Automatic Groove Identification via Bayesian Changepoint Detection

Provides functionality to automatically detect groove locations via a Bayesian changepoint detection method to be used in the data preprocessing step of forensic bullet matching algorithms. The methods in this package are based on those in Stephens (1994) . Bayesian changepoint detection will simply be an option in the function from the package 'bulletxtrctr' which identifies the groove locations.


The goal of 'bulletcp' is to easily automate the identification of groove locations via a Bayesian changepoint model on data which are 2D crossections of 3D bullet land scans. Ultimately, this package will potentially support other packages implementing automated bullet land matching algorithms for use by forensic scientists or statisticians. The only function that should ideally be used by a user or another function is get_grooves_bcp(), which takes minimal arguments (though several optional arguments can be supplied) and returns a list. Of the items in the list, the only one that should ideally be needed by anyone is the one called "groove": a two element vector of estimated groove locations.


You can install the released version of 'bulletcp' from CRAN with:



The ideal usage of the package is now demonstrated on the example data included. First, we show what the data should look like.

#> Loading required package: dplyr
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#>     filter, lag
#> The following objects are masked from 'package:base':
#>     intersect, setdiff, setequal, union
#> Loading required package: assertthat
#>       x value
#> 1 0.000    NA
#> 2 0.645    NA
#> 3 1.290    NA
#> 4 1.935    NA
#> 5 2.580    NA
#> 6 3.225    NA
ggplot(data = raw_data) +
  geom_point(aes(x = x, y = value)) +
  theme_bw() +
  ylab("Height") +
#> Warning: Removed 323 rows containing missing values (geom_point).

Next, we use the get_grooves_bcp() function on the raw data to get the groove locations. Downsampled data are used here for speed, but in practice the full data should be used.

# Estimate the groove locations by supplying additional arguments 
raw_data <- raw_data[seq(from = 1, to = nrow(raw_data), by = 30),]
cp_gibbs <- get_grooves_bcp(x = raw_data$x, value = raw_data$value, adjust = 30, iter = 2000)
# Estimated groove locations
#> [1]  68.7000 178.0647
ggplot(data = raw_data) +
  geom_point(aes(x = x, y = value)) +
  theme_bw() +
  ylab("Height") +
  xlab("Width") +
  geom_vline(aes(xintercept = cp_gibbs$groove[1])) +
  geom_vline(aes(xintercept = cp_gibbs$groove[2]))
#> Warning: Removed 11 rows containing missing values (geom_point).


Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


1.0.0 by Nathaniel Garton, a year ago

Browse source code at

Authors: Nathaniel Garton [aut, cre] , Kiegan Rice [ctb]

Documentation:   PDF Manual  

GPL-3 license

Imports Rdpack

Depends on mvtnorm, dplyr, stats, assertthat

Suggests knitr, rmarkdown, ggplot2

See at CRAN