Fine alignment, flexible fusion: A novel framework of multi-model score fusion in face identification.

Multi-model score fusion was considered a bottleneck problem in forensic face identification. While the score distribution of different face models varies greatly, the existing score processing methods cannot achieve accurate alignment. This paper proposed a score fusion framework named fine alignment and flexible fusion framework (FAFF). In FAFF, we took score-based likelihood ratios as the reference values to align the similarity scores generated by different face models. First, we set up a unified calibration test workflow based on the forensic likelihood ratio test. Then, 3 LLR anchor-based methods (LLRBA1, LLRBA2, and LLRBA3) and LLR curve-based methods (LLRBC) were proposed. Finally, we conducted fusion experiments on four face models (VGGface, Facenet, Arcface, and SFace). The experimental results show that on the CelebA dataset, compared with the existing MOEBA and PAN methods, LLRBC increased the TPR@ 10-7 FPR by 175.4 % and 162.9 %, and LLRBA increased by 55.6 % and 48.5 %.

Two heads are better than one: Dual systems obtain better performance in facial comparison.

Forensic facial image comparison based on recognition algorithms has been widely applied in forensic science. Previous researches have been concentrating on the cases of using single system during comparison, while how to use multiple systems has not yet been studied. In this paper, a dual-systems model (including SeetaFace and FaceNet) for facial comparison was constructed, and Bayesian networks were utilized as the basic frame. In order to prove its superiority, a large-scale experiment (on the dataset CelebA) has been carried on to evaluate the score-based likelihood ratio. We used three likelihood ratio evaluation tools (Empirical Cross-Entropy, Cost Likelihood Ratio, Limit Tippett Plots) to assess the performance of the model. The Wasserstein distance was also used to evaluate the detailed likelihood ratio performance. The experimental results show that the likelihood ratio performance of our dual-systems model is better than single system. Besides, our method of model building and evaluation can also be used in the condition of triple or more systems.

Deep learning features in facial identification and the likelihood ratio bound.

In recent years, the score-based likelihood ratio (SLR) method for facial comparison has attracted considerable research attention. This method relies on the match scores that are calculated from the features obtained from facial recognition systems, deep learning based in particular. However, this concept has not been completely understood. Therefore, this study is aimed at investigating deep learning facial features, and the SLR levels of their match scores. We propose a new interpretation that the deep learning feature is a class characteristic. Based on a large-scale data set experiment, we present evidence that the log SLR value of deep learning features can reach 8 in some data sets. The study results imply that the SLR of deep learning features is a useful method for facial identification, especially when the suspected image is obtained via a CCTV camera.

Structural and optical characterization of banknotes using spectral-domain optical coherence tomography.

Counterfeiting of banknotes remains a severe threat to economic security and social stability. The characterization of banknote has mainly relied on the assessment of various security features applied to the surface of the note. However, the surface features are easy to forge and contain insufficient information to discover the source. In this paper, a novel approach for banknote characterization has been proposed by employing spectral-domain optical coherence tomography (SD-OCT) that can provide structural and optical features. Three groups of counterfeit Chinese 100 Yuan banknotes produced by different printing manners and one group of authentic banknotes were examined by SD-OCT without any sample preparation and four distinct areas were selected for imaging. High-resolution tomographic and three-dimensional (3D) volumetric OCT images were obtained and a set of features were first revealed to characterize the banknotes qualitatively and quantitatively. The results demonstrated that SD-OCT was effective to detect and classify different types of counterfeit banknotes and could potentially be used to link counterfeit banknotes to their sources in a fast, simple and nondestructive manner.

Initial Study for the Determination of the Sequence of Intersecting Lines between Gel Pens and Seals by Optical Coherence Tomography.

Determining the sequence of intersecting lines is a significant issue in the forensic document examination that can reveal the fraud or distinguish between different allegations. Optical coherence tomography (OCT) is a high-resolution cross-sectional imaging technique that has been introduced into forensic science field recently. The potential of OCT as a novel method to determine the sequence of intersecting lines was examined for the first time. In this study, a spectral-domain OCT system with a center wavelength of 900 nm was employed to perform nondestructive examination on determining the sequence of 18 heterogeneous intersecting line samples produced using three types of gel pens and three brands of stamp pad ink seals. Two-dimensional (2D) cross-sectional, and three-dimensional (3D) volumetric images of the intersecting lines were obtained by the OCT system. Several features were noted and analyzed to successfully determine the sequence of all the 18 samples. Blind tests were also conducted to demonstrate the effectiveness of OCT technique. The results illustrate that OCT technology can provide an effective and accurate method for sequencing intersecting lines of gel pen ink and seal ink, which may complement the conventional methods used in the examination of questioned documents.

Detection of latent fingerprint hidden beneath adhesive tape by optical coherence tomography.

Adhesive tape is one type of common item which can be encountered in criminal cases involving rape, murder, kidnapping and explosives. It is often the case that a suspect deposits latent fingerprints on the sticky side of adhesive tape material when tying up victims, manufacturing improvised explosive devices or packaging illegal goods. However, the adhesive tapes found at crime scenes are usually stuck together or attached to a certain substrate, and thus the latent fingerprints may be hidden beneath the tapes. Current methods to detect latent fingerprint hidden beneath adhesive tape need to peel it off first and then apply physical or chemical methods to develop the fingerprint, which undergo complicated procedures and would affect the original condition of latent print. Optical coherence tomography (OCT) is a novel applied techniques in forensics which enables obtaining cross-sectional structure with the advantages of non-invasive, in-situ, high resolution and high speed. In this paper, a custom-built spectral-domain OCT (SD-OCT) system with a hand-held probe was employed to detect fingerprints hidden beneath different types of adhesive tapes. Three-dimensional (3D) OCT reconstructions were performed and the en face images were presented to reveal the hidden fingerprints. The results demonstrate that OCT is a promising tool for rapidly detecting and recovering high quality image of latent fingerprint hidden beneath adhesive tape without any changes to the original state and preserve the integrity of the evidence.

Characterization of automotive paint by optical coherence tomography.

Automotive paint is common trace evidence that plays a significant role in many vehicle-related criminal cases. However, the conventional methods of obtaining tomographic images tend to damage the samples. Optical coherence tomography (OCT) is a novel method to obtain high-resolution and cross-sectional images of the automotive paints in a non-destructive, and high-speed manner. In this study, OCT was applied to image and analyze the automotive paint, using scanning electron microscope (SEM) as reference. Eight automotive paint samples of different brands were examined. The images of multi-layer structures provided by OCT system with 5µm depth resolution were consistent with those by SEM. To distinguish different paints with similar visual appearance, we extracted internal structural features from the images using peak analysis and optical attenuation fit. Six characterized parameters were found to distinguish the samples including the optical path length (OPL) of base coat, the optical attenuation coefficient (OAC) of base coat, the OPL of clear coat, the back-scattering ratio (BSR) of clear coat and base coat, the OPL of primer surfacer, and the BSR of base coat and primer. Statistical differences were evaluated by an independent t-test with p<0.05. OCT was applied to analyze repainted paint as well. Three-dimensional OCT reconstruction of the paints was also implemented to create en face (transverse section) images for morphology examination and comparison. These results suggest that OCT imaging can provide additional new features for analyzing the automotive paints and thereby may be a promising supplement to traditional methods. Meanwhile, the OCT system is favorable for achieving in-situ and real-time examination at the scene of crime.