International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
Distance: 5 ns Distance: 6 ns 
= 
DN ® OO 
Soooo 
2-05 
oO ooo 
   
oe 
  
percentage of successful detection (%) 
en 
eo 
percentage of successful detection (76) 
2 3 2 3 4 
Ratio (A2 / A1) Ratio (A2 / A1) 
. 2 echo ; 
Echo with: 5 ns À Se Echo with: 5 ns 
Distance: 5 ns |Bilimore than 2 echo Distance: 6 ns 
(a) Zero-crossing (b) 
  
Distance: 5 ns Distance: 6 ns 
   
1 2 3 4 5 
1 2 3 4 
Ratio (A2 / A1) Ralio (A2 / A1) 
à M2 echo i 
Echo with: 5 ns |Ezi1 echo Echo with: 5 ns 
Distance: Sns Mimorethan2echo) | Distance: 6 ns 
(c) Wavelet-based (d) 
percentage of successful detection (%) 
percentage of successful detection (%) 
  
  
Figure 8. Results of overlapped echo detection under variant 
relative ratio cases. CR2: blue, MR2: green, RR2: red, (a)(b) (b) 
verasetessitig (Nd) wavelet-tased Figure 9. (a) the study area, (b) the point clouds colored by 
35 Real waveiprm data test elevation in the rectangle area of (a). 
  
  
Figure 9 shows the study area where tall and shrubs plants grow 
irregularly. The real waveform data is captured by Leica ALS60 Waveforms = Wavelet-based 
system. We chose the waveform data in the rectangle area detector 
shown in Figure 9 (a). The objective of this test is to see if the 
wavelet-based detector can effectively detect the weak and 
  
  
  
  
  
  
  
  
  
  
  
  
  
overlapped echoes so that the number of points increases 
obviously. Export point Remove echoes 
xport points where 
3.5.1 Set up: The results of simulated experiments can be SNR, 4, « 16 dB 
used as reference to set up the strategy of detecting echoes in : s 
real waveform data. From section 3.3.2 the wavelet-based Figure 10. Flow chart ofthe wavelet-based detection method 
detector can confidently detect the echoes if the SNR value of 
an echo is greater than 16. Therefore a threshold is set as a 
condition to remove the uncertain echoes which may be 
incorrectly detected due to the noises. To obtain the SNR value 
of each echo, the noises in the waveform needs to be estimated. 
Fortunately that there only exists noises in some interval of each 
waveform, the noises can be therefore easily estimated using 
these waveform sections without target echo information. 
Figure 10 shows the procedure of the wavelet-based echo 
detection method applied to a real waveform dataset. 
3.5.2 Results: 
The total number of waveforms is 459708 and the number of 
points provided by the instrument is 499997. Based on the 
wavelet-based detector, the number of detected points is 590074. 
It means that the detector increasingly found 1896 more of the 
original number of echoes. This result shows that the ability of 
the proposed detector in finding weak and overlapped returns 
from waveforms. Figure 11 shows a side view of a selected 
profile drawn by the system points and the points extracted by (b) 
our method respectively. One can see that the ground and the 
trees can be better described by the point clouds extracted by 
our methods. 
  
Figure 11. The side view of a selected profile. (a) points 
provided from the instrument, (b) points extracted from 
waveforms using the wavelet-based detector. The arrows show 
some of the extra-points detected by our method. 
833