Test F3B - Number of Images
Goal: SPC derives much of its ability to remove error by using a large number of images, and as such, error becomes Gaussian noise. The purpose of the test is to measure the rate and magnitude of improvement with number of images.
Test Set Up
Data:
- Approach, extra set of RC (Rotation Characterization) and closer set of RC
- Centered around the middle of TAG Site #1, -8.030 97.218
- Take the following images and add perturbed sets.
- Based upon F1b/DS results (Bennu2) with MAPLET EE0452 at 5cm resolution, -8.023 97.220.
- Test using 4, 7, 14, 70, and 140 images
- Uses the following images (Figure 1). Add displacements for this of x2, x10, x20, x40
- P601293196, P601372856, P601372862, P601372869, P601372875
Start: Preliminary Survey Model, TAG Site only, based on Shape 3.7.
Build:
- TAG site, near 0 latitude (for best evaluation).
- Create a 25x25m BIGMAP with MAPLETS at 5cm.
- Evaluate the center 20x20m to the truth model.
Sub-Tests
F3B-1 |
Use only three images: P601293196, P601372856, P601372869 |
F3B-2 |
Use all five images: P601293196, P601372856, P601372862, P601372869, P601372875 |
F3B-3 |
Use ten images, start with the base fives images and perturb them twice. |
F3B-4 |
Use fifty images, start with the base fives images and perturb them ten times |
F3B-5 |
Use one hundred images, start with the base fives images and perturb them twenty times |
Evaluate:
- Accuracy
- Relative Accuracy