Theory and Experimental Assessment of Real-Time Sea-State Estimation Via Sunglint Inversion G. P. Cureton 1 , S. J. Anderson 2 , M. J. Lynch 1 , B. T. McGann 1 1 Remote Sensing and Satellite Research Group Curtin University of Technology GPO Box U1987, Perth WA 6845 2 Electronics and Surveillance Research Laboratory Defence Science and Technology Organisation PO Box 1500, Salisbury SA 5108 Abstract— A nonlinear retrieval model was used to invert wave image data containing sunglint to obtain the elevation power spec- trum. The sunglint images were thresholded to obtain binary glint images, from which the glint autocorrelation was calculated in one direction. A relationship was determined connecting the glint and slope random variables, which was then inverted to obtain the slope autocorrelation, from which the elevation power spectrum was obtained by integration and Fourier transform. I. I NTRODUCTION The measurement of ocean directional wave spectra has historically been pursued by means of in situ sensors which estimate the surface elevation and/or surface slopes from the dynamics of a floating buoy or from some other electro- mechanical device immersed in the water body. Such tech- niques can provide very direct measurements of the physical quantities of interest, but they are costly to build, deploy and operate, sample only a single location, and yield reasonable accuracy over only a limited band of wave frequencies. In recent years, an emerging requirement for near- simultaneous, high spatial resolution observations over entire regions has focussed attention on remote sensing technologies which can conduct measurements over vast swathes of ocean from airborne or space-borne platforms. Both active and pas- sive sensor systems have been developed, amongst the best known of which are methods which retrieve wavenumber spec- tra by imaging the surface in an extended spatial sense using photographic or video techniques. An essential component of any image-based remote sensing technique is a model connecting the measured observable (im- age intensity) with the surface parameters of interest (the ele- vation and the slope). Early attempts to exploit natural illumin- ation imagery (using smoothly varying or constant sky radi- ance) revealed that this relationship is complicated by depen- dencies on the scattering geometry, detector or film response, and the particular sky radiance distribution, as well as the im- pact of wave breaking, shadowing, turbidity and the presence of surfactants [1–6]. Glint imagery has the advantage of a low signal to noise ratio where glints are present, and in such lo- cations the surface slope could be precisely determined. How- ever, such imagery could not be used with models derived for use with sky radiance data, as it did not furnish a continuous relationship between the surface slope and image intensity [5]. These considerations led to a less than favourable assessment of the potential of glint image interpretation as a directional wave spectrum measurement technique. In 1993, Alvarez-Borrego [7] proposed a model relating the slope and glint autocorrelations, which is less susceptible to the discontinuous nature of the input data. Working from a linear representation of the sea surface dynamics, Alvarez-Borrego developed an inversion procedure and demonstrated that it was possible to recover the spatial wave spectra of parametrically- defined ocean surfaces. The work reported in this paper addresses three issues. First, we describe a Monte Carlo simulation environment, based on the Alvarez-Borrego model, which can be used as a test-bed for model validation. Second, we present the results of an ex- periment conducted using an airborne imaging system which recorded sun glint data from a specific patch of sea, at twelve different scattering geometries, to test the robustness of the in- version procedure. Finally, we consider the ramifications of ex- tending the surface representation to allow for nonlinear wave interactions induced by the free surface boundary conditions. II. RETRIEVAL MODEL We have the surface elevation η(x, y) with variance σ 2 η . The slope of this surface is given by M x (x, y) and M y (x, y), with variance σ 2 M . The surface is illuminated by a source S with zenith angle θ s and angular subtense β. The beam is reflected at the specular point, at which the surface normal ˆ n makes an angle θ with the zenith, and reaches the detector D with zenith angle θ d (Fig. 1). When we look at an image of the ocean surface, the pres- ence of a glint at a particular point indicates that the surface normal ˆ n is orientated in such a way that the specular condition is satisfied between S and D. Because we are dealing with an extended source (the Sun), there will be a range of slopes that This work was supported by the Defence Science and Technology Organisation.