rseg.event: Segments SX90 data using noise-up and top-down.

Description Usage Arguments

View source: R/rseg.event.R

Description

Segments SX90 data using noise-up and top-down.

Usage

1
2
3
4
5
6
7
8
9
rseg.event(event, t = 1, cores = 1, memsize = 1e+09, noise = NULL,
  kern = 21L, wt = 5, noisethr = 8, schoolthr = function(S, SBR) 4.4 +
  0.013 * S + 0.1 * SBR, thr = NULL, ind = list(-(1:300), -(49 + (-5:5))),
  subset = NULL, range = NULL, beamend = kern, type = c("median",
  "mean"), noisesmooth = 100, startn = NULL, fresh = FALSE,
  keep.temp = FALSE, save.trh = FALSE, save.data = FALSE,
  save.ind = FALSE, rawevent = NULL, dir.data = NULL, owtsd = TRUE,
  temp = FALSE, track = FALSE, ellipsoid = NA, cmpred = NA,
  nlmp = 200, avoid = NULL, plot = FALSE, ...)

Arguments

event

is the identifier of the event, either given as the number of the event, a string contained in the name of the event, or the path of the event directory.

t

is a vector of the time steps to segment, or a two element vector of ftim values between which rawfiles are extracted from a directory (see the description of 'event').

cores

is the number of cores to use for parallel processing.

noise

can be given as an array of background noise values (either of dimension [J], or [J, I], where J is the length of the beams and I is the number of beams), in which case the background will not be calculated for each time step.

kern

is an integer if the Profus segmentation method is applied, smoothing the data by a median filter of 21 voxels along the beams. If specified as double (not using the "L" behind the number), Gaussian kernel is used along the beams, with a warning.

wt

is the number of time steps used as the width of the median filter along time steps.

noisethr

is a vector of segmentation threshold levels above the noise. One segmentation file is written for each value of 'noisethr'.

schoolthr

is a function of size and mean volume backscattering strength (S_V) defining the threshold level below the 90-percentile of the school using the initial above-noise-threshold. Can also be a vector of values, where one segmentation file is written for each value of 'schoolthr'.

thr

is the fixed threshold to use, in which case 'noisethr' and 'schoolthr' are ignored.

ind

is a list of indexes, as typed into the [] of an array, where 0 and NULL denotes all indexes, and is used to discard regions of the sampling volume of the sonar, such as the inner 100 voxels and the 11 beams behind the vessel (default)..

subset

is a numeric or logical vector/expression indicating elements or rows to keep. Missing values are taken as false, and subset=0 or subset=NULL indicates no subsetting.

range

is a list of elements with names matching names of 'data', specifying the range of the corresponding elements.

beamend

is the number of voxels at the end of the beams which should be discarded from the segmentation.

type

is the function to use across beams (one of "mean" and "median", where "mean" is used in PROFUS up to 2014, while "median" is more robust to the valus of the school).

noisesmooth

is the standard deviation of the Gaussian kernel smoother for the background noise/signal estimate, set by CMR (Christian Michelsen Research) to 100.

startn

is the file segmentation file number written to the first segmentation file. If given as a vector of two elements, the second element is the segmentation file number of the first file using below-school-threshold.

fresh

should be set to FALSE if existing noise-up files should be kept. Otherwise, these are overwritten without warning.

keep.temp

should be set to TRUE to keep all temporary files written by the cores.

save.trh

should be set to TRUE to save the smoothed background noise/signal estimate and the threshold value, 'lenb' values, for each ping. Requires typically 20 KB per ping and 20 MB per 1000 pings. For a mini survey of 12 hours and one ping per 2 seconds, the two variables require more than 500 MB.

save.data

is TRUE to save the actual segmented acoustic data.

rawevent

specifies a directory with raw files which are copied to 'event' and converted to TSD filess before running the segmentation.

dir.data

is the path to the directory in which the projects are stored, defaulted by the variable Acoustics_datasets_directory().

owtsd

is FALSE to not overwrite existing data when copying data from 'rawevent'.

temp

is used in extract_event(), and is TRUE to only return the temporary directory if raw files already exist and toTSD==TRUE, in which case the existing and temporary tsd files are not merged into the existing.

track

is TRUE to track the school recursively in time, to locate the midpoint of the ellipsoid restricting the segmentation of the schools.

ellipsoid

is an optional vector of up to three elements specifying the semi axes of an ellipsoid centered at the center of mass of a user selected region, outside which voxels will not be included in the segmentation. The ellispoid can be set to move with the center of mass of the segmented voxels.

cmpred

is a three column matrix holding the centers of mass of the ellipsoid outside which no samples are discarded from the segmentation mask. This matrix must have one row per time step, starting at the first time step.

nlmp

is the number of center of mass positions of the school used to predict the next center of mass.

avoid

defines a region in which segmentation will not be performed, such as a regio around the vessel track, typically avoid=list(numt=50, dist=30, vessel=TRUE), indicating that all points within 30 m of the last 50 vessel positions should be discarded.

plot

is TRUE to plot the segmentations for each time step, or for each 'plot' time step if 'plot' is numeric.

...

parameters passed to merge_TSD(), particularly 'chunksize'.


arnejohannesholmin/cpplot3d documentation built on March 19, 2018, 9:15 a.m.