One of the primary means to evaluate the accuracy of a shape model is to measure the deviation between a truth model (if available) and the shape model. Typically, this is done by calculating the square root of the average error squared of all the points, i.e the root mean squared error (RMS).This technique provides valuable insight into the error distribution of a shape model, as well as providing an objective measurement of deviations. However, it does not fully explain the error and especially the quality of a digital terrain model. Systematic errors can obscure poorly performing regions and may over-report errors.We have begun an extensive analysis of using normalized cross-correlation to evaluate the quality of shape models compared to truth topography, as well as the agreement between images rendered from the model with the original images. This technique provides a tool to differentiate between local accuracy and global accuracy. It also provides an effective way to decompose the error vector into horizontal and vertical displacements. It is especially useful for stereophotoclinometry (SPC) because it allows a clear determination of the quality of the model at the resolution of the source images (i.e. if the source images have a 5cm pixel size, it shows how well the SPC solution is at 5cm). Additionally, it demonstrates how essential a good imaging plan is to the quality of the shape model.We are using these techniques in support of the OSIRIS-REx mission to the asteroid Bennu.
AAS/Division for Planetary Sciences Meeting Abstracts #48
- Pub Date:
- October 2016