Classification of Glass Fragments Based on Elemental Composition and Refractive Index*

Abstract: The aim of this study was to assess the efficiency of likelihood ratio (LR)‐based measures when they are applied to solving various classification problems for glass objects which are described by elemental composition, and refractive index (RI) values, and compare LR‐based methods to other classification methods such as support vector machines (SVM) and naïve Bayes classifiers (NBC). One hundred and fifty‐three glass objects (23 building windows, 25 bulbs, 32 car windows, 57 containers, and 16 headlamps) were analyzed by scanning electron microscopy coupled with an energy dispersive X‐ray spectrometer. Refractive indices for building and car windows were measured before (RI_b), and after (RI_a) an annealing process. The proposed scheme for glass fragment(s) classification demonstrates some efficiency, although the classification of car windows (c) and building windows (w) must be treated carefully. This is because of their very similar elemental content. However, a combination of elemental content and information on the change in RI during annealing (ΔRI = RI_a?RI_b) gave very promising results. A LR model for the classification of glass fragments into use‐type categories for forensic purposes gives slightly higher misclassification rates than SVM and NBC. However, the observed differences between results obtained by all three approaches were very similar, especially when applied to the car window and building window classification problem. Therefore, the LR model can be recommended because of the ease of interpretation of LR‐based measures of certainty.     

Recovery of Spray Paint Traces from Clothing by Beating

Manual recovery of spray paints from textiles using a microscope, the routine method in many laboratories, is often laborious. Beating the clothing with a plastic rod, the routine method used for recovery of glass traces within the authors’ laboratory, is proposed as an alternative. The efficiency of the method was evaluated by spray tests with fluorescent paint. In these tests, paint particles in the acquired debris samples, as well as those remaining on the textiles, were investigated. The results show that beating is an efficient way to recover and concentrate paint particles. A good efficiency for jeans fabric and rough knitwear is reported. The results appeared to be less satisfactory for smooth woven fabric. Application of the method in casework was effective for graffiti paints as well as for flaked car paint.     

Application of Infrared and Raman Spectroscopy in Paint Trace Examination

Raman spectroscopy has proved to be a promising technique in forensic examinations, where optical microscopy, micro‐infrared spectroscopy, and microspectrophotometry in the visible and UV range are used for identification and differentiation between paint traces. Often no organic pigments are detected using Fourier transform infrared spectroscopy, because their content in a trace is usually undetectable. Application of a micro‐Raman spectrometer equipped with several excitation lasers helps forensic experts in paint analysis enabling identification of main pigments. Three cases concerning comparative examination of car paint are discussed in detail. The comparison of Raman spectra of paint chips found on clothing of a victim or smears found on body of a damaged car to these of paint chips originated from the suspected car enabled us to identify the car involved in the accident. When no comparative material is available, the method can be useful in establishing the color and make of the car.     

Trace Elemental Analysis of Titanium Dioxide Pigments and Automotive White Paint Fragments for Forensic Examination Using High‐Energy Synchrotron Radiation X‐Ray Fluorescence Spectrometry*

Abstract: High‐energy synchrotron radiation x‐ray fluorescence spectrometry (SR‐XRF) utilizing 116 keV x‐rays was used to characterize titanium dioxide pigments (rutile) and automotive white paint fragments for forensic examination. The technique allowed analysis of K lines of 9 trace elements in 18 titanium dioxide pigments (rutile), and 10 trace elements in finish coat layers of seven automotive white paint fragments. High‐field strength elements (HFSE) were found to strongly reflect the origin of the titanium dioxide (TiO₂) pigments, and could be used as effective parameters for discrimination and classification of the pigments and paint fragments. A pairwise comparison of the finish coat layers of seven automotive white paint fragments was performed. The trace elements in the finish coat layers detected by the high‐energy SR‐XRF were especially effective for identification. By introducing the trace element information of primer and electrocoat layers, all the automotive white paint fragments could be discriminated by this technique.     

Progress Toward the Determination of Correct Classification Rates in Fire Debris Analysis,,

Principal components analysis (PCA), linear discriminant analysis (LDA), and quadratic discriminant analysis (QDA) were used to develop a multistep classification procedure for determining the presence of ignitable liquid residue in fire debris and assigning any ignitable liquid residue present into the classes defined under the American Society for Testing and Materials (ASTM) E 1618‐10 standard method. A multistep classification procedure was tested by cross‐validation based on model data sets comprised of the time‐averaged mass spectra (also referred to as total ion spectra) of commercial ignitable liquids and pyrolysis products from common building materials and household furnishings (referred to simply as substrates). Fire debris samples from laboratory‐scale and field test burns were also used to test the model. The optimal model's true‐positive rate was 81.3% for cross‐validation samples and 70.9% for fire debris samples. The false‐positive rate was 9.9% for cross‐validation samples and 8.9% for fire debris samples.     

Applications of Blue Light‐curing Acrylic Resin to Forensic Sample Preparation and Microtomy

This study discusses the results of an evaluation of a one‐part blue light‐curing acrylic resin for embedding trace evidence prior to the preparation of thin sections with a microtome. Through a comparison to several epoxy resins, the physical properties relevant to both trace evidence examination and analytical microscopy in general, including as viscosity, clarity, color, hardness, and cure speed, were explored. Finally, thin sections from paint samples embedded in this acrylic resin were evaluated to determine if, through smearing or impregnation, the resin contributed to the infrared spectra. The results of this study show that blue light‐curing acrylic resins provide the desired properties of an embedding medium, generate high‐quality thin sections, and can significantly simplify the preparation of paint chips, fibers and a multitude of other types of microscopic samples in the forensic trace evidence laboratory.     

5.6. Hochauflösende ultradünnschicht-isoelektrofokussierung zur differenzierung der PGM1-subtypen in der forensischen blutgruppenkunde

The analytical data (five elemental concentrations plus refractive index) for 349 glass samples have been assessed for the purpose of glass classification by the application of multivariate statistical procedures. The requirements of a classification scheme for glass samples are discussed, and such a scheme for unknown samples in presented. The use of this scheme to classify a glass sample involves a comparison of the glass analysis with a reference collection of analyses. By identifying glasses of similar composition to the unknown it is possible in many cases to classify the sample as sheet, container, tableware, headlamp, etc. The results are expressed as definite classifications, suggested classifications and as unclassifiable. The classification scheme was assessed by means of a blind trial set up within the Metropolitan Police Forensic Science Laboratory. For the type of sample occurring most often in casework the scheme produced a rate of successful definite classifications of 91%. No wrong classifications were obtained in the trial.     

Characterization of Blue Pigments Used in Automotive Paints by Raman Spectroscopy

Micro‐Raman spectroscopy was applied to forensic identification of pigments in paint chips and provided differentiation between paint samples. Sixty‐six blue automotive paint samples, 26 solid and 40 metallic were examined. It was found that the majority of the collected Raman spectra provided information about the pigments present. However, in some cases, fluorescence precluded pigment identification. Using laser excitation at longer wavelengths or pretreatment to effect photobleaching often resulted in reduced fluorescence, particularly for solid color samples, and allowed pigment identification. The examined samples were compared pairwise taking into account number, location, and intensity of absorption bands in their infrared spectra. The estimated discrimination power ranged from 97% for solid paint samples to 99% for metallic paint samples.     

红外差谱技术用于微量混合油漆的比对检验

对交通事故中的漆片检材进行同一物质比对时,某些情况下难以提取到单一的油漆碎片而是两辆车的油漆混合在一起,给物证比对带来困难。红外差谱技术具有辨别混合物中不同成分的功能,因此可应用于交通事故中混合油漆检材的比对检验。本文对一起交通事故的漆片检材,分别用三种方式提取、制样,用红外光谱仪进行分析,借助红外差谱技术对谱图进行处理,排除混合油漆的干扰。处理后得到的谱图与样本的谱图进行比对,结果显示两者主要官能团吸收峰位置、峰型、峰强度基本一致,可出具比对一致的结论。针对混合油漆检材的比对检验,红外差谱技术可提供支持。     

Information‐Theoretical Assessment of the Performance of Likelihood Ratio Computation Methods

Performance of likelihood ratio (LR) methods for evidence evaluation has been represented in the past using, for example, Tippett plots. We propose empirical cross‐entropy (ECE) plots as a metric of accuracy based on the statistical theory of proper scoring rules, interpretable as information given by the evidence according to information theory, which quantify calibration of LR values. We present results with a case example using a glass database from real casework, comparing performance with both Tippett and ECE plots. We conclude that ECE plots allow clearer comparisons of LR methods than previous metrics, allowing a theoretical criterion to determine whether a given method should be used for evidence evaluation or not, which is an improvement over Tippett plots. A set of recommendations for the use of the proposed methodology by practitioners is also given.     

A Comparison of Logistic Regression,Classification and Regression Tree,and Neural Networks Models in Predicting Violent Re-Offending

Previous studies that have compared logistic regression (LR), classification and regression tree (CART), and neural networks (NNs) models for their predictive validity have shown inconsistent results in demonstrating superiority of any one model. The three models were tested in a prospective sample of 1225 UK male prisoners followed up for a mean of 3.31 years after release. Items in a widely-used risk assessment instrument (the Historical, Clinical, Risk Management-20, or HCR-20) were used as predictors and violent reconvictions as outcome. Multi-validation procedure was used to reduce sampling error in reporting the predictive accuracy. The low base rate was controlled by using different measures in the three models to minimize prediction error and achieve a more balanced classification. Overall accuracy of the three models varied between 0.59 and 0.67, with an overall AUC range of 0.65–0.72. Although the performance of NNs was slightly better than that of LR and CART models, it did not demonstrate a significant improvement.     

Progress Toward the Determination of Correct Classification Rates in Fire Debris Analysis II: Utilizing Soft Independent Modeling of Class Analogy (SIMCA)

A multistep classification scheme was used to detect and classify ignitable liquid residues in fire debris into the classes defined by the ASTM E1618‐10 standard method. The total ion spectra (TIS) of the samples were classified by soft independent modeling of class analogy (SIMCA) with cross‐validation and tested on fire debris. For detection of ignitable liquid residue, the true‐positive rate was 94.2% for cross‐validation and 79.1% for fire debris, with false‐positive rates of 5.1% and 8.9%, respectively. Evaluation of SIMCA classifications for fire debris relative to a reviewer's examination led to an increase in the true‐positive rate to 95.1%; however, the false‐positive rate also increased to 15.0%. The correct classification rates for assigning ignitable liquid residues into ASTM E1618‐10 classes were generally in the range of 80–90%, with the exception of gasoline samples, which were incorrectly classified as aromatic solvents following evaporative weathering in fire debris.     

Automation of high volume MPS mixture interpretation using CaseSolver

Sixty samples were sequenced using the ForenSeq kit to produce complex STR DNA mixtures, represented with three separate formats capturing different degrees of sequence information. All mixtures were run through the (open-source) CaseSolver software for comparison against both 10 reference profiles. By comparing the performance of the qualitative and the quantitative models for the different allele formats we found the following preliminary results: The quantitative model performed better than the qualitative model, however the gain was only substantial when the LR was already large. The performance gain of using the longest uninterrupted stretch over using only repeat units was large, whereas the further gain of also using the whole sequence information was small.     

Probabilistic Classification of Age by Third Molar Development: The Use of Soft Evidence

Abstract: This study introduces a method to classify individuals according to an age threshold, given sex and third molars’ dental maturity measured on the Demirjian scale by expressing uncertainty on dental evidence (soft evidence). We introduced a procedure to learn the parameters of the Naïve Bayes model, and we discussed two classification rules. The model was estimated and tested on 559 Italians aged 16–22. Two experts provided the dental evaluations, and the model was estimated for each of them. We evaluated the coherence of the evidence provided by the experts. Some indexes have been proposed to evaluate the effectiveness of the models, emphasizing how the experts’ ability and the technology affect the results. We introduced two benchmarks, one based on the sample distribution per sex and age: in this case, probability of correct classification increases 22% and the proportion of false adults impressively decreases 80.2%; the other benchmark, obtained by simulating hard evidence, shows how the use of soft evidence increases the proportion of correct classification 3.1% and decreases the crucial proportion of false adults about 20%. Similarly, the proportion of false minors decreases about 5.3%.     

Locating,Identifying, and Comparing Sub‐visible Paint Particles

Forensic paint comparisons are generally conducted on samples which, while small relative to their source, are still visible to the unaided eye and are thus located and analyzed without great difficulty. Here we demonstrate that a more detailed examination of candidate transfer surfaces can capture materials (questioned samples), even when such traces are invisible to the unaided eye. While certain analytical details (such as layer sequence or a pure FTIR spectrum) may not be obtainable from such traces due to their size and condition, a detailed analysis of the sample characteristics that are analytically accessible may still provide sufficient analytical data to arrive at a probative result. Here we present the application of this approach to a suspected paint transfer case, involving particles of paint as small as 40 μm in size. Using a combination of stereomicroscopy, polarized light microscopy, infrared microspectroscopy, Raman microspectroscopy, and SEM/EDS, all performed on a single, subsample of the original minute particle, it was possible to demonstrate evidence of a two‐way transfer between the suspected sources. Furthermore, the transferred paint particle in one direction could be classified as automotive in nature based on a combination of polymer composition, microscopic texture, and pigment package (which included three specifically identified pigments). This work demonstrates (i) the potential for improving detection limits when searching for a questioned sample, (ii) the potential benefits of higher resolution analyses on samples that would be traditionally labeled as “sample‐size limited,” and (iii) the value of case‐specific interpretation over standardized, one‐size fits all report templates.     

The Evidentiary Significance of Automotive Paint from the Northeast: A Study of Red Paint

This population study was conducted to assess the frequency of physical, microscopical, and chemical properties of automotive paint chips. Population studies of trace evidence provide valuable analytical data for criminalists to assess evidentiary significance. Two‐hundred automotive paint chips were collected from auto body shops from the Northeastern United States. All samples were analyzed using stereomicroscopy, brightfield, and polarized light microscopy. Red paints were targeted for further analysis using a sequence of modern instrumental techniques commonly used by forensic paint examiners: Fourier‐transform infrared (FT‐IR), Raman, and ultraviolet–visible (UV–Vis) microspectroscopy. The discrimination potential of each analytical method was evaluated by inter‐comparing the paint samples. Results demonstrated that macroscopic and microscopic properties were able to differentiate 99.995% of the population (one undifferentiated pair out of 19,900). When combined with either FT‐IR or UV–Vis microspectroscopy, all paints were differentiated. The results of this research lead to the conclusion that one would not expect to encounter two indistinguishable paint chips originating from different sources during the investigation of a single event.     

Identification of tiny and thin smears of automotive paint following a traffic accident

Three complementary methods in the order of stereomicroscopy, micro-FT-IR spectroscopy and solubility tests were used for the purpose of matching known and questioned paint samples in an auto accident case. Grayish green paint smears scattered on a silvery gray coated plastic bar were taken from a blue car and referred to as questioned samples. Green paint chips were collected as known samples from a green car. These were analyzed to determine whether the paint smears found in the blue car could have been the transfers from the green car. Although each of the three methods, when used alone, suffered from unequal bases for making comparison (i.e., layering whole paint vs. smeared paint), insufficient specificity of methodology and the interfering background coating beneath the smeared paint, the limitations were significantly relieved when three methods were used in combination. Based on the results presented in this report, the questioned grayish green paint smears collected from the blue car and the known green paint chips from the green car are of the same class of paint; that is, the possibility of the above stated paint transfers cannot be eliminated.     

Comparison of feature selection and data fusion of Fourier transform infrared and Raman spectroscopy for identifying watercolor ink

The identification of different kinds of watercolor inks is an important work in the field of forensic science. Four different kinds of watercolor ink Spectroscopy data fusion strategies (Fourier Transform Infrared spectroscopy and Raman spectroscopy) combined with a non-linear classification model (Extreme Learning Machine) were used to identify the brand of watercolor inks. The study chose Competitive Adaptive Reweighted Sampling (CARS), Random Frog (RF), Variable Combination Population Analysis-Genetic Algorithm (VCPA-GA), and Variable Combination Population Analysis-Iteratively Retains Informative Variables (VCPA-IRIV) to extract characteristic variables for mid-level data fusion. The Cuckoo Search (CS) algorithm is used to optimize the extreme learning machine classification model. The results showed that the classification capacity of the mid-level fusion spectra model was more satisfactory than that of single Infrared spectroscopy or Raman spectroscopy. The CS-ELM models based on infrared spectroscopy used to recognize the watercolor ink according to brands (ZHENCAI, DELI, CHENGUANG, and STAEDTLER) obtained an accuracy of 66.67% in the test set using all spectral datasets. The accuracy of CS-ELM models based on Raman spectroscopy was 67.39%. The characteristic wavelength selection algorithms effectively improved the accuracy of the CS-ELM models. The classification accuracy of the mid-level spectroscopy fusion model combined with the VCPA-IRIV algorithm was 100%. The data fusion method increased effectively spectral information. The method could satisfactorily identify different brands of watercolor inks and support the preservation of artifacts, paintings, and forensic document examination.     

A New Index for the MMPI‐2 Test for Detecting Dissimulation in Forensic Evaluations: A Pilot Study

This pilot study is the starting point of a potentially broad research project aimed at identifying new strategies for assessing malingering during forensic evaluations. The forensic group was comprised of 67 males who were seeking some sort of certification (e.g., adoption, child custody, driver's license, issuance of gun permits, etc.); the nonforensic group was comprised of 62 healthy male volunteers. Each participant was administered the MMPI‐2. Statistical analyses were conducted on obtained scores of 48 MMPI‐2 scales. In the first step, parametric statistics were adopted to identify the best combination of MMPI‐2 scales that differentiated the two groups of participants. In the second step, frequency‐based, nonparametric methods were used for diagnostic purposes. Results: A model that utilized the best three predictors (“7‐Pt”, “L,” and “1‐Hs”) was developed and used to calculate the Forensic Evaluation Dissimulation Index (FEDI), which features satisfactory diagnostic accuracy (0.9), sensitivity (0.82), specificity (0.81), and likelihood ratio indices (LR+ = 4.32; LR? = 0.22).     

The random presence of glass and paint on the clothing and footwear of members of the general population: A US baseline survey at various seasons

This study assists the interpretation of glass and paint evidence by filling an existing gap in the background occurrence that reflects the socioeconomic and demographic circumstances in the United States. The collection was performed in a college US city (Morgantown, West Virginia) to determine the effect of the type of clothing worn at different seasons on the presence of glass and paint. Tape lifts and sole scrapings (1038) were collected from 210 participants and up to six clothing and footwear areas per individual. Glass fragments were analyzed via polarized light microscopy (PLM), refractive index (RI), micro-X-Ray fluorescence (μXRF), and scanning electron microscopy-energy dispersive spectroscopy (SEM-EDS), while paint specimens were examined by light microscopy and infrared spectroscopy (μFTIR). Higher occurrences of glass and paint were found in the winter season. The winter collection yielded 10 glass fragments and 68 paint particles, whereas the summer collection resulted in one glass fragment and 23 paint particles. The percentage of individuals with traces varied between seasons; 7% of individuals in the winter and 0.9% in the summer had glass, whereas 36% of individuals in the winter and 19% in the summer bore paint. Lastly, when considering the overall garment and footwear areas, glass was detected in 1.4% of the winter set, compared to 0.2% in the summer collection; paint was found in 9.2% of the winter collection, whereas only 4.2% was found in the summer set. There were no instances where both glass and paint were detected on the clothing and footwear of the same individual.