Full text: Proceedings of the Symposium on Global and Environmental Monitoring (Pt. 1)

126 
classified pixels and the manually traced ones for landings, roads and skidroads were 756 (0.92% of the total 
areas of Unit #29), 206 (0.25%) and 3921 (4.77%) pixels, respectively. For Unit #32, only 1412, 5384 and 2385 
pixels within areas for landings, roads and skidroads, respectively, were classified as the total number of pixels 
for both "Class A" and "Class B", whereas, the manually traced areas within the landings, roads and skidroads 
were 2194, 5413 and 3891 pixels, respectively. The differences between the number of computer threshold 
classified pixels and the manually traced ones for landings, roads and skidroads were 782 (1.01%), 29 (0.04%) 
and 1506 (1.95%) pixels, respectively. These differences were due to the fact that pixels containing pockets 
of non-"Class A and B" pixels (green vegetation, tree shadows) were not classified as landings, roads or 
skidroads in the computer threshold classification, whereas manually traced pixels of the landings, roads and 
skidroads included the non-"Class A and B" pixels as areas of soil disturbance. Field checks on these pockets 
of non-"Class A and B" may indicate that these pockets are not considered as disturbed areas and may not 
need to be included in the soil disturbance guidelines. 
The results of the computer threshold classification (Table 2) indicated that Unit#29 contained 82116 pixels 
(10.225 ha) of which 3444 pixels or 4.19% (0.429 ha) were areas of "Class A" pixels (brightest pixels in the 
digitized photo image), and 13390 pixels or 16.31% (1.667 ha) were areas of "Class B" pixels (next brightest 
pixels in the digitized photo image). The results (Table 2) also indicated that Unit #32 contained 77393 pixels 
(6.833 ha) of which 5941 or 7.68% (0.524 ha) were areas of "Class A" pixels and 13566 pixels or 17.53% (1.198 
ha) were areas of "Class B" pixels. The estimated areas of "Class A" and "Class B" could only be surveyed with 
great effort in the field. It is impossible to delineate those "Class A" and "Class B" areas within Units #29 and 
#32 using conventional photo interpretation techniques at 1:5000 scale, even with enlarged photos. 
Furthermore, areas of "Class A" and "Class B" outside the landings, roads and skidroads, can be estimated by 
subtracting the computer threshold "Class A & B" within the manually traced pixels for landings, roads and 
skidroads, from those of the total threshold classified pixels. There were 452 pixels or 0.056 ha (0.55%) and 
7849 pixels or 0.977 ha (9.56%) for "Class A" and "Class B", respectively, outside of the landings, roads and 
skidroads in Unit #29. There were 388 pixels or 0.034 ha (0.50%) and 9938 pixels or 0.877 ha (12.84%) for 
"Class A" and "Class B", respectively, outside of the landings, roads and skidroads in Unit #32. These "Class 
A & B" areas outside of the landings, road and skidroads should be checked in the field for severity of 
disturbance and if these were significant soil disturbed areas, then these should be included in the soil 
disturbance guidelines. 
It appears that this threshold classification procedure can consistently obtain "Class A" and "Class B" pixels 
from different digitized photo image data sets of clearcut units. The reason for this consistency is that "Class 
A" pixels are always in grey level 250 to 255 (the brightest pixels in the digitized photo image) and "Class B" 
pixels are always in grey level 200 to 249 (the next brightest pixels in the digitized photo image) after contrast 
stretching, using minimum and maximum grey levels, has been performed on the image data sets. However, 
maximum-likelihood classification techniques (Lee 1990) cannot consistently obtain "Class A" and "Class B" 
pixels from different digitized photo image data sets of clearcuts, because one cannot obtain the same training 
samples from different image data sets and different training samples will produce different classification results 
even from the same digitized photo image. 
4.0 CONCLUSION 
The use of computer digitized photo image data sets from large-scale (1:5000 or larger) aerial photographs in 
combination with the image analytical technique allows the estimation of areas of landings, roads, skidroads, 
"Class A" - "deeply gouged or exposed compacted mineral soil" and "Class B" - "exposed mineral soil or woody 
debris". An advantage of the computer digitized photo image data was that one could use the contrast 
stretching routine to produce three enhanced image channels from the three original red, green and blue 
channels with emphasis on the study site. One could then assign a different color for each of the enhanced 
red, green and blue filtered channels to obtain the best combination of color and brightness for the image. 
The best color image can then be used for performing the necessary digital image analysis. This could not be 
done on aerial photographs. Pixel sizes of landings, roads, skidroads, can easily be calculated from the 
manually traced pixels using the image analysis system within a matter of hours. "Class A" and "Class B" pixels
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.