Estimating error rates for firearm evidence identifications in forensic science

Forensic Science International, Год журнала: 2017, Номер 284, С. 15 - 32

Опубликована: Дек. 13, 2017

Estimating error rates for firearm evidence identification is a fundamental challenge in forensic science. This paper describes the recently developed congruent matching cells (CMC) method image comparisons, its application to identification, and usage initial tests rate estimation. The CMC divides compared topography images into correlation cells. Four parameters are defined quantifying both similarity of correlated cell pairs pattern congruency registered locations. A declared match requires significant number CMCs, i.e., that meet all requirements. Initial testing on breech face impressions set 40 cartridge cases fired with consecutively manufactured pistol slides showed wide separation between distributions numbers observed known non-matching pairs. Another test 95 from different by same process also yielded widely separated distributions. results were used develop two statistical models probability mass function scores. applied framework estimating cumulative false positive negative individual matches non-matches this population impressions. prospect applying large populations realistic case work discussed. can provide foundation identifications, thus emulating methods DNA evidence.

Язык: Английский

---

Valid data based normalized cross-correlation (VDNCC) for topography identification

Neurocomputing, Год журнала: 2018, Номер 308, С. 184 - 193

Опубликована: Май 4, 2018

Язык: Английский

---

Evaluating firearm examiner conclusion variability using cartridge case reproductions

Journal of Forensic Sciences, Год журнала: 2021, Номер 66(5), С. 1704 - 1720

Опубликована: Май 31, 2021

Abstract The forensic science pattern comparison areas, including fingerprints, footwear, and firearms, have been criticized for their subjective nature. While much research has attempted to move these disciplines more objective methods, examiners are still coming conclusions based on own training experience. To complement this subjectivity, black box studies necessary establish the accuracy of feature‐comparison methods. However, when cartridges fired by a firearm create cartridge case test sets there may be significant variability within resulting impressions. This can result in different participants receiving with varying levels difficulty differences impression quality. Therefore, between is not straightforward. compare examiners, method called double‐casting was used plastic reproductions. Double‐casts twenty‐one master cases were created mailed examiners. double‐casts ensured that all comparing exhibits same level detail. tasked determining if unknown each set as three knowns. Automated comparisons also set. results from study showed examiner examining evidence. Furthermore, it shown automated metrics would benefit quality control measure correct any potential errors strengthen conclusions.

Язык: Английский

---

Firearm examination: Examiner judgments and computer‐based comparisons

Journal of Forensic Sciences, Год журнала: 2020, Номер 66(1), С. 96 - 111

Опубликована: Сен. 24, 2020

Abstract Forensic firearm examination provides the court of law with information about source fired cartridge cases. We assessed validity decisions a computer‐based method and 73 examiners who compared breechface firing pin impressions 48 comparison sets. also method's scores examiners' degree‐of‐support judgments latter. The true‐positive rate (sensitivity) true‐negative (specificity) (for both impressions) were 94.4% at least 91.7%, respectively. For examiners, was 95.3% 86.2%. improved when evaluations combined for perceived difficulty decreased. reluctant to provide "difficult" comparisons even though their mostly correct. correlation between low same‐source negligible different‐source comparisons. Combining outcomes methods could increase examinations. numerical not well‐calibrated showed clear signs overconfidence. suggest studying merits performance feedback calibrate these judgments.

Язык: Английский

---

Development and Validation of a Virtual Examination Tool for Firearm Forensics, ,

Journal of Forensic Sciences, Год журнала: 2017, Номер 63(4), С. 1069 - 1084

Опубликована: Окт. 16, 2017

Abstract The transition from 2D imaging to 3D scanning in the discipline of firearms and toolmark analysis is likely provide examiners an unprecedented view microscopic surface topography. digital examination measured topographies has been referred as virtual microscopy ( VM ). approach offers several potential advantages over traditional comparison microscopy. Like any new analytic method, must be validated prior its use a crime laboratory. This paper describes one first validation studies Fifty‐six participants at fifteen laboratories used tools complete two proficiency‐style tests for cartridge case identification. All participating trained correctly reported 100% identifications (known matches) while reporting no false positives. also allowed annotate compared surfaces. These annotations insight into types marked utilized comparative analysis. Overall, results study demonstrate that can successfully conduct support technology

Язык: Английский

---

Results of the 3D Virtual Comparison Microscopy Error Rate (VCMER) Study for firearm forensics

Journal of Forensic Sciences, Год журнала: 2020, Номер 66(2), С. 557 - 570

Опубликована: Окт. 26, 2020

Abstract The digital examination of scanned or measured 3D surface topography is referred to as Virtual Comparison Microscopy (VCM). Within the discipline firearm and toolmark examination, VCM enables review comparison microscopic toolmarks on fired ammunition components. In coming years, this technique may supplement potentially replace light microscope primary instrument used for examination. This paper describes a error rate validation study involving 107 participants. included 40 test sets cartridge cases from firearms with variety makes, models, calibers. Participants commercially available software which allowed data distribution, specimen visualization, submission conclusions. also participants annotate areas similarity dissimilarity support their cohort 76 qualified United States Canadian examiners that completed had an overall false‐positive 3 errors 693 comparisons (0.43%) false‐negative 0 491 (0.0%). accuracy supplemented by participant's provided annotations provide insight into cause consistency across independent examinations conducted in study. ability obtain highly accurate conclusions fires wide range supports hypothesis useful tool within crime laboratory.

Язык: Английский

---

Chromatic confocal microscopy to rapidly reveal nanoscale surface/interface topography by position-sensitive detection

Applied Physics Letters, Год журнала: 2018, Номер 113(8)

Опубликована: Авг. 20, 2018

This paper presents a confocal microscope for the rapid and noninvasive capture of three-dimensional (3D) images showing surface/interface topography in far-field configuration. The proposed features nanoscale optical ruler capable performing fast depth measurements based on chromatic aberration objectives. A diffraction grating position-sensitive detector were used to extract spectral shift, which directly maps height surface with resolution. Because neither mechanical scanning nor spectrometer, limit speed 3D imaging, are needed axial direction, full resolution can be acquired rapidly using conjunction 2D lateral beam scanning. presented technique overcomes limitation detection conventional microscopy. In experiments, static dynamic profiles several representative optoelectronic electronic devices demonstrated. non-invasive approach measurement has considerable promise academic research industrial applications.

Язык: Английский

---

Applying 3D measurements and computer matching algorithms to two firearm examination proficiency tests

Forensic Science International, Год журнала: 2016, Номер 271, С. 98 - 106

Опубликована: Дек. 22, 2016

Язык: Английский

---

Automated Firearm Classification From Bullet Markings Using Deep Learning

IEEE Access, Год журнала: 2020, Номер 8, С. 78236 - 78251

Опубликована: Янв. 1, 2020

Firearm violence is one of the leading causes death in many countries around world, including Thailand. This work proposes a fast and accurate automated method to classify firearm brands from bullet markings. Specifically, panoramic image collected crime scene was captured using developed mobile phone application custom-built portable hardware. The top three state-of-the-art CNNs pretrained on ImageNet-DenseNet121, ResNet50, Xception-were further trained same training set, which composed 718 bullets eight different brands-Beretta, Browning, CZ, Glock, Norinco, Ruger, Sig Sauer, Smith & Wesson-using five-fold cross validation technique. DenseNet121 provided highest AUC 0.99 for CZ classification (the most common registered brand Thailand) average (0.9780 ± 0.0130 SD), significantly higher than those ResNet50 Xception. In addition, there were no interaction effects between CNN model AUC. DenseNet121, had AUC, evaluated test set (72 bullets), results showed that Beretta classifications lowest accuracy (91.18%), followed by Browning Norinco (96.88%), whereas Wesson (98.41%). These suggest based deep learning algorithm hardware have promising potential use at scenes firearms By narrowing down list suspects, this convenient approach can potentially accelerate identification processes forensic science examiners.

Язык: Английский

---

Combined interpretation of objective firearm evidence comparison algorithms using Bayesian networks

Journal of Forensic Sciences, Год журнала: 2024, Номер 69(6), С. 2028 - 2040

Опубликована: Авг. 22, 2024

Abstract Traditionally, firearm and toolmark examiners manually evaluate the similarity of features on two bullets using comparison microscopy. Advances in microscopy have made it possible to collect 3D topographic data, several automated algorithms been introduced for bullet striae these data. In this study, open‐source approaches cross‐correlation, congruent matching profile segments, consecutive striations, a random forest model were evaluated. A statistical characterization was performed four datasets consecutively manufactured firearms provide challenging scenario. Each approach applied all samples pairwise fashion, classification performance compared. Based findings, Bayesian network empirically learned constructed leverage strengths each individual approach, relationship between results, combine them into posterior probability given comparison. The evaluated similarly approaches, results developed classified 99.6% correctly, resultant distributions significantly separated more so than when used isolation.

Язык: Английский

---