, Martin Macek [aut]
| Title: | Processing Hemispherical Canopy Images |
|---|---|
| Description: | Import and classify canopy fish-eye images, estimate angular gap fraction and derive canopy attributes like leaf area index and openness. Additional information is provided in the study by Chianucci F., Macek M. (2023) <doi:10.1016/j.agrformet.2023.109470>. |
| Authors: | Francesco Chianucci [aut, cre]
, Martin Macek [aut]
|
| Maintainer: | Francesco Chianucci <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 1.1.4 |
| Built: | 2024-11-04 05:04:35 UTC |
| Source: | https://github.com/cran/hemispheR |
The function calculates a single threshold of a single-channel raster image using the autothresholdr::auto_thresh() functionality.
The single thresholding is also applied at sub-image level if zonal=TRUE.
The available methods are described at https://imagej.net/plugins/auto-threshold.
The thresholding value is then used to make a binary raster image of canopy (0) and gap (1) pixels.
binarize_fisheye( img, method = "Otsu", zonal = FALSE, manual = NULL, display = FALSE, export = FALSE )binarize_fisheye( img, method = "Otsu", zonal = FALSE, manual = NULL, display = FALSE, export = FALSE )
img |
SpatRaster. A single layer fisheye image imported by |
method |
Character. The method used to threshold the image, using the |
zonal |
Logical. If is set to TRUE, it divides the images in four (N, W, S, E) regions and classify each region separately. Useful in case of uneven illumination condition in the image. |
manual |
Numeric. It uses a manual thresholding instead of automatic one. If selected, it overrides automatic thresholding. |
display |
Logical. If is set to TRUE, it plots the classified binary image. Default to FALSE. |
export |
Logical. If is set to TRUE, it saves the binary fisheye image as tif file. Default to FALSE. |
A binary single-layer image (SpatRaster)
https://imagej.net/plugins/auto-threshold
c.im<-system.file('extdata/circular_coolpix4500+FC-E8_chestnut.jpg',package='hemispheR') c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye(display=TRUE) #zonal thresholding: c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye(zonal=TRUE,display=TRUE) #manual thresholding: c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye(manual=55,display=TRUE)c.im<-system.file('extdata/circular_coolpix4500+FC-E8_chestnut.jpg',package='hemispheR') c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye(display=TRUE) #zonal thresholding: c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye(zonal=TRUE,display=TRUE) #manual thresholding: c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye(manual=55,display=TRUE)
Provide circular mask parameters from known camera+fisheye lens models
camera_fisheye(model = NULL)camera_fisheye(model = NULL)
model |
Character. An input camera+lens model |
A list of three parameters (xc, yc, rc) of the circular mask
#available camera+lenses: list.cameras camera_fisheye(model='Coolpix4500+FC-E8')#available camera+lenses: list.cameras camera_fisheye(model='Coolpix4500+FC-E8')
The function calculate canopy attributes from angular distribution of gap fraction. It returns both the effective (Le) and actual (L) leaf area index following the Miller theorem (1967). The Lang and Xiang (1986) clumping index LX is calculated as the ratio of Le to L; two additional clumping indices (LXG1, LXG2) are derived from ordered weighted average gap fraction as in Chianucci et al. (2019). The mean leaf angle (MTA) and the ellipsoidal x are derived from Norman and Campbell (1989). Canopy openness is also provided as weighted diffuse non-interceptance (DIFN), following the LAI-2200 manual (Li-Cor Inc., Nebraska US).
canopy_fisheye(rdfw)canopy_fisheye(rdfw)
rdfw |
Dataframe. The input dataframe generated from |
A dataframe of canopy attributes from classified fisheye images.
Chianucci F., Zou J., Leng P., Zhuang Y., Ferrara C. 2019. A new method to estimate clumping index integrating gap fraction averaging with the analysis of gap size distribution. Canadian Journal of Forest Research 49 doi:10.1139/cjfr-2018-0213
LAI-2200C Plant Canopy Analyzer - Instruction Manuals. Licor.
Lang A.R.G., Xiang Y. 1986. Estimation of leaf area index from transmission of direct sunlight in discontinuous canopies. Agricultural and Forest Meteorology 37, 228-243. doi:10.1016/0168-1923(86)90033-X
Miller J.B. 1967. A formula for average foliage density. Australian Journal of Botany 15(1) 141 - 144. doi:10.1071/BT9670141 .
Norman J.M., Campbell G.S. 1986. Canopy structure. In: Plant Physiological Ecology, pp. 301-325 doi:10.1007/978-94-009-2221-1_14.
c.im<-system.file('extdata/circular_coolpix4500+FC-E8_chestnut.jpg',package='hemispheR') c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye() |> gapfrac_fisheye(lens='FC-E8',nrings=7,nseg=8,endVZA=70) |> canopy_fisheye() #Zenith rings similar to LAI-2000/2200: c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye() |> gapfrac_fisheye(lens='FC-E8',nrings=5,nseg=8,endVZA=75) |> canopy_fisheye() #The hinge angle method close to 1 radian (57 degree): c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye() |> gapfrac_fisheye(lens='FC-E8',nrings=1,nseg=8,startVZA=55,endVZA=60) |> canopy_fisheye()c.im<-system.file('extdata/circular_coolpix4500+FC-E8_chestnut.jpg',package='hemispheR') c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye() |> gapfrac_fisheye(lens='FC-E8',nrings=7,nseg=8,endVZA=70) |> canopy_fisheye() #Zenith rings similar to LAI-2000/2200: c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye() |> gapfrac_fisheye(lens='FC-E8',nrings=5,nseg=8,endVZA=75) |> canopy_fisheye() #The hinge angle method close to 1 radian (57 degree): c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye() |> gapfrac_fisheye(lens='FC-E8',nrings=1,nseg=8,startVZA=55,endVZA=60) |> canopy_fisheye()
The function calculates the gap fraction for a number of zenith annuli (rings) and azimuth sectors (segments). A list of lens is available for correcting for lens distorsion. Type 'list.lenses'.
gapfrac_fisheye( img.bw, maxVZA = 90, lens = "equidistant", startVZA = 0, endVZA = 70, nrings = 7, nseg = 8, message = FALSE, display = FALSE )gapfrac_fisheye( img.bw, maxVZA = 90, lens = "equidistant", startVZA = 0, endVZA = 70, nrings = 7, nseg = 8, message = FALSE, display = FALSE )
img.bw |
SpatLayer. A binary, single-layer fisheye image generated from |
maxVZA |
Numeric. The maximum Zenith angle (in degrees) corresponding to the image radius. Default= 90. |
lens |
Character. The lens type for fisheye-lens correction. A list of lenses is available by typing list.lenses. If missing, it is assumed equidistant. |
startVZA |
Numeric. The minimum Zenith angle (in degrees) considered in the analysis. Default is 0. |
endVZA |
Numeric. The maximum Zenith angle (in degrees) considered in the analysis. Default is 70. |
nrings |
Numeric. The number of equiangular zenith rings considered in the analysis. Default is 7. |
nseg |
Numeric. The number of azimuth segments considered in the analysis. Default is 8. |
message |
Logical. If set to TRUE, it reports the circular mask used in the analysis. |
display |
Logical. If set to TRUE, it desplays the zenith rings and azimuth segments overlaid on the fisheye image. |
A dataframe of gap fraction (GF) for zenith rings (rows) and azimuth segments (columns).
Francesco Chianucci
Lens correction functions have been retrieved from the following sources:
Pekin and Macfarlane 2009: doi:10.3390/rs1041298
Paul Bourke:http://www.paulbourke.net/dome/fisheyecorrect/
Hemisfer: https://www.schleppi.ch/patrick/hemisfer/index.php
c.im<-system.file('extdata/circular_coolpix4500+FC-E8_chestnut.jpg',package='hemispheR') #List of lenses for fisheye projection correction: list.lenses #Zenith rings similar to LAI-2000/2200: c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye() |> gapfrac_fisheye(lens='FC-E8',nrings=5,nseg=8,endVZA=75,display=TRUE) #The hinge angle method close to 1 radian (57): c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye() |> gapfrac_fisheye(lens='FC-E8',nrings=1,nseg=8,startVZA=55,endVZA=60,display=TRUE)c.im<-system.file('extdata/circular_coolpix4500+FC-E8_chestnut.jpg',package='hemispheR') #List of lenses for fisheye projection correction: list.lenses #Zenith rings similar to LAI-2000/2200: c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye() |> gapfrac_fisheye(lens='FC-E8',nrings=5,nseg=8,endVZA=75,display=TRUE) #The hinge angle method close to 1 radian (57): c.im |> import_fisheye(circ.mask=camera_fisheye('Coolpix4500+FC-E8')) |> binarize_fisheye() |> gapfrac_fisheye(lens='FC-E8',nrings=1,nseg=8,startVZA=55,endVZA=60,display=TRUE)
This function imports fisheye images using terra::rast() functionality, by selecting a single channel, or a combination of channels.
The default option (blue channel) is generally preferred for canopy image analysis as it enables high contrast between canopy and sky pixels, which ease image thresholding.
A circular mask is then applied to mask outside pixel in case of circular fisheye images. It can be manually inserted, or retrieved using the camera_fisheye() function.
Alternatively, it is automatically calculated. Additional functions include a gamma correction and a contrast stretch.
import_fisheye( filename, channel = 3, circ.mask = NULL, circular = TRUE, gamma = 2.2, stretch = FALSE, display = FALSE, message = TRUE )import_fisheye( filename, channel = 3, circ.mask = NULL, circular = TRUE, gamma = 2.2, stretch = FALSE, display = FALSE, message = TRUE )
filename |
Character. The input image filename. |
channel |
Character. Either the band number corresponding to an image channel or a mixing channel method (Available options are: 'first','GLA','Luma','2BG','BtoRG','B','GEI','RGB'). Default is 3 (Blue channel). |
circ.mask |
List. The circular mask parameters (xc,yc,rc) to be applied to the image. It can be created from a list of available cameras using the |
circular |
Logical. It indicates if the fisheye image is circular (circular=TRUE) or fullframe (circular=FALSE) type. This influences the way the radius is calculated if circ.mask is not inserted. Default is circular. |
gamma |
Numeric. It indicates the input gamma, which is then back-corrected to unity. Default is 2.2 (typical in jpeg images). If no gamma is required, just set gamma=1. |
stretch |
Logical. It indicates if a linear stretch should be applied to enhance contrast. Default FALSE. |
display |
Logical. If is set to TRUE, it plots the image along with the applied mask and a circle radius. Default to FALSE. |
message |
Logical. If is set to TRUE, it prints the mask used for importing the image. Default to TRUE. |
A single-channel image (SpatRaster).
c.im<-system.file('extdata/circular_coolpix4500+FC-E8_chestnut.jpg',package='hemispheR') #set the circular mask automatically: import_fisheye(c.im,circ.mask=list(xc=1136,yc=850,rc=754),channel='B',gamma=2.2,display=TRUE) #list of cameras for circular mask: list.cameras #set the circular mask using camera_fisheye(): import_fisheye(c.im,circ.mask=camera_fisheye('Coolpix4500+FC-E8'), gamma=2.2) #automatic calculating circular mask: import_fisheye(c.im,channel='B',gamma=2.2,display=TRUE) #import a fullframe image: f.im<-system.file('extdata/fullframe_D90_Nikkor-10.5_beech.jpg',package='hemispheR') import_fisheye(f.im,circular=FALSE,channel='B',gamma=2.2,display=TRUE)c.im<-system.file('extdata/circular_coolpix4500+FC-E8_chestnut.jpg',package='hemispheR') #set the circular mask automatically: import_fisheye(c.im,circ.mask=list(xc=1136,yc=850,rc=754),channel='B',gamma=2.2,display=TRUE) #list of cameras for circular mask: list.cameras #set the circular mask using camera_fisheye(): import_fisheye(c.im,circ.mask=camera_fisheye('Coolpix4500+FC-E8'), gamma=2.2) #automatic calculating circular mask: import_fisheye(c.im,channel='B',gamma=2.2,display=TRUE) #import a fullframe image: f.im<-system.file('extdata/fullframe_D90_Nikkor-10.5_beech.jpg',package='hemispheR') import_fisheye(f.im,circular=FALSE,channel='B',gamma=2.2,display=TRUE)
This function imports a SpatRaster image using terra::rast() functionality, and divide into four masks, using the image centre and borders as vertices.
The four zonal masks are then returned as a RasterStack.
zonal_mask(img)zonal_mask(img)
img |
SpatRaster. The input single layer image generated from |
A 4-layers stacks of image masks
image<-system.file('extdata/circular_coolpix4500+FC-E8_chestnut.jpg',package='hemispheR') zmsk<-zonal_mask(terra::rast(image, lyrs=3)) terra::plot(zmsk,col=gray.colors(5),main=c('N','W','S','E'))image<-system.file('extdata/circular_coolpix4500+FC-E8_chestnut.jpg',package='hemispheR') zmsk<-zonal_mask(terra::rast(image, lyrs=3)) terra::plot(zmsk,col=gray.colors(5),main=c('N','W','S','E'))