Lunar and Planetary Science XXXIII (2002) 1771.pdf

THE TRUE COLOR OF YOGI: AN ACCURATE METHOD FOR REMOVING DIFFUSE ILLUMINATION FROM MULTISPECTRAL IMAGES OF Carol Stoker and Kathy Rages M.S. 245-3 NASA Ames Research Center, Moffett Field, CA 94035, [email protected]..gov, [email protected]

Background: The apparent color of surfaces on Mars also requires knowledge of the incidence and reflec- is strongly influenced by the color of the light illumi- tance angles and phase angle. These, in turn, require nating them. scenes are illuminated by both detailed knowledge of the 3-D position and orientation direct sunlight and diffuse skylight caused by the scat- of the surface. The difficulty of obtaining this infor- tering of sunlight by dust particles in the atmosphere. mation has previously prevented accurate correction of This diffuse illumination is reddened compared to the spectra for diffuse illumination [1]. We use three- solar spectrum and is a substantial fraction of the total dimensional models of the terrain derived from stereo illumination [1]. For example, Figure 1 shows the rock images using the methods described by Stoker and Yogi taken at two times of day. The color differences others [4] to compute the normal to the surface for between the center and right faces (points A and B on each considered point, and from these compute the Fig. 1) in the morning image were interpreted [2] to be scattering geometry. evidence that the East face of Yogi, which faces the mean wind, was relatively free of an oxidized coating. Results: To study the effect of varying illumination Abrasion due to wind-carried particles might be the conditions on Yogi, and to illustrate our method for mechanism for removal of the coatings. However, retrieving true color, we constructed a geometric Thomas and others [1] argued that the apparent color model which roughly approximates the shape and po- of Yogi can be fully explained by variations in the sition of Yogi rock by a truncated dodecahedron, relative amounts of direct and diffuse light on the dif- placed at an azimuth of 330o from an observer. The ferent faces of the rock. Thus, it is important to ac- reflectance properties were represented as Lambertian count for and remove the effects of illumination con- with Albedo=1. The light scattered into the direction ditions to correctly interpret reflectance spectra ob- of the observer for each visible surface was calculated tained from a Mars lander or rover. for 9:00, 12:00 and 15:00 local solar time on July 20, 1997 at the Pathfinder latitude on Mars. Figure 2 shows the model as it would appear on a dust free Mars, at the three times of day. Table 1 shows the ra- tio of radiances in red (670 nm) and blue (440 nm) channels. The diffuse (sky) illumination is a substan- tial contribution to the total illumination of all faces at all times. Also, the diffuse illumination is substantially redder than the solar spectrum. Even though the direct (sun) illumination is somewhat blue, as a result of pas- sage through the atmosphere and preferential extinc- Figure 1 . Images of Yogi taken at 9:00 (left) and tion of red light by dust, the total illumination on all 15:00 (right) showing the change in appearance at dif- faces is always redder than the solar spectrum due to ferent illumination conditions. the diffuse illumination.

Methods: We have developed an accurate method for removing the effects of diffuse illumination which uses a doubling-adding radiative transfer computer code to model the direct (solar) and diffuse (sky) illumination of surfaces as separate contributions to the total re- flected light intensity. The method can be used to pro- duce spectra as they would appear if there were no dust in the atmosphere, provided we know the scattering Figure 2. Geometric model of Yogi as it would ap- geometry and the light scattering and absorption prop- pear at three times of day illuminated by the sun with erties of the atmospheric dust and the surfaces. We are no dust in the atmosphere. The facets 1-4 in Table 1 applying our method to interpretation of multispectral are labeled on the 9:00 AM model. images obtained from the IMP cam- era. Diffuse illumination is produced by the scattering We used our radiative transfer code to investigate of sunlight by atmospheric dust which we model using whether the color differences between points A and B the values for the particle size, complex refractive in- on Yogi are true color or illumination effects. Using 3- dex, and optical depth reported by [3]. The modeling D models deduced from a stereo pair of Yogi acquired Lunar and Planetary Science XXXIII (2002) 1771.pdf

THE TRUE COLOR OF YOGI C. Stoker and K. Rages

in the Superpan imaging sequence, we determined the an efficient way to produce spectral products which are surface normal and scattering geometry for the points independent of illumination conditions. This method A and B. We analyzed a multispectral sequence of should be applied to interpretation of other multispec- Yogi acquired at four times of day. We determined an tral image products from Mars Pathfinder and to the equivalent Albedo for Yogi at these two upcoming Mars Exploration Rover as it eliminates the points by dividing the calibrated image radiance by the ambiguities in the interpretation of spectra obtained radiance of a Lambert surface with unity Albedo tan- under differing illumination conditions. gent to the rock surface at these points and illuminated by our Mars lighting model. Results are illustrated in Plans: The method we used to interpret the color of Table 2. We conclude that point B is less red (and by Yogi still has the simplification of not accounting for implication less coated with bright red dust) than point the wavelength-dependant scattering phase function of A. the surface. We plan to show results using a Hapke model for the surface reflectance properties. With sur- Table 1. Synthetic model of Yogi face properties modeled, we can produce "true color" Time Face sky sun total spectra, that is, spectra as they would appear on Mars (I/F)red (I/F)red (I/F)red without atmospheric dust. (I/F)blue (I/F)blue (I/F)blue 9:00 1 1.35 0.94 1.08 References: [1] Thomas, N. et al., J. G. R. 104, 8795-8808, 1999. 2 2.27 ND 2.27 [2] McSween, H. et al., J. G. R. 104, 8679-8716, 1999. 3 1.36 0.94 1.09 [3]Tomasko, M. et al., J. G. R. 104, 8987-9008,1999. 4 1.31 0.94 1.07 [4] Stoker, C. et al., J. G. R. 104, 8889-8906, 1999. 12:00 1 1.27 0.95 1.05 Acknowledgements: We gratefully acknowledge 2 1.38 0.95 1.09 sponsorship of the Mars Data Analysis program for 3 1.37 0.95 1.09 support of this study. 4 1.42 0.95 1.11 3:00 1 1.35 0.94 1.08 2 1.34 0.94 1.08 3 2.00 0.94 1.69 4 2.39 ND 2.39

Table 2. Equivalent Lambert Albedo and red/blue Al- bedo ratio for Yogi at points A and B Time Wave Point A Point A Point B Point B

Length Albedo Ared/Ablue Albedo Ared/Ablue (nm) 7:41 440 0.119 4.06 0.067 2.57 670 0.483 0.172 9:04 440 0.089 4.54 0.067 2.76 670 0.404 0.185 15:00 440 0.055 4.64 0.129 2.74 670 0.255 0.353 16:43 440 0.051 4.69 0.134 2.44 670 0.239 0.327

Conclusions: We conclude that the Eastern facet of Yogi, which faces into the mean wind direction, is sub- stantially more blue (less red) than the orthogonal (Southern) facet. This corroborates the conclusion of McSween and others [2] that coatings of bright red material, probably dust, are responsible for the red color of rocks at the Pathfinder site. The method we applied to Yogi can be performed quickly and provides