1074 
2.5- 
Figure 8 : Leaf area estimation : 
Figure 7 : Leaf area azimuth distribution (dashed a) leaf area density vertical profile : 
line ) and the midrib azimuth distribution (solid - by plant profiles method (dotted line ) with 
line ). the confidence intervals (dashed line ); 
- by stereovision (solid line •). 
b) leaf area index horizontal profile in the cross row 
direction; 
The zenith angle distribution was estimated per classes of 10 degrees. The same study was performed 
by using the plant profiles method and both data plots showed an almost planofile leaf distribution with a mean 
zenith angle of 11° (figure 6) and a standard deviation a(Q) = 1.86°. The equation of the regression 
constrained to the origin was : 
0 STEREO -- 0 979 0 PLANTPROFILES /j-2 = Q 414) 
The foliage azimuth distribution was calculated per classes of 10 degrees from 0° to 360° (figure 7) 
and the result showed a nearly uniform distribution with slight maxima in the parallel to rows direction. The 
same tendency was observed on the midrib azimuth angle distribution shown on the same figure. 
Figure 8 compares the vertical profile of LAD evaluated from the 3-D maize model with the 
estimations from the plant profiles method. Dividing the vegetation into 8 horizontal layers of 0.3 m high each, 
the total leaf surface inside these layers was calculated. Standard deviation between LAD estimated by 
stereovision and that the plant profiles method was a = 0.393 m^/mL Equation of the regression constrained to 
the origin was : 
LAD STER£0 (Z) = 0.922 LAD PLANTPR0FILES (Z) (r 2 = 0.878) 
Figure 8 shows that the LAD STERE0 (Z) was slightly underestimated in regards to the 
LAD PLANT PR0F,LES (Z) in the extreme leaf layers and overestimated in the 1.8 m and 2.1 m layers. 
4 DISCUSSION AND CONCLUSION 
4.1. Precision of the (X, Y, Z) calculation 
We must outline that the precision obtained by our method is not as good as the precision reached by other 
photogrammetric techniques. Let us consider the main error causes in our processing : 
• The small B / H ratio. 
• Image distortion due to the photographic process and of paper enlargements. 
• Insufficient resolution of the digitizing device which was operated manually. 
The two last problems could be solved by using special devices allowing to directly process the slides so as to 
avoid paper enlargements and manual digitizing. Another mean to simplify the treatment is to use CCD 
cameras (Coupled Charge Device) providing digital images with the appropriate pixel resolution which should 
allow automation of manual procedures.