cmcR: Open-Source Implementation of the Congruent Matching Cells method

class: center, middle, inverse, title-slide

# cmcR: Open-Source Implementation of the Congruent Matching Cells method
### Joe Zemmels, Dr. Heike Hofmann, Dr. Susan VanderPlas
### 2020-11-30

---

class: primary-blue
## Cartridge Case Comparisons

A general pipeline for comparing two cartridge case scans:

1. **Preprocess** the scans to highlight breech face impressions.

2. **Compare** the two scans and extract similarity features.

3. Formulate a **Decision Rule** to determine whether similarity features support the same-source hypothesis.

Comparison algorithms aim to automate most, if not all, of this process.

---
class: primary-green
# CMC method

- Algorithm developed at NIST that addresses Comparison and Decision Rule steps.

- Involves dividing a reference scan into a grid of cells that search for optimal alignment in target scan.

- Use estimated alignment (translation & rotation) and CCF as similarity features to classify matches.

---
class: primary-blue
## cmcR Package

- CMC method requires user to make many processing decisions.

- How sensitive is the method to different decisions?

- Is there an "optimal" set of decisions?

- cmcR package provides a general framework for investigation:

- `preProcess_*` functions perform initial processing
 
 - `comparison_*` functions extract similarity features
 
 - `decision_*` functions classify CMCs

- Each step is further divided into modularized pieces.

<center> <b> <a href="https://csafe-isu.github.io/cmcR/">https://csafe-isu.github.io/cmcR/</a> </b> </center>

---
class: secondary-green
## Preprocessing Example

```r
fadul1.1 <- x3ptools::x3p_read("https://tsapps.nist.gov/NRBTD/Studies/CartridgeMeasurement/DownloadMeasurement/2d9cc51f-6f66-40a0-973a-a9292dbee36d")

fadul1.1 %>%
  preProcess_crop(region = "exterior",...) %>%
  preProcess_crop(region = "interior",...) %>%
  preProcess_removeTrend(...) %>%
  preProcess_gaussFilter(...)
```

---
class: secondary-blue
## Comparison & Decision Rule

```r
comparisonData <- comparison_allTogether(reference = fadul1.1,
                                         target = fadul1.2,
                                         theta = -24, #degrees
                                         numCells = 64) #8x8 grid
comparisonData <- mutate(comparisonData,
                         originalMethodClassif = decision_CMC(cellIndex,x,y,theta,pairwiseCompCor))
cmcPlot(fadul1.1,fadul1.2,
        reference_v_target_CMCs = comparisonData)
```
<img src="slides_files/figure-html/unnamed-chunk-8-1.png" style="display: block; margin: auto;" />

---
class: secondary-green
## Conclusion

- cmcR provides modularized, open-source tools for algorithmically comparing cartridge case scans.

- As cartridge case comparison algorithms mature, foundation set by cmcR can be developed upon.

<center> <big> <b> Thank You! </b> </big> </center>

---
class: primary-blue
# References

Fadul, T. et al. (2011). _An Empirical Study to Improve the Scientific
Foundation of Forensic Firearm and Tool Mark Identification Utilizing
10 Consecutively Manufactured Slides_.

Hofmann, H. et al. (2020). _x3ptools: Tools for Working with 3D Surface
Measurements_. R package version 0.0.3.

Peng, R. D. (2011). "Reproducible Research in Computational Science".
In: _Science_ 334.6060, pp. 1226-1227.

Song, J. (2013). "Proposed â€œNIST Ballistics Identification System
(NBIS)â€ Based on 3D Topography Measurements on Correlation Cells".
In: _American Firearm and Tool Mark Examiners Journal_ 45.2, p. 11.

Tong, M. et al. (2015). "An Improved Algorithm of Congruent Matching
Cells (CMC) Method for Firearm Evidence Identifications". In: _Journal
of Research of the National Institute of Standards and Technology_ 120,
p. 102. DOI:
[10.6028/jres.120.008](https://doi.org/10.6028%2Fjres.120.008).