming from the scene background, which in many cases is the major limiting factor of detectability. In this paper, we evaluate the results of an edge detection algorithm on images from two commercial off-the-shelf camera system - one visual light and one thermal. We then draw conclusions on the detectability of objects which commonly needs to be detected at the sea surface.