Assessing Geologic Image Interpretation Errors Occurring in Extraterrestrial Robotic Exploration

Robotic exploration of the Martian surface requires numerous interpretations of imaged data, where incorrect results can have drastic consequences. The imaging process transforms and reduces the amount of information available. Three experiments measured the differences in interpretation between imaged sediments and physical sediments. Three characteristics were analyzed: grain length, grain shape, and grain distribution. The results found the difference between the grain length measured on an image and the true length to +/- 2.333 pixels (p <0.0001); the difference is similar to the amount of blurring introduced by the camera. Both grain roundness and grain sphericity were classified on a scale from 1 to 6 in the shape experiment. The roundness classification differed by 0.114 categories (p = 0.0082) with the imaged grains being rounder. The sphericity classification differed by 0.151 categories (p = 0.0010) with the imaged grains being less spherical. In the distribution experiment, the subjects determined the percentage of the total image area covered by grains in six specified size ranges. The average error for each size range was 11.112 % of the total area (p < 0.0001). In all three experiments, the measurements taken using the imaged specimens significantly differed from the measurements taken using the physical specimens. The magnitudes of the differences were small and may not be scientifically significant.