Test11J:
Building topography with s/c azimuth every 90deg and 5cm images.
Results Summary:
Maplet continues to change even after 150 iterations. I am not sure why CompareOBJ gives such low numbers for the first few steps; as you can see the boulder is basically flat. Even though the truth does not have a tilt, the solution is tilted quite a bit. Note that once I started using auto alignment the top of the boulder dropped down a bit.
RMS values:
STEP |
COMPAREOBJ(cm)(w/ opt trans and rot) |
compMapVec(cm) |
NOTES |
step01 |
2.19 |
31.4 |
aligned images by hand |
step02 |
3.53 |
27.9 |
|
step03 |
4.49 |
25.3 |
|
step05 |
5.73 |
21.6 |
|
step10 |
5.49 |
17.0 |
|
step15 |
5.11 |
14.6 |
|
step20 |
4.76 |
12.9 |
|
step30 |
4.27 |
10.6 |
|
step31 |
4.23 |
10.4 |
autoalign from here down |
step32 |
4.18 |
9.33 |
|
step40 |
3.88 |
7.29 |
|
step50 |
3.50 |
6.76 |
|
step100 |
2.52 |
4.28 |
Traces:
Landmark Display:
step01 
step150 
step250 
Run the model even further:
Run the model out many iterations further and track the RMS for CompareOBJ with and without translate.
step250 
STEP |
COMPAREOBJ(cm)(w/ opt trans and rot) |
COMPAREOBJ(cm)(w/o opt trans and rot) |
NOTES |
step250 |
2.19 |
12.0 |
made truth at same location as maplet |