Evaluation of a moving-window algorithm for detecting individual tree stems from scanned aerial photographs in Victoria's eucalypt forests

Summary

An algorithm-based method developed in Canadian boreal forest conditions for automatically detecting individual stems from digitally scanned 1:17000 aerial photographs was tested on an even-aged eucalypt forest in Victoria, Australia. It was found that the algorithm functioned reasonably well for areas that were visually uniform on aerial photographs, for example, the 95% confidence intervals for the number of stems per ha at a given location were reduced by 18% through the use of the algorithm. The information extracted by the algorithm from the aerial photographs, in combination with forest height values obtained from forest maps, led to a reduction in the 95% confidence interval for total volume of 50%. However, it was the height variable that was the most significant in the relationship and not the algorithm variable. Moreover, the forest area that would be considered visually uniform on aerial photographs is only about 25%.