A survey of image processing techniques and statistics for ballistic specimens in forensic science

Validation of image stream hashing: A forensic method for content verification

How do forensic examiners know if they have altered an image stream when converting a digital image from one codec or file container to another for analysis? Forensic standards and best practices recommend avoiding alteration or degradation of multimedia data during transcoding. An image stream hashing method was recently introduced to the forensic science community to answer the question above. This paper offers an initial validation study of image stream hashing method that may answer the question above. The first half of the study's experiments tested the image stream hashing method to measure fitness for use in forensic science while identifying errors and limitations. The study's second phase analyzed the systematic errors detected in initial tests to discover error causation. Causation analysis identified four method limitations subsequently used to develop proposed standard controls of method operations. The final study phase repeated the initial experiments used in the first phase while implementing the proposed standard controls of method operations. Initial test results indicated the method had significant error rates, limiting the effectiveness of the method to only three of the five file types used in the study. The final testing phase revealed that implementing proposed standard controls of method operations reduced the potential systematic errors to a negligible level when using the image stream hashing method for content verification. The validation study concluded that examiners could use the image stream hashing method for forensic science only by implementing error mitigation techniques that utilize the proposed standard controls of method operations.

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.

Classification of firing pin impressions using HOG-SVM

Crimes, such as robbery and murder, often involve firearms. In order to assist with the investigation into the crime, firearm examiners are asked to determine whether cartridge cases found at a crime scene had been fired from a suspect's firearm. This examination is based on a comparison of the marks left on the surfaces of cartridge cases. Firing pin impressions can be one of the most commonly used of these marks. In this study, a total of nine Ruger model 10/22 semiautomatic rifles were used. Fifty cartridges were fired from each rifle. The cartridge cases were collected, and each firing pin impression was then cast and photographed using a comparison microscope. In this paper, we will describe how one may use a computer vision algorithm, the Histogram of Orientated Gradient (HOG), and a machine learning method, Support Vector Machines (SVMs), to classify images of firing pin impressions. Our method achieved a reasonably high accuracy at 93%. This can be used to associate a firearm with a cartridge case recovered from a scene. We also compared our method with other feature extraction algorithms. The comparison results showed that the HOG-SVM method had the highest performance in this classification task.

Evaluation of photometric stereo and elastomeric sensor imaging for the non-destructive 3D analysis of questioned documents - A pilot study

Photometric Stereo and Elastomeric Sensor Imaging were assessed for measuring the 3-dimensional (3D) morphology of questioned document samples. Photometric stereo is shown to be a powerful non-contact technique for revealing micron level detail of the samples examined. Elastomeric Sensor Imaging is shown to complement photometric stereo by yielding equivalent results. Additionally, this technique allows quantification of the morphological depth information. The techniques were applied to 2 different types of questioned document sample. Firstly, written signatures were examined. Both techniques were able to reveal characteristic features that could be used to infer stroke direction and ink line application sequence. Secondly toner/ink intersections were examined. Both techniques allowed visualisation of 3D features which were used to infer the sequence of application.

Planning, design and logistics of a decision analysis study: The FBI/Ames study involving forensic firearms examiners

This paper describes design and logistical aspects of a decision analysis study to assess the performance of qualified firearms examiners working in accredited laboratories in the United States in terms of accuracy (error rate), repeatability, and reproducibility of decisions involving comparisons of fired bullets and cartridge cases. The purpose of the study was to validate current practice of the forensic discipline of firearms/toolmarks (F/T) examination. It elicited error rate data by counting the number of false positive and false negative conclusions. Preceded by the experimental design, decisions, and logistics described herein, testing was ultimately administered 173 qualified, practicing F/T examiners in public and private crime laboratories. The first round of testing evaluated accuracy, while two subsequent rounds evaluated repeatability and reproducibility of examiner conclusions. This project expands on previous studies by involving many F/T examiners in challenging comparisons and by executing the study in the recommended double-blind format.

The performance of automatic ballistics identification system, BALİSTİKA, for 7.62 mm × 39 mm cartridge case correlation

The one-by-one examination of the increasing Open Case File (OCF) archive by comparison macroscope has become a time-consuming process and the last quarter of the 20th century witnessed the development of automated ballistic identification systems. After the utilization of these systems, government institutes, scholars and the companies devising such systems conducted various performance tests to analyze duration and correlation. The performance of BALİSTİKA and similar automated firearms identification systems are frequently measured using the pistol cartridges. This study brings a new perspective to automated ballistic identification system performance tests through using same and different brand steel 7.62 × 39 mm cartridge cases. The performance of BALİSTİKA system was measured using 4387 comparison cartridge cases according to breech face, firing pin, ejector marks and combined evaluation criteria. It was determined that correlation performance of BALİSTİKA for the same brand cartridges are higher and ejector marks are the most distinct matching criteria for 7.62 × 39 mm steel cartridge cases. The test results are presented under the headings of weapon-manufacturer countries, automated ballistic identification system matching criteria for both same and different cartridge brands.

Statistical models for firearm and tool mark image comparisons based on the congruent matching cells (CMC) method

In the branch of forensic science known as firearm evidence identification, various similarity scores have been proposed to compare firearm marks. Some similarity score comparisons, for example, congruent matching cells (CMC) method, are based on pass-or-fail tests. The CMC method compares the pairwise topography images of breech face impressions, from which the similarity score is derived for quantifying their topography similarity. For an image pair, the CMC method determines a certain number of correlated cell pairs. Next, each correlated pair is determined to be a congruent match cell (CMC) pair, or not based on several identification parameters. The number of CMC pairs as a threshold is required so that the two images of surface topographies can be either identified as matching or determined to be non-matching. To reliably estimate error rates or evaluate likelihood ratio (LR), the key is to find an appropriate probability distribution for the frequency distribution of the observed CMC results. This paper discusses four statistical models for CMC measurements, which are binomial and three binomial-related probability distributions. In previous studies, for a sequence of binomial distributed or other binomial-related distributed random variables (r.v.), the number of Bernoulli trials N for each r.v. is assumed to be the same. However, in practice, N(the number of cell pairs in an image pair) varies from one r.v. (or one image pair) to another. In that case, the term, frequency function, of the CMC results is not appropriate. In this paper, the generalized frequency function is introduced to depict the behavior of the CMC values and its limiting distribution is provided. Based on that, nonlinear regression models are used to estimate the model parameters. The methodology is applied to a set of actual CMC values of fired cartridge cases.

Convergence-improved congruent matching cells (CMC) method for firing pin impression comparison

A firing pin impression is usually concave in shape with a small textured area, which makes it difficult to perform automated algorithm-based comparison. The congruent matching cells (CMC) method was invented for accurate breech face impression comparison, in which a reference impression is divided into correlation cells. Each cell is registered to a cell-sized area of the comparison impression that has maximum similarity in surface topography. Four parameters are used to quantify the congruent matching pattern of the registration position and orientation. This paper aims to further develop the cell-division-matching method based on a convergence feature and to develop practical convergence-improved algorithms for firing pin impression comparison. The convergence feature refers to the tendency of the x-y registration positions of correlated cell pairs to converge at the correct registration angle when comparing same-source samples at different orientations. The areal Gaussian filter is employed to extract high-frequency micro-features; the least-squares matching method is used to improve each cross-correlation precision and reach convergence in the registration positions of correlated cell pairs; and a density-based clustering algorithm is introduced to collect the registration positions of dense cell pairs relative to a virtual common center and to remove outliers. Improvements are achieved in the reliability and accuracy of the number of congruent matching cell pairs (CMCs) collected, which represents the quantification of the degree of pairwise impression similarity. Experiments in this report used 40 firing pin impression samples on cartridge cases fired from 10 pistols. The results included no false identifications or false exclusions.

Bullet signature measurement with chromatic confocal sensor

A new, non-contact, three-dimensional (3D) bullet signature measuring system based on a chromatic confocal sensor is developed. The system is composed of a precision rotary table and a chromatic confocal sensor. The measurement uncertainty of the system is less than 1 µm. When measuring the surface topography of the object, the sensor acquires wavelength information reflected from the object instead of intensity information. This advantage is very suitable to bullet signature measurements. The chromatic confocal sensor works in the point measuring mode and can acquire data continuously with high speed. One round section measurement on the bullet body takes less than 1 minute.

Performance Evaluation of a Registered Ballistic Database Using the Evofinder® System

Firearms for police in China are registered along with their fired bullets and cartridge cases. A Registered Ballistic Database (RBD) of 1000 Norinco QSZ-92 pistols with registered ammunition was established and was evaluated through the Evofinder® system. In this research, 1000 bullets and 1000 cartridge cases were randomly selected and correlated against an RBD of 2996 bullets and 2999 cartridge cases. Examiners found that successful identifications all ranked 1st, supported with land (100%), groove (97%) engraved areas, and primary marks (85.6%) for bullets, and firing pin impressions (99.8%), and breech face marks (99.9%) for cartridge cases. Two known matches (KM) for the same pistol rank in the top two (100%). The distribution of similarity scores varies from marks; however, the Evofinder® system could still effectively distinguish known matches from known nonmatches (KNM) for either bullets or cartridge cases. This study demonstrates the efficiency of the RBD.

Pilot study of feature-based algorithm for breech face comparison

A novel feature-based method, which is scale invariant feature transform (SIFT) and RANdom SAmple Consensus (RANSAC) integration algorithm, is introduced to promote the automated identification of the breech face impression, the most common mark left on the cartridge used for firearm evidence. SIFT algorithm is employed to extract the local extrema from examined impression as keypoints representing its invariant features, and to build the feature descriptor for each keypoint based on its local gradients in neighborhood. RANSAC is used to improve the matching performance among these keypoints and feature descriptors. With hypothesize-and-verify methods, RANSAC is able to construct the best model fitting initial matching pairs of keypoints and to guarantee the robust comparison result. Validation tests using 40 cartridge cases fired from pistols with 10 consecutively manufactured slides yielded a clear separation result, which strongly supports the effectiveness of the ensemble algorithm of SIFT and RANSAC. This application indicates the practical feasibility of feature-based algorithm and image processing technique in forensic science.

Influence of the axial rotation angle on tool mark striations

A tool's axial rotation influences the geometric properties of a tool mark. The larger the axial rotation angle, the larger the compression of structural details like striations. This complicates comparing tool marks at different axial rotations. Using chisels, tool marks were made from 0° to 75° axial rotation and compared using an automated approach Baiker et al. [10]. In addition, a 3D topographic surface of a chisel was obtained to generate virtual tool marks and to test whether the axial rotation angle of a mark could be predicted. After examination of the tool mark and chisel data-sets it was observed that marks lose information with increasing rotation due to the change in relative distance between geometrical details on the tool and the disappearance of smaller details. The similarity and repeatability were high for comparisons between marks with no difference in axial rotation, but decreasing with increased rotation angle from 0° to 75°. With an increasing difference in the rotation angles, the tool marks had to be corrected to account for the different compression factors between them. For compression up to 7.5%, this was obtained automatically by the tool mark alignment method. For larger compression, manually re-sizing the marks to the uncompressed widths at 0° rotation before the alignment was found suitable for successfully comparing even large differences in axial rotation. The similarity and repeatability were decreasing however, with increasing degree of re-sizing. The quality was assessed by determining the similarity at different detail levels within a tool mark. With an axial rotation up to 75°, tool marks were found to reliably represent structural details down to 100μm. The similarity of structural details below 100μm was dependent on the angle, with the highest similarity at small rotation angles and the lowest similarity at large rotation angles. Filtering to remove the details below 100μm lead to consistently higher similarity between tool marks at all angles and allowed for a comparison of marks up to 75° axial rotation. Finally, generated virtual tool mark profiles with an axial rotation were compared to experimental tool marks. The similarity between virtual and experimental tool marks remained high up to 60° rotation after which it decreased due to the loss in quality in both marks. Predicting the rotation angle is possible under certain conditions up to 45° rotation with an accuracy of 2.667±0.577° rotation.

The sharing of ballistics data across Europe and neighbouring territories

The current study explored the use of ballistic examinations and cross-border information sharing across 14 European countries. The presented data were collected using a mixed methods technique consisting of semi-structured interviews and questionnaires that were completed by participants. The results painted a very heterogeneous picture of the use of automated ballistic systems across these countries, as well as how ballistic analyses are integrated in the fight against gun-enabled crime. Three super-ordinates themes emerged from the thematic analysis: use of automated ballistic systems; Ballistic evidence recovery and analysis; knowledge exchange and best practices. The ability to draw firm conclusions regarding the value of ballistics comparison systems, either on a national or cross-border basis, is hampered by inconsistencies regarding data recording practices and definitions. Therefore, key recommendations are suggested to establish better cross border cooperation between member states and develop a better understanding of data sharing procedures.

Computed Tomography Image Origin Identification Based on Original Sensor Pattern Noise and 3-D Image Reconstruction Algorithm Footprints

In this paper, we focus on the "blind" identification of the computed tomography (CT) scanner that has produced a CT image. To do so, we propose a set of noise features derived from the image chain acquisition and which can be used as CT-scanner footprint. Basically, we propose two approaches. The first one aims at identifying a CT scanner based on an original sensor pattern noise (OSPN) that is intrinsic to the X-ray detectors. The second one identifies an acquisition system based on the way this noise is modified by its three-dimensional (3-D) image reconstruction algorithm. As these reconstruction algorithms are manufacturer dependent and kept secret, our features are used as input to train a support vector machine (SVM) based classifier to discriminate acquisition systems. Experiments conducted on images issued from 15 different CT-scanner models of 4 distinct manufacturers demonstrate that our system identifies the origin of one CT image with a detection rate of at least 94% and that it achieves better performance than sensor pattern noise (SPN) based strategy proposed for general public camera devices.

The reference ballistic imaging database revisited

A reference ballistic image database (RBID) contains images of cartridge cases fired in firearms that are in circulation: a ballistic fingerprint database. The performance of an RBID was investigated a decade ago by De Kinder et al. using IBIS(®) Heritage™ technology. The results of that study were published in this journal, issue 214. Since then, technologies have evolved quite significantly and novel apparatus have become available on the market. The current research article investigates the efficiency of another automated ballistic imaging system, Evofinder(®) using the same database as used by De Kinder et al. The results demonstrate a significant increase in correlation efficiency: 38% of all matches were on first position of the Evofinder correlation list in comparison to IBIS(®) Heritage™ where only 19% were on the first position. Average correlation times are comparable to the IBIS(®) Heritage™ system. While Evofinder(®) demonstrates specific improvement for mutually correlating different ammunition brands, ammunition dependence of the markings is still strongly influencing the correlation result because the markings may vary considerably. As a consequence a great deal of potential hits (36%) was still far down in the correlation lists (positions 31 and lower). The large database was used to examine the probability of finding a match as a function of correlation list verification. As an example, the RBID study on Evofinder(®) demonstrates that to find at least 90% of all potential matches, at least 43% of the items in the database need to be compared on screen and this for breech face markings and firing pin impression separately. These results, although a clear improvement to the original RBID study, indicate that the implementation of such a database should still not be considered nowadays.

Quantitative comparison of striated toolmarks

A comparison of striated toolmarks by human examiners is dependent on the experience of the expert and includes a subjective judgment within the process. In this article an automated method is presented for objective comparison of striated marks of screwdrivers. The combination of multi-scale registration (alignment) of toolmarks, that accounts for shift and scaling, with global cross correlation as objective toolmark similarity metric renders the approach robust with respect to large differences in angle of attack and moderate toolmark compression. In addition, a strategy to distinguish between relevant and non-relevant spatial frequency ranges (geometric details) is presented. The performance of the method is evaluated using 3D topography scans of experimental toolmarks of 50 unused screwdrivers. Known match and known non-match similarity distributions are estimated including a large range of angles of attack (15, 30, 45, 60 and 75°) for the known matches. It is demonstrated that the system has high discriminatory power, even if the toolmarks are made at a difference in angle of attack of larger than 15°. The probability distributions are subsequently employed to determine likelihood ratios. A comparison of the results of the automated method with the outcome of a toolmark comparison experiment involving three experienced toolmark examiners reveals, that the automated system is more powerful in correctly supporting the hypothesis of common origin for toolmarks with a large difference in angle of attack (30°). In return, the rate of toolmark comparisons that yield incorrect support for the hypothesis of common origin is higher for the automated system. In addition, a comparison between estimating known match and known non-match distributions using 2D and 3D data is presented and it is shown that for toolmarks of unused screwdrivers, relying on 3D is slightly better than relying on 2D data. Finally, a comparison between estimating known match and known non-match distributions for two different types of screwdrivers suggests, that the method may be used for comparing marks of other tools as well.