Development of a 3D-based automated firearms evidence comparison system

Since the early 1990's, the idea of automated systems for the comparison of microscopic firearms evidence has received considerable attention. The main objective of such systems is to enable the analysis of large amounts of evidence, therefore, transforming the comparison of firearms evidence from an evidence verification tool into a crime-fighting tool. Two such systems have been widely used in United States forensic laboratories: namely, the Integrated Ballistics Identification System (IBIS) (1) and DRUGFIRE (2). Both IBIS and DRUGFIRE have in common the fact that their characterization of a specimen is based on a two-dimensional (2D) representation of the specimen's surface. Although these systems have provided satisfactory results in the identification of cartridge cases, their performance in the identification of bullets has not yet met firearms examiner's expectations. This project was motivated by the premise that a better characterizations of the bullet's surface should translate into better performance of automated identification systems. A three-dimensional (3D) characterization of the bullet's surface is proposed as an alternative to a 2D characterization. This paper discusses the development and preliminary results obtained with SciClops, an automated microscopic comparison system based on the use of a 3D characterization of a bullet's surface.     

A Robust Approach to Automatically Locating Grooves in 3D Bullet Land Scans

Land engraved areas (LEAs) provide evidence to address the same source–different source problem in forensic firearms examination. Collecting 3D images of bullet LEAs requires capturing portions of the neighboring groove engraved areas (GEAs). Analyzing LEA and GEA data separately is imperative to accuracy in automated comparison methods such as the one developed by Hare et al. (Ann Appl Stat 2017;11, 2332). Existing standard statistical modeling techniques often fail to adequately separate LEA and GEA data due to the atypical structure of 3D bullet data. We developed a method for automated removal of GEA data based on robust locally weighted regression (LOESS). This automated method was tested on high-resolution 3D scans of LEAs from two bullet test sets with a total of 622 LEA scans. Our robust LOESS method outperforms a previously proposed “rollapply” method. We conclude that our method is a major improvement upon rollapply, but that further validation needs to be conducted before the method can be applied in a fully automated fashion.     

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

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

Three-dimensional surface reconstruction for cartridge cases using photometric stereo

In forensic science, automated firearms identification is an important and yet unsolved problem. On the way to the solution, one of the most important phases is data acquisition. To be able to identify firearms in a reliable way, all the striated and impressed marks on metallic surfaces of cartridge cases should be visible. But two-dimensional images of cartridge cases are very sensitive to the type and direction of the light source(s). Depending on illumination conditions, the images of marks change drastically and sometimes they simply disappear. But, if the three-dimensional (3D) topography of the surface is obtained, the geometry of the marks, which is independent of the illumination, is available. Thus, by providing illumination independent features that can be used for automated matching, 3D data have the potential to make automated matching much reliable. In the literature on data acquisition for automated firearms identification, a few different ways of three-dimensional surface extraction are described, like laser interferometry or laser profilometry. This study presents a real life application of another method, photometric stereo, for the acquisition of 3D topographic data for cartridge cases, which is the one used in BALISTIKA Ballistics Image Analysis and Recognition System. In order to construct 3D topographic data, first of all, two-dimensional images were acquired using a specially designed set-up. After the images were calibrated radiometrically, photometric stereo method was applied. In order to minimize the low-frequency errors in the final surface, a surface-fitting algorithm was used. The method uses low-cost equipment and image acquisition is not time-consuming. Results were compared to interferometric measurement values for error assessment.     

Automated detection of regions of interest in cartridge case images using deep learning

This paper explores a deep-learning approach to evaluate the position of circular delimiters in cartridge case images. These delimiters define two regions of interest (ROI), corresponding to the breech face and the firing pin impressions, and are placed manually or by an image-processing algorithm. This positioning bears a significant impact on the performance of the image-matching algorithms for firearm identification, and an automated evaluation method would be beneficial to any computerized system. Our contribution consists in optimizing and training U-Net segmentation models from digital images of cartridge cases, intending to locate ROIs automatically. For the experiments, we used high-resolution 2D images from 1195 samples of cartridge cases fired by different 9MM firearms. Our results show that the segmentation models, trained on augmented data sets, exhibit a performance of 95.6% IoU (Intersection over Union) and 99.3% DC (Dice Coefficient) with a loss of 0.014 for the breech face images; and a performance of 95.9% IoU and 99.5% DC with a loss of 0.011 for the firing pin images. We observed that the natural shapes of predicted circles reduce the performance of segmentation models compared with perfect circles on ground truth masks suggesting that our method provide a more accurate segmentation of the real ROI shape. In practice, we believe that these results could be useful for firearms identification. In future work, the predictions may be used to evaluate the quality of delimiters on specimens in a database, or they could determine the region of interest on a cartridge case image.     

Pilot Study of Automated Bullet Signature Identification Based on Topography Measurements and Correlations*†

Abstract: A procedure for automated bullet signature identification is described based on topography measurements using confocal microscopy and correlation calculation. Automated search and retrieval systems are widely used for comparison of firearms evidence. In this study, 48 bullets fired from six different barrel manufacturers are classified into different groups based on the width class characteristic for each land engraved area of the bullets. Then the cross‐correlation function is applied both for automatic selection of the effective correlation area, and for the extraction of a 2D bullet profile signature. Based on the cross‐correlation maximum values, a list of top ranking candidates against a ballistics signature database of bullets fired from the same model firearm is developed. The correlation results show a 9.3% higher accuracy rate compared with a currently used commercial system based on optical reflection. This suggests that correlation results can be improved using the sequence of methods described here.     

Use of an Automated System to Evaluate Feature Dissimilarities in Handwriting Under a Two-Stage Evaluative Process*†

The two-stage evaluative process is an established framework utilized by forensic document examiners (FDEs) for reaching a conclusion about the source(s) of handwritten evidence. In the second, or discrimination, stage, the examiner attempts to estimate the rarity of observations in a relevant background population. Unfortunately, control samples from a relevant background population are often unavailable, leaving the FDE to reach this determination based on subjective experience. Automated handwriting feature recognition systems are capable of performing both feature comparison and discrimination, yet these systems have not been subjected to empirical validation studies. In the present study, we repurposed a commercially available automated system to generate empirical distributions for ranking feature dissimilarity scores among pairs of handwritten phrases. The blinded results of this automated process were used to survey an international cohort of 36 FDEs regarding their strength of support for same- and different-writer propositions. The survey served to cross-validate FDE decision-making under the two-stage approach. Results from the survey demonstrated a clear pattern of response consistent with ground truth. Predictive regression analyses indicated that the automated feature dissimilarity scores and the log of their cumulative distribution functions accounted for 72% of the variability in FDE opinions. This study demonstrated that feature dissimilarity scores acquired using automated processes and their distributions are closely aligned with FDE decision-making processes supporting the heuristic value of the two-stage evaluative framework.     

Defending the scientific foundations of the firearms and tool mark identification discipline: responding to recent challenges

Recent challenges have brought the discipline of firearms and tool mark identification to the forefront in recent court cases. This article reviews those challenges and offers substantial support for the scientific foundations of the firearms and tool mark identification discipline. A careful review of the available literature has revealed that firearms and tool mark identification is rooted in firm scientific foundations, critically studied according to the precepts of the scientific method culminating in the Association of Firearms and Toolmark Examiners' Theory of Identification. Firearms and tool mark identification has been validated in a manner appropriate for evidence of the kind to be expected in firearms and tool mark examinations. Proficiency tests and error rates have been studied and can provide consumers of the disciple with a useful guide as to the frequency with which misidentifications are reported in the community using appropriate methodologies and controls. As a result, the primary issues in recent challenges do not invalidate the firearms and tool mark discipline as a science nor should it detract it from its admissibility in a court of law.     

Automatic Comparison and Evaluation of Impressions Left by a Firearm on Fired Cartridge Cases

Recent years have been characterized by a series of publications in the field of firearms investigation questioning the reliability and objectivity of such examination. This research investigates new solutions to decrease the subjective component affecting the evaluation that follows the comparison of impressions left by a firearm on the surface of spent cartridge cases. An automatic comparison system based on 3D measurements has been developed and coupled to a bivariate evaluative model allowing assigning likelihood ratios. Based on a dataset of 79 pistols (all SIG Sauer 9 mm Luger caliber), the system shows a very high discriminating power and the LRs that it provides are very indicative of the true state under both the prosecution and the defense propositions. For example, likelihood ratios exceeding a billion are predominantly obtained when impressions originating from the same source are compared. The system is also characterized by relatively low rates (≤1%) of misleading evidence depending on the firearm considered.     

Development of a Mobile Toolmark Characterization/Comparison System

Since the development of the striagraph, various attempts have been made to enhance forensic investigation through the use of measuring and imaging equipment. This study describes the development of a prototype system employing an easy‐to‐use software interface designed to provide forensic examiners with the ability to measure topography of a toolmarked surface and then conduct various comparisons using a statistical algorithm. Acquisition of the data is carried out using a portable 3D optical profilometer, and comparison of the resulting data files is made using software named “MANTIS” (Mark and Tool Inspection Suite). The system has been tested on laboratory‐produced markings that include fully striated marks (e.g., screwdriver markings), quasistriated markings produced by shear‐cut pliers, impression marks left by chisels, rifling marks on bullets, and cut marks produced by knives. Using the system, an examiner has the potential to (i) visually compare two toolmarked surfaces in a manner similar to a comparison microscope and (ii) use the quantitative information embedded within the acquired data to obtain an objective statistical comparison of the data files. This study shows that, based on the results from laboratory samples, the system has great potential for aiding examiners in conducting comparisons of toolmarks.     

A Fully Automatic Method for Comparing Cartridge Case Images,

When a gun is fired, it leaves marks on cartridge cases that are thought to be unique to the gun. In current practice, firearms examiners inspect cartridge cases for “sufficient agreement,” in which case they conclude that they come from the same gun, testifying in courts as such. A 2016 President's Council of Advisors on Science and Technology report questioned the scientific validity of such analysis (President's Committee of Advisors on Science and Technology, Washington, DC, Executive Office of the President). One recommendation was to convert firearms analysis to an objective method. We propose a fully automated, open‐source method for comparing breechface marks on cartridge cases using 2D optical images. We improve on existing methodology by automating the selection of marks, and removing the effects of circular symmetry. We propose an empirical computation of a “random match probability” given a known database, which can be used to quantify the weight of evidence. We demonstrate an improvement in accuracy on images from controlled test fires.     

Bite mark documentation and analysis: the forensic 3D/CAD supported photogrammetry approach

Bite mark identification is based on the individuality of a dentition, which is used to match a bite mark to a suspected perpetrator. This matching is based on a tooth-by-tooth and arch-to-arch comparison utilising parameters of size, shape and alignment. The most common method used to analyse bite mark are carried out in 2D space. That means that the 3D information is preserved only two dimensionally with distortions. This paper presents a new 3D documentation, analysis and visualisation approach based on forensic 3D/CAD supported photogrammetry (FPHG) and the use of a 3D surface scanner. Our photogrammetric approach and the used visualisation method is, to the best to our knowledge, the first 3D approach for bite mark analysis in an actual case. The documentation has no distortion artifacts as can be found with standard photography. All the data are documented with a metric 3D measurement, orientation and subsequent analysis in 3D space. Beside the metrical analysis between bite mark and cast, it is possible using our method to utilise the topographical 3D feature of each individual tooth. This means that the 3D features of the biting surfaces and edges of each teeth are respected which is--as shown in our case--very important especially in the front teeth which have the first contact to the skin. Based upon the 3D detailed representation of the cast with the 3D topographic characteristics of the teeth, the interaction with the 3D documented skin can be visualised and analysed on the computer screen.     

3D documentation of footwear impressions and tyre tracks in snow with high resolution optical surface scanning

The three-dimensional documentation of footwear and tyre impressions in snow offers an opportunity to capture additional fine detail for the identification as present photographs. For this approach, up to now, different casting methods have been used. Casting of footwear impressions in snow has always been a difficult assignment. This work demonstrates that for the three-dimensional documentation of impressions in snow the non-destructive method of 3D optical surface scanning is suitable. The new method delivers more detailed results of higher accuracy than the conventional casting techniques. The results of this easy to use and mobile 3D optical surface scanner were very satisfactory in different meteorological and snow conditions. The method is also suitable for impressions in soil, sand or other materials. In addition to the side by side comparison, the automatic comparison of the 3D models and the computation of deviations and accuracy of the data simplify the examination and delivers objective and secure results. The results can be visualized efficiently. Data exchange between investigating authorities at a national or an international level can be achieved easily with electronic data carriers.     

Quantification of Perspective‐Induced Shape Change of Clavicles at Radiography and 3D Scanning to Assist Human Identification

Change in perspective between antemortem and postmortem imaging sessions (radiograph to radiograph and surface scan to radiograph) may cause different 2D renderings of the same osseous element complicating comparisons for identification. In this study, clavicle shape changes due to radiographic positioning and 3D laser scanning were examined in 20 right‐side specimens, as pertinent to chest radiograph comparisons. Results indicate substantial changes in clavicle form with short source‐to‐image receptor distance, elevation of the element from the image receptor, and movement of the element away from the center beam (10% mean square for shape). Although quantitative shape differences were small when the clavicle was in close opposition to the image receptor (3% mean square), important qualitative differences remained with large distances from the center beam (e.g., conoid tubercle presence/absence). The significance of these results for image superimposition and computer‐automated‐shape‐based searches of radiographic libraries to find matching candidates is discussed.     

Identifying Diagonal Cutter Marks on Thin Wires Using 3D Imaging

We present work on matching 2‐mm‐thick wires using optical 3D imaging methods. Marks on such small surfaces are difficult to match using a comparison microscope as this 2D imaging method does not provide height data about the sample surface. Moreover, these 2D microscopy images may be affected by illumination. Hence, the reference and investigated sample should be present at the same time. We employed scanning white light interferometry and confocal microscopy to provide quantitative 3D profiles for reliable comparison of samples that are unavailable for simultaneous analysis. We show that 3D profiling offers a solution by allowing illumination‐independent sample comparison. We correctly identified 74 of 80 profiles using consecutive matching striae (CMS) criteria, and we were able to match samples based on profiles measured using different 3D imaging devices. The results suggest that the used methods allow matching cutter marks on thin wires, which has been difficult previously.     

Human Identification via Lateral Patella Radiographs: A Validation Study,

This research examines the utility of patella outline shape for matching 3D scans of patellae to knee radiographs using elliptical Fourier analysis and subjective methods of human visual comparison of patellae across radiographs for identification purposes. Repeat radiographs were captured of cadaver's knees for visual comparison before patellae were extracted and skeletonized for quantitative comparisons. Quantitative methods provided significant narrowing down of the candidate pool to just a few potential matches (<5% of original sample), while the human analysts showed high capacity for correctly matching radiographs, irrespective of educational level (positive predictive value = 99.8%). The successful computerized matching based on a single quantified patella trait (outline shape) helps explain the potency achieved by subjective visual examination. This work adds to a growing body of studies demonstrating the value of single isolated infracranial bones for human identification via radiographic comparison.     

Portable high‐resolution automated 3D imaging for footwear and tire impression capture

The forensic science community raised the need for improved evidence recognition, collection, and visualization analytical instrumentation for field and laboratory use. While the 3D optical techniques for imaging static objects have been extensively studied, there is still a major gap between current knowledge and collecting high‐quality footwear and tire impression evidence. Among optical means for 3D imaging, digital fringe projection (DFP) techniques reconstruct 3D shape from phase information, achieving camera‐pixel spatial resolution. This paper presents a high‐resolution 3D imaging technology using DFP techniques dedicated to footwear and tire impression capture. We developed fully automated software algorithms and a graphical user interface (GUI) that allow anyone without training to operate for high‐quality 3D data capture. We performed accuracy evaluations and comparisons comparing with the commercial high‐end 3D scanner and carried out qualitative tests for various impressions comparing with the current practices. Overall, our technology achieves similar levels of accuracy and resolution with a high‐end commercially available 3D scanner, while having the merits of being (1) more affordable; (2) much easier to operate; and (3) more robust. Compared with the current practice of casting, our technology demonstrates its superiority because it (1) is non‐destructive; (2) collects more evidence detail than casts, especially when an impression is fragile; (3) requires less time and money to collect each piece of evidence; and (4) results in a digital file that can easily be shared with other examiners.     

Objective Evaluation of Subclass Characteristics on Breech Face Marks

Subclass characteristics can be found on the breech face marks left on spent cartridge cases. Even if they are assumed to be rare and their reported number is small, they can potentially lead to false associations. Subclass characteristics have been studied empirically allowing examiners to recognize them and to understand in which conditions they are produced. Until now, however, their influence on the identification process has not been studied from a probabilistic point of view. In this study, we aim at measuring the effect of these features on the strength of association derived from examinations involving subclass characteristics. The study takes advantage of a 3D automatic comparison system allowing the calculation of likelihood ratios (LRs). The similarities between cartridge case specimens fired by thirteen S&W .40S&W Sigma pistols are quantified, and their respective LRs are computed. The results show that the influence of subclass characteristics on the LRs is limited, even when these features are prevalent among the potential sources considered in a case. We show that the proportion of firearms sharing subclass characteristics should be larger than 40% of the pool of potential firearms for the effect to be significant.     

Validation of chest X-ray comparisons for unknown decedent identification

Comparing skeletal structures between antemortem and postmortem chest radiographs is widely used by forensic specialists from many disciplines to positively identify unknown decedents. However, validity assessments of this method have been fairly limited. This study had three objectives: 1) to quantify the reliability of ante- and postmortem chest radiograph comparison for decedent identification; 2) to identify useful radiologic features supporting decedent identification; and 3) to recognize sources of error in decedent identification related to use of comparative radiographs. A forensic pathologist, a forensic anthropologist, and two radiologists participated in the study. Our results showed that chest radiograph comparisons proved reliable, if basic decedent information was provided, and antemortem and postmortem radiographs were adequately positioned and exposed. A "morphological approach" using normal anatomical structures for comparison may provide the most efficient method for accurate identification.