Digitally scanned aerial photographs (1:15840) of 40 forest areas were processed by two texture algorithms, a high pass filter and a diversity filter. The texture indices generated by these algorithms and the textural statistics on the raw data were regressed with forest traits of age, height, crown closure, crown width and trees/acre. Five pixel sizes varying from 100 to 500 pixels per inch were evaluated for their effect on the regression analysis. Correlations were much stronger when variations related to species composition were reduced.
Mike L. Hoppus,
Don T. Evans,
"Role of texture in interpreting images of forest land", Proc. SPIE 2023, Airborne Reconnaissance XVII, (20 December 1993); doi: 10.1117/12.165539; https://doi.org/10.1117/12.165539