An R package to compare cartridge cases and compute similarity scores based on their breech face impressions.
You can install the development version of scored from GitHub with:
# install.packages("devtools")
devtools::install_github("jzemmels/scored")A cartridge case is the metal casing that holds the projectile, propellant, and primer of a cartridge (you might refer to them colloquially as a “brass” or “shell”). When a firearm is discharged and the projectile travels out of the barrel, the cartridge case pushes against the back wall of the barrel, known as the breech face, with great force. Markings on the breech face surface, attributable to manufacturing, use, wear, etc., are “stamped” into the surface of the cartridge case, leaving so-called breech face impressions. Analogous to a fingerprint, forensic examiners have used these breech face impressions to identify the firearm from which a cartridge case was fired.
It is possible to capture digital representations, such as images or
topographic scans, of cartridge case surfaces and compare these
representations on a computer. We have developed a suite of tools
available in this R package for automatically comparing cartridge case
scans. The primary intended use of the scored package is to compute
numerical features between pairs of cartridge case scans that measure
the similarity of their breech face impressions.
As an example, consider two cartridge cases fired from the same handgun
(see here to learn
more). The cartridge cases are given unique character-based IDs
K013sA1 and K013sA2. Because these cartridge cases were fired from
the same firearm, we assume that they share similar markings left by the
firearm barrel on their surfaces. The code below loads important
packages and the two cartridge cases.
library(scored)
library(impressions)
library(tidyverse)
data("K013sA1","K013sA2")Below is a visual of the two cartridge case scans using the
impressions::x3pPlot() function. These scans have already undergone
some pre-processing to isolate the breech face impressions, but we skip
the details here. The similarity between these cartridge cases is not
immediately apparent. We can calculate similarity features between these
two scans using functions available in the scored package.
x3pPlot(K013sA1,K013sA2)
To compare the two cartridge case scans, we first try to align them by
determining the rotation/translation at which they are most similar.
This is a standard image processing technique called registration that
is implemented in the comparison_fullScan() function. The thetas
argument specifies one or more rotation angles (in degrees) that we want
to explore for possible registration. The comparison_fullScan()
function performs the registration in both “directions”, meaning it
first registers K013sA2 to K013sA1 and then registers K013sA1 to
K013sA2. The direction column differentiates between these two sets
of results. The output is a data frame-like tibble() object containing
results of the registration process, which are discussed in greater
detail in the vignettes.
comparisonDat_fullScan_estimRotation <- comparison_fullScan(reference = K013sA1,
target = K013sA2,
thetas = seq(-30,30,by = 3))
head(comparisonDat_fullScan_estimRotation)
#> # A tibble: 6 × 12
#> cellIndex x y fft_ccf pairwiseCompCor theta refMissingCount
#> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 1, 1 -91 -130 0.0692 0.470 -30 126310
#> 2 1, 1 -6 -8 0.0654 0.0869 -27 126310
#> 3 1, 1 -87 -110 0.0632 0.363 -24 126310
#> 4 1, 1 -83 -99 0.0717 0.386 -21 126310
#> 5 1, 1 -77 -89 0.0684 0.351 -18 126310
#> 6 1, 1 19 8 0.0714 0.145 -15 126310
#> # ℹ 5 more variables: targMissingCount <dbl>, jointlyMissing <dbl>,
#> # cellHeightValues <named list>, alignedTargetCell <named list>,
#> # direction <chr>comparisonDat_fullScan_estimRotation %>%
group_by(direction) %>%
feature_aLaCarte(features = "all")
#> Parameter 'eps' not specified. Defaulting to eps = 5.
#> Parameter 'minPts' not specified. Defaulting to minPts = 5.
#> Only one cell found. Skipping the density-based feature calculation.
#> Parameter 'threshold' not specified. Defaulting to threshold = 1.
#> direction ccfMean ccfSD pairwiseCompCorAve pairwiseCompCorSD
#> 1 reference_vs_target 0.2695168 NA 0.3998884 NA
#> 2 target_vs_reference 0.2718716 NA 0.4096494 NA
#> xTransSD yTransSD thetaRotSD neighborhoodSizeAve_ave neighborhoodSizeAve_sd
#> 1 NA NA NA 113.09091 NA
#> 2 NA NA NA 97.35443 NA
#> neighborhoodSizeSD_ave neighborhoodSizeSD_sd differenceCor_ave
#> 1 940.7352 NA 0.07415392
#> 2 822.4063 NA 0.08575254
#> differenceCor_sd filteredRatio_ave filteredRatio_sd
#> 1 NA 4.752316 NA
#> 2 NA 4.858481 NAcomparisonData <- comparison_cellBased(reference = K013sA1,target = K013sA2,
numCells = c(8,8),
thetas = seq(-30,30,by = 3),
direction = "both",
returnX3Ps = TRUE)
comparisonData %>%
group_by(direction) %>%
feature_aLaCarte(features = "all",eps = 5,minPts = 5)
#> Parameter 'threshold' not specified. Defaulting to threshold = 1.
#> direction ccfMean ccfSD pairwiseCompCorAve pairwiseCompCorSD
#> 1 reference_vs_target 0.3015478 0.08546018 0.5411682 0.1543828
#> 2 target_vs_reference 0.2488325 0.07746423 0.5513807 0.1503855
#> xTransSD yTransSD thetaRotSD thetaDiff translationDiff clusterSize clusterInd
#> 1 39.20518 37.7141 19.54130 0 1.159502 11 TRUE
#> 2 31.77217 26.2262 14.59369 0 1.159502 11 TRUE
#> neighborhoodSizeAve_ave neighborhoodSizeAve_sd neighborhoodSizeSD_ave
#> 1 54.71608 22.58417 135.9947
#> 2 54.36036 20.13450 138.9213
#> neighborhoodSizeSD_sd differenceCor_ave differenceCor_sd filteredRatio_ave
#> 1 41.95552 0.4287395 0.1986363 2.126379
#> 2 35.80590 0.3636990 0.2136124 2.218079
#> filteredRatio_sd
#> 1 2.013289
#> 2 1.405319