Idemia forensic software extracts human faces, tattoos for investigative leads

Even when a facial recognition system is integrated within a state or federal investigative agency, human intervention is necessary. In cases involving 1:N identification, the process involves facial recognition by a software system followed by facial comparison by a human operator.

The top two algorithms recognized by NIST for facial review for investigative and operational leads cover mugshot images in a database of 12 million individuals with a high accuracy of 99.88 percent. However, in most forensic cases, the images gathered have low quality, necessitating the involvement of humans in adjudicating the facial recognition results.

In a webinar, Idemia says the accuracy of 1:N facial recognition drops to 50 percent when conducted by both laymen and agency employees. However, this accuracy significantly improves when trained facial examiners or reviewers are involved.

Idemia offers 3-day and 5-day training courses for facial examiners or reviewers to ensure maximum accuracy is achieved. Additionally, the Facial Identification Scientific Working Group (FISWG) provides minimum training criteria when using facial recognition systems to ensure that facial reviewers are trained to a certain level at least.

For investigative agencies, Idemia has developed Face Expert, a web-based application built on the MBIS platform. Depending on the customer’s needs, this application can be deployed on-premise or in the cloud. This system enables law enforcement agencies to create photo arrays and conduct descriptor searches.

According to Idemia, the software has the capability to compare unidentified faces with new enrollment records, aiding in the resolution of cold cases, which can be a potential use case.

When utilizing Face Expert, users are required to upload images or videos, which the biometric software then uses to extract faces (even with tattoos). These faces or tattoos are subsequently compared against various databases by trained facial reviewers. The investigative agency has the choice to select the database to be used, the number of probes and candidates, as well as the size at which the images appear on the screen.

In videos, the software identifies individuals based on their eyes and normalizes the images, capturing various expressions and poses. The identified faces are presented as the “best shot” image.

Idemia states that the software offers editing tools for the extracted facial images, enabling users to crop images, adjust the eye platforms, and apply standard FISWG-approved image enhancements. These enhancements may include adjustments to brightness, contrast, and sharpening.

Furthermore, the software has the capability for 3D pose correction.

“Our system is going to let you correct what it thinks these placements are. So it’s going to look and say, well, this is where I found the corners of the eyes, but I’m not sure if I’m right,” explains Nicole Spaun, technical problem manager for facial recognition at Idemia, in the webinar.

Spaun further explains “So you can go in and you can move these points through the image so that instead of having your computer’s best guess at what this face would look frontal, you can go in and correct it and make a frontal approximation of your off-angle image.”

The Face Expert software provides the capability to compare only facial tattoos too. Idemia emphasizes its consistent top performance in NIST tattoo evaluation (Tatt-E). The company says the same tools used for standard facial recognition are also available for tattoo recognition.

Once the operator identifies the most precise match from the database, it undergoes peer review.

Article Topics

biometric identification  |  biometrics  |  Facial Identification Scientific Working Group (FISWG)  |  facial recognition  |  forensics  |  Idemia Public Security  |  MBIS  |  tattoo recognition