Scanopy: canopy analysis with fish-eye lens and digital camera

An accurate classification of pixels into the sky (gaps) and canopy categories is a pre-requisite to get precise canopy analyses from hemispherical images. WinSCANOPY offers different methods to do this classification and to modify it after if required.

All WinSCANOPY versions have a global automatic threshold method which has been improved in 2006. This method uses grey levels information (light intensity from a color or grey levels image) to decide in which class (sky or canopy) pixels belongs to. With a global threshold, the classification criteria is the same for all pixels of the hemisphere.

The Pro DSLR version offers four additional methods to classify pixels, two of which are specific to hemispherical images.

An adaptive threshold that changes value in function of the location in the hemisphere. It adapts itself in function of the lighting variations.
An hemispherical threshold that takes into account the light variation of hemispherical lenses which are brighter at the zenith and darker at the horizon.
A threshold that takes into account the light variations due to the sun position in the image (indicated by the operator). The attenuation compensation can be linear or like the SOC diffuse radiation distribution.
Classification based on true color. This algorithm is more tolerant to sky conditions variations. For example, it allows to analyse images with white clouds and dark blue sky, a task difficult to do with simple grey levels thresholds (the dark blue sky tend to be classified as canopy).

The result of the pixels classification can be viewed before the analysis or after. As you change the parameters, the resulting classification is shown in the displayed image allowing you to choose the best method.

The pixels classification can be verified and modified for specific image regions at any moment of the analysis. Pixels that fall into the canopy group are drawn a different color over the original image as the threshold (pixel classification criteria) is modified by moving a slider bar. This allows for a simultaneous view of the original and pixels classification images.

Select a region to be classified. It can be the whole image, a sub-region of any shape defined by outlining it or a pre-defined circular regions corresponding to the sky grid's zenith rings or circular regions centered on the sky brightest position (below).

As you move the slider, pixels classified into the canopy groups are drawn in green over the original image.

Adjust the slider so that all canopy elements are covered by green pixels (but not the sky). The analysis is updated automatically.

The image to the left has a non-uniform sky light distribution. It is well analysed with our solar threshold which adjusts its strength in function of position in the hemisphere. In this case, a global threshold or a threshold adjusted in function of the sky grid (centered on the zenith) is not efficient.

Color analysis is more tolerant to sky condition variations. Images with clouds and blue sky or blue sky alone can often be analyzed.
	The left side image is more easily analysed in color than in grey levels due to the presence of dark blue sky which tends to be classified as canopy in a grey levels analysis. White stems (right image) are often misclassified as sky with a threshold based method. With color classification this is not a problem (when no white clouds are present)

10 to 16 bits of grey levels per pixels images (like those produced by DSLR cameras) offers new pixels classification opportunities. An image with more bits per pixel has a wider dynamic range (difference between darkest and brightest light levels) and more tonal variations (smallest light variations that can be distinguished). When the number of grey levels is not sufficient, light variations are merged into the same grey level information is lost. When subtle light variations are preserved, this gives more choices for pixels classification into canopy and sky


WinSCANOPY 2006a Measurements & Features
Gap Size Distribution Analysis (Pro DSLR) - (new since 2006a) Gap size distribution (GSD), i.e. the number of gaps in function of their size, can be used in combination with gap fractions to quantify the degree of clumpiness at the tree level and to use this information to increase the accuracy of LAI measurements. For a canopy of a given gap fraction with randomly distributed foliage elements, it is possible to make a theoretical probability of gaps occurring in function of their size. By comparing the measured GSD to this theoretical distribution, foliage clumpiness can be measured. At the base of GSD analysis, is the classification of gaps in two categories; those which are normally expected for a given randomly distributed leaf area and those which are not. The latter are larger gaps that are present because of foliage clumping at the crown level and can be seen between tree crowns. These are called between-crown gaps while random gaps are called within-crown gaps. WinSCANOPY has two methods of classifying gaps into these two groups; Chen and Cihlar 95's method based on transect length (a one dimensional data), which is also used in a sunfleck based commercial instrument, and a new simpler, but efficient, method of our own based on gap area (a two dimensional data).

GSD analyses can be done on hemispherical or cover images. Between-crown gaps are drawn in blue, within-crown gaps are drawn in yellow.
Other features of GSD > On-screen visualisation of between-crown and within-crown gaps. Can also be saved to standard tiff image files. > The automatic gap classification can be modified with simple mouse clicks. It can also be done completely manually. > Clumping index is measured in function of view zenith angle and globally for the hemisphere or for any view angles range that you choose. Clumping index in function of zenith can be displayed in the graphic above the image during the analysis.

Canopy Cover Images Analysis (Pro DSLR) - (new since 2006a) Canopy cover images have a narrow view angle (5 to 25 degrees) directed toward the zenith or close to it (see figure below). This kind of analysis is an alternative method to hemispherical images analysis to compute LAI and other canopy structural parameters (crown porosity, crown cover, foliage cover, clumping index). Full-Frame Fish-Eye Images Analysis (Pro DSLR) - (new since 2006a) Full-frame fish-eye images are acquired with a fish-eye lens but do not have a circular projection. The 180 degrees (or less) typical field-of-view spans over the diagonal of the image sensor rather than the vertical image dimension. For a given image sensor size, a full frame fish-eye image is obtained by using a fish-eye lens with a longer focal length than a regular fish-eye circular projection lens or simply by zooming in (Point & Shoot camera). One of their advantage is to increase the effective image resolution as all pixels are used for canopy and sky information (no black pixels). The analysis that can be done on these images are identical to those of regular fish-eye images.

Standard fish-eye image	Full-frame fish-eye image	Cover image
Two methods to measure LAI or leaf density of isolated tree - (new since 2006a) One method (Pro DSLR) was first described by Lindsey and Bassuk 1992 and later modified and tested by Peper and McPherson 1998. The other (Reg) is a modification to the LAI2000 LAI method which consists in substituting the default normalized path lengths for those of the tree canopy (length traveled by light in the canopy at the five rings view angle). Individual leaf area measurement from non fish-eye images (Pro) Turns WinSCANOPY into a basic individual leaf area meter, disease quantifier (see WinFOLIA for more sophisticated measurements) and soil foliage cover quantifier. Leaf projection coefficient in function of view zenith angle (Reg)