*Review* **Forensic Facial Comparison: Current Status, Limitations, and Future Directions**

**Nicholas Bacci \* , Joshua G. Davimes , Maryna Steyn and Nanette Briers**

Human Variation and Identification Research Unit, School of Anatomical Sciences, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg 2193, South Africa; joshua.davimes@wits.ac.za (J.G.D.); maryna.steyn@wits.ac.za (M.S.); nanette.briers@wits.ac.za (N.B.)

**\*** Correspondence: nicholas.bacci@wits.ac.za; Tel.: +27-11-717-2204

**Simple Summary:** Facial identification is an emerging field in forensic anthropology, largely due to the rise in closed circuit television presence worldwide, yet there is little published research in it. Our research group has conducted a series of studies testing the validity and reliability of the facial identification practice of morphological analysis. In this paper, we summarize the results of our studies and other latest advances in facial identification practice. In addition, we present a review of relevant technical literature on the limiting factors imposed on facial identification by closed circuit television, while making recommendations for practice and the future of this research niche based on a combination of our results and the technical know-how available. Facial identification research is a multidisciplinary task, with involvement from the field of anatomy, forensic anthropology, photography, image science, and psychology, among others. The value of this brief review is the bridging of these multiple disciplines to discuss the relevant needs and requirements of facial identification in forensic practice and future research.

**Abstract:** Global escalation of crime has necessitated the use of digital imagery to aid the identification of perpetrators. Forensic facial comparison (FFC) is increasingly employed, often relying on poorquality images. In the absence of standardized criteria, especially in terms of video recordings, verification of the methodology is needed. This paper addresses aspects of FFC, discussing relevant terminology, investigating the validity and reliability of the FISWG morphological feature list using a new South African database, and advising on standards for CCTV equipment. Suboptimal conditions, including poor resolution, unfavorable angle of incidence, color, and lighting, affected the accuracy of FFC. Morphological analysis of photographs, standard CCTV, and eye-level CCTV showed improved performance in a strict iteration analysis, but not when using analogue CCTV images. Therefore, both strict and lenient iterations should be conducted, but FFC must be abandoned when a strict iteration performs worse than a lenient one. This threshold ought to be applied to the specific CCTV equipment to determine its utility. Chance-corrected accuracy was the most representative measure of accuracy, as opposed to the commonly used hit rate. While the use of automated systems is increasing, trained human observer-based morphological analysis, using the FISWG feature list and an Analysis, Comparison, Evaluation, and Verification (ACE-V) approach, should be the primary method of facial comparison.

**Keywords:** human identification; facial identification; CCTV; photography; forensic facial comparison; morphological analysis; FISWG; face mapping; disguises

## **1. Introduction**

Cameras and photographic imagery have been used in surveillance, identification, and detection of criminals as early as the 19th century [1]. Anthropological standards have been used to depict portraits of regular criminals for law enforcement registries, similar to today's mugshot system. These registries were intended as a means for witnesses and

**Citation:** Bacci, N.; Davimes, J.G.; Steyn, M.; Briers, N. Forensic Facial Comparison: Current Status, Limitations, and Future Directions. *Biology* **2021**, *10*, 1269. https:// doi.org/10.3390/biology10121269

Academic Editors: Ann H. Ross and Eugénia Cunha

Received: 3 November 2021 Accepted: 1 December 2021 Published: 3 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

victims to conduct a facial review of potential suspects. However, the lack of standardization in image capture processes made these registries ineffective. The advent of judicial photography, in the late 19th century, incorporated anthropometry and relied on standardized conditions of image capture, featuring the well-known anterior and lateral facial views with neutral expression and stance [1,2] routinely used to this day by many police departments throughout the world. The facial anthropometry application was abandoned in favor of the more accepted fingerprint identification system [3], yet the facial image capture standards it relied on endured in facial depiction practices throughout the 20th century [1].

Depicting faces [1], facial anthropometry [2], and facilitating crime scene investigations [4,5] have relied on the use of photography in a forensic context almost since its development [1]. Probably the most recognized use of photography in a forensic setting, and its derivative in the form of video recording, is surveillance. Closed-circuit television (CCTV) was the natural progression of improved use of video technology that allowed for consistent monitoring and review of potential criminal activities [6]. CCTV surveillance systems have since the 1990s become increasingly more common and relied upon throughout the world [7–10] and are in fact considered by many communities the norm in public areas [11,12].

Deployment of CCTV surveillance is considered to act as a deterrent for local crime in monitored areas [8,13,14], often shifting criminal incidents to nearby unmonitored areas instead of completely eliminating them [10]. However, perhaps its most valuable contribution is its frequent use in criminal investigations [8,15]. An analysis of CCTV data in the United Kingdom showed that when CCTV data are available, criminal activity is substantially more likely to be resolved [15]. When the data were not of use, it was primarily due to its lack of availability or some fixed parameter of the surveillance system being suboptimal, such as the incident not being covered by CCTV, the system being faulty, or the images being of insufficient quality [15]. The criteria of usefulness of CCTV recordings vary greatly based on the intended use.

Other than general surveillance and criminal activity monitoring, facial examination is often of interest for the data extracted from many CCTV surveillance systems. This has become more evident as the deployment of CCTV systems and increases in crime have led to an increase in demand for facial identification [16–18]. This rise in demand is a direct outcome of the increased availability of image data, from both CCTV data [7,16] and photographic and video evidence from other sources, such as mobile phones [19].

Forensic facial identification falls under the discipline of facial imaging, which involves the use of visual facial data to assist the identification process [20]. Through the analysis of photographic or video evidence, forensic facial identification is routinely utilized to associate persons of interest to criminal activity [17]. Craniofacial identification involves multiple disciplines, such as facial approximation, facial composites and sketches, age progression and regression, photographic superimposition, molecular photofitting, facial depiction, and facial comparison [20]. Some of these techniques, such as facial approximation and facial composites and sketches, have been researched in some depth [20]. However, forensic facial comparison (FFC) for identification remains largely untested, despite its increasing demand [17,21].

Understanding that forensic facial comparison is a niche of research that needs further development requires the use of clear terminology. A colloquial confusion in terminology between facial identification and recognition is prominent throughout many discussions. This misnomer has been discussed by Schüler and Obertová [22], who clarified that identification is reliant on perfect agreement, which is different from recognition, understood as the innate psychological process humans employ at a glance to recognize a face, usually based on familiarity. Therefore, to attempt facial identification from a forensic anthropological perspective, a strict process of facial comparison is employed. Due to the innate process of recognition in any forensic facial comparison process, the distinction needs to be made clear. Recognition is employed generally as part of the investigative process of facial comparison and is holistic, rapid, and methodologically inconsistent with a high predisposition to error [23,24]. Identification, however, requires further systematic analysis involving standardized, detailed, comprehensive, and meticulously recorded methodology [22]. As such, forensic facial comparison must involve the human-based detailed examination of facial images for identity confirmation [25–27].

Another prominent misconception in facial identification (ID) involves the misuse of the term "facial recognition" to specifically refer to automated or semi-automated facial recognition systems, with this being fully adopted by many in the field of automated facial recognition (e.g., [28,29]). To avoid this miscommunication, certain studies refer to automated facial recognition as facial recognition technology (FRT) or systems [30]; however, this practice is not universally applied.

The misnomer of FRT and facial ID is often closely associated to the misconception of FRT being considered the ideal approach to facial ID. FRT systems apply a variety of computer-based methods to attempt confirmation of facial identity [29,31] and have proven high levels of accuracy in constrained circumstances [28,29]. While great advances have been achieved in the field of FRT [28,32], it remains associated with high false positive rates [32,33], strong racial biases [34], and other ethical concerns around privacy and consent that require resolution prior to the employment of FRT in a legal context. Most concerns revolve around the reliance of FRT systems on biometric information [35] and highly standardized images [36–38], which are often not available in the realistic unstandardized organization of most surveillance installations. As a result, while there are strong commercial and government incentives to deploy FRT systems, in part due to their large market share (USD 3.72 billion) [39], they are still reliant on human-based validation in their operating loops [40]. The need for human validation is further enhanced by the lack of varied databases used to develop and test these FRT systems [41]. Hence, until further varied and realistic databases are used to test and develop these FRTs, human observerbased facial image comparison is considered the preferred approach to facial ID [25,42–44] and will likely persist as the validation method of choice despite the improvement and widespread deployment of FRT systems.

Understanding the limitations and permissible applications of FRTs is crucial to conducting research in both FFC and FRT. The misconceptions and assumptions around FRT and FFC may pose a risk of driving researchers and funders away from conducting research in facial identification. This is primarily because most funders and new researchers would consider facial identification, and particularly FFC, as redundant in an era where FRT has become the norm. Despite these misconceptions, human-based facial identification methods, which are currently employed routinely in the judicial system, rely on forensic facial comparison [17,42].

Facial examination, also referred to as forensic facial comparison (FFC), must be applied using the Analysis, Comparison, Evaluation, and Verification (ACE-V) approach [27], commonly used in other forensic practices, such as fingerprint identification [45]. The ACE-V methodological approach is meant to integrate principles of the scientific method in forensic comparisons in order to enhance their implementation and reliability [45].

In the past, approaches to FFC included photo-anthropometry, facial superimposition, and morphological analysis (MA) [20,27], with morphological analysis being the currently accepted method as advised by both the Facial Identification Scientific Working Group (FISWG) (https://fiswg.org/index.htm accessed on 30 October 2021) and the European Network of Forensic Science Institutes (ENFSI) (https://enfsi.eu/ accessed on 30 October 2021) [27,46]. Application of MA relies on the detailed examination of specific facial features to reach a conclusion with regard to the similarity or dissimilarity of two or more faces [27]. The facial features are assessed subjectively, evaluated, and compared between the faces [27]. The selection of individual facial features often depends on the feature list utilized. Feature lists generally include both overall face composition and structure, individual anatomical feature components (e.g., hairline shape, ear helix morphology, nasal alae protrusion, etc.), and distinguishing characteristics such as scars, blemishes, piercings, and tattoos (e.g., [47]). The current standard feature list used for facial comparison relies on criteria developed by the FISWG for facial comparison by MA [47]. An example of how this analysis is conducted is shown in Figure 1, using sample facial images from the Wits Face Database [41]. alae protrusion, etc.), and distinguishing characteristics such as scars, blemishes, piercings, and tattoos (e.g., [47]). The current standard feature list used for facial comparison relies on criteria developed by the FISWG for facial comparison by MA [47]. An example of how this analysis is conducted is shown in Figure 1, using sample facial images from the Wits Face Database [41].

features to reach a conclusion with regard to the similarity or dissimilarity of two or more faces [27]. The facial features are assessed subjectively, evaluated, and compared between the faces [27]. The selection of individual facial features often depends on the feature list utilized. Feature lists generally include both overall face composition and structure, individual anatomical feature components (e.g., hairline shape, ear helix morphology, nasal

*Biology* **2021**, *10*, × FOR PEER REVIEW 4 of 27

**Figure 1.** Example of a forensic facial comparison analysis process between a wildtype (WT) photograph and a standardized (ST) photograph from the Wits Face Database [41] sample images in the SAPS court chart format. The individual facial features are numbered, analyzed, compared, and evaluated between the two images using the FISWG feature list [47]. Features marked in blue indicate morphological similarity between the two images, while features marked in red indicate morphological dissimilarity. In the example provided, skin color appears different due to lighting discrepancies in the two images (red 1); however, skin texture appears similar (blue 1). The facial images used for Figure 1 are images of the corresponding author of the present manuscript and are part of the sample images of the Wits Face Database [41], reproducible under an open access license distributed under the terms of the Creative Commons Attribution License. This license permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The images can be found in the Wits Face Database data note, including the supplementary material for the Wits Face Database [41]. Recently, our research group (https://www.wits.ac.za/anatomicalsciences/hviru/ ac-**Figure 1.** Example of a forensic facial comparison analysis process between a wildtype (WT) photograph and a standardized (ST) photograph from the Wits Face Database [41] sample images in the SAPS court chart format. The individual facial features are numbered, analyzed, compared, and evaluated between the two images using the FISWG feature list [47]. Features marked in blue indicate morphological similarity between the two images, while features marked in red indicate morphological dissimilarity. In the example provided, skin color appears different due to lighting discrepancies in the two images (red 1); however, skin texture appears similar (blue 1). The facial images used for Figure 1 are images of the corresponding author of the present manuscript and are part of the sample images of the Wits Face Database [41], reproducible under an open access license distributed under the terms of the Creative Commons Attribution License. This license permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The images can be found in the Wits Face Database data note, including the supplementary material for the Wits Face Database [41].

> cessed on 30 October 2021) has conducted a series of validation studies to test the validity and reliability of FFC using the FISWG list (https://fiswg.org/index.htm accessed on 30 October 2021) of morphological features [21,41,48,49]. The aim of this paper is to summarize the results of these findings, thus elucidating the reliability and potential uses of FFC. Recently, our research group (https://www.wits.ac.za/anatomicalsciences/hviru/ accessed on 30 October 2021) has conducted a series of validation studies to test the validity and reliability of FFC using the FISWG list (https://fiswg.org/index.htm accessed on 30 October 2021) of morphological features [21,41,48,49]. The aim of this paper is to summarize the results of these findings, thus elucidating the reliability and potential uses of FFC. Potential areas of caution and observed shortcomings are also discussed. Finally, recommendations as to the minimum standards for CCTV equipment are given, as well as guidelines for future directions in research.
