Abstract
Hemispherical photography (HP) is a long-standing tool for forest canopy characterization. Currently, there are low-cost fisheye lenses to convert smartphones into high-portable HP equipment; however, they cannot be used whenever since HP is sensitive to illumination conditions. To obtain sound results outside diffuse light conditions, a deep-learning-based system needs to be developed. A ready-to-use alternative is the multiscale color-based binarization algorithm, but it can provide moderate-quality results only for open forests. To overcome this limitation, I propose coupling it with the model-based local thresholding algorithm. I call this coupling the MBCB approach.
Methods presented here are part of the R package CAnopy IMage ANalysis (caiman), which I am developing. The accuracy assessment of the new MBCB approach was done with data from a pine plantation and a broadleaf native forest.
The coefficient of determination (R2) was greater than 0.7, and the root mean square error (RMSE) lower than 20 %, both for plant area index calculation.
Results suggest that the new MBCB approach allows the calculation of unbiased canopy metrics from smartphone-based HP acquired in sunlight conditions, even for closed canopies. This facilitates large-scale and opportunistic sampling with hemispherical photography.
Competing Interest Statement
The authors have declared no competing interest.
