The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
(b) 8,857 Craters for the image strips in Iani Vallis
Figure 6. Crater Detection results and assessment.
Green = True Positives, Red = False Positives (not in count),
Blue = False Negatives, Yellow = Interpolated from duplicates
Region
Mean %
Detection
Mean %
Quality
Mean
Branching
Factor
Elysium
85.59
80.09
0.09
Iani
83.31
77.74
0.09
Overall
84.45
78.92
0.09
Table 1.2D crater assessment results (diameter >8 pixels).
3.2 Assessment of 3D crater profiles
The assessment of 3D craters was performed in three ways: 1)
an inter-comparison between optical images and hill-shaded
images based on reconstructed DTMs, or constructed DTMs
and standard 3D models; 2) manual check; 3) comparison with
other 3D information sources such as HiRlSE, CTX and MOLA
track profiles.
At first the inter-comparison between optical images and DTM
was designed for a kind of automated verification stage of the
3D processing. The hill shaded image was constructed using the
illumination conditions from the corresponding HRSC image
and a correlation value with median filtered optical images was
derived. A thresholding value to discriminate ill-constructed
DTM is roughly > 0.35 but depends on crater size. We used
0.35-0.5 depending on crater size. In addition, the convolved
value between the constructed DTM with pre-defined standard
3D models was checked. If the maximum convolved value with
a standard DTM is lower than 0.7, it is classified as a poor case.
We processed 40 craters (r>400m) over parts of F1RSC images
hi 190, h2099 and hi923. Figure 7 and Table 2 show some
examples of 3D crater DTM and the assessment results.
V»
1
a
ECH.ÖUJ»
BimKm
8—»1
O 1
o
0
©
Àh
-3543-3423-3301
- . 5 - 3276
h2099
h2099
h2099
h2099
(r=0.6km)
(r=0.86km)
(r=0.57km)
(r=0.61km)
J
*
' L\
*)
U 1
i-f H ¡Km
0 I
f Li i ìsì
epwn
o
Q
imi
1 % % / j* 1 1
-3543-3423-3301
-P.41B -83?»» -*H»-
Jtns * -t- -
-sSI
h2099
h2099
h2099
h0923
(r=0.36km)
(r=0.61km)
(r= 1.08km)
(r=0.35km)
jfjj
#1
HmS
n
01
' i
0 1
0 I
FTHlKm
0 1
mm
1JÜ
CS**
ÜH
4
di
mmmm
-3721
mmmrnjm iu
-2415.4
—
h0923
h0923
h0923
h0923
(r=0.59km)
(r=0.32km)
(r=0.82km)
(i=l.53km)
Figure 7. Hill shaded images of individual crater DTMs
Automated
Verification 3
Manual
verification 11
Manual
verification c
Overall
detection
76.6%
96.6%
DP :78.5%
BF : 0%
OP :78.5%
86.6%
DP :88%
BF : 0%
OP :88%
r> 1000m
100%
100%
100%
1000m<r<450m
73.3%
100%
86.7%
r<450m
70%
80%
90%
(a) HRSC orbit 2099 (29 crater DTMs)
Automated
Verification 3
Manual
verification 11
Manual
verification 11
Overall
detection
40.9%
81.8%
DP :41.1 %
BF : 0.14
QP :38.8%
54.5%
DP : 45.4%
BF : 0.6
QP : 35.7
r> 1000m
66.6%
66.6%
50%
1000m<K450m
27.2%
90.9%
63.6%
r<450m
40%
80%
40%
(b) HRSC orbit 0923 (21 crater DTMs)
a using DTM and hill shading
b manual check, if depth/diameter is available, true
c manual check if crater rim is clear, true
