Why an RGB-IR camera is crucial for liveness detection in facial recognition systems

In this blog, e-con Systems explores how these RGB-IR cameras work, their benefits in liveness detection, and how they can be used to overcome the challenges posed by spoofing attacks.

Facial recognition systems have progressed far in recent years, with advancements like RGB-IR camera technology playing a crucial role. These cameras offer enhanced imaging capabilities by capturing both visible and infrared (IR) light.

The IR component, often less affected by ambient light conditions, helps detect facial features even in challenging environments. Meanwhile, the RGB component ensures that the system can accurately map these features to their correct colour counterparts.

Recently, e-con Systems has launched See3CAM_CU83 – a state-of-the-art 4K RGB-IR camera with a USB interface. We are the only company offering such a unique 4K RGB-IR camera system with FPGA-based processing.

Before you learn more about this new camera – let’s take a look at how RGB-IR cameras work in facial detection systems and their impact on liveness detection.

Role of RGB-IR cameras in facial recognition systems

Unlike standard RGB cameras, which use a Colour Filter Array (CFA) that only filters red, green, and blue wavelengths, RGB-IR cameras use an additional set of pixels to capture IR light. They can simultaneously capture images in the visible and infrared spectra without the need for mechanical switches.

It also enables multi-band imaging in facial recognition applications. For example, in low-light or dark environments, the IR component can illuminate and capture details that would otherwise be lost. In contrast, the RGB component ensures accurate colour representation in well-lit conditions.

However, several key components must be carefully selected and integrated, such as:

• Selecting a sensor with built-in IR sensitivity is critical. Manufacturers like onsemi and OmniVision offer sensors equipped with CFAs that include IR-sensitive pixels.
• Traditional lenses for colour cameras come with an IR cut filter that blocks wavelengths above 650nm. In an RGB-IR setup, it prevents the camera from capturing the infrared spectrum. So, a dual-band pass filter is used for wavelengths in the 400–650nm range for visible light and 800–950nm for the IR spectrum.
• The ISP needs to separate the combined RGB and IR data into distinct frames. It involves interpolating the RGB output while simultaneously handling the IR frame. Also, an algorithm must be implemented to subtract any IR contamination from the RGB channels.

How RGB-IR cameras improve liveness detection in facial recognition systems

Facial recognition systems are susceptible to spoofing attacks, where fake images, videos, or masks are used to deceive the recognition algorithms. To counter these vulnerabilities, RGB-IR cameras have become important in enhancing liveness detection.

RGB-IR cameras‘ dual-spectrum capability improves the accuracy of liveness detection. The RGB component of the camera captures standard images in visible light, providing colour and texture information. This is used to create a digital representation of the person’s face while comparing against stored templates to identify or verify the individual.

However, the RGB image alone is insufficient for detecting liveness, as it can’t distinguish between a live person and a high-quality photograph or mask. So, the IR camera captures the heat signature emitted by the human body, which is a natural characteristic of living beings.

Since static images and masks do not emit infrared radiation, the presence or absence of this heat signature becomes a reliable indicator of whether the face in front of the camera is live or a spoof.

Hence, when a face is presented to the camera, the system first captures an image in visible light using the RGB sensor. Simultaneously, the IR sensor captures the infrared radiation emitted by the face. The system then compares these two sets of data to determine if they correspond to a live human face.

Benefits of RGB-IR cameras’ liveness detection capabilities

Provides an additional security layer in environments such as financial institutions, government facilities, and critical infrastructure

Ensures quicker and more accurate determinations without the need for separate imaging systems

Minimises the likelihood of false positives, leading to reliable identification and verification processes

Strengthens the facial system’s reliability across a range of scenarios

Proves to be beneficial in low-light or nighttime conditions, where traditional RGB cameras might struggle to capture clear images

e-con Systems’ See3CAM_CU83 RGB-IR camera

e-con Systems has 20+ years of experience in designing, developing, and manufacturing OEM cameras. One of their latest releases is See3CAM_CU83 – an 8 MP, RGB-IR camera with a USB interface to capture both visible and infrared light. This camera is based on a 1/2.9″ AR0830 image sensor from onsemi.

See3CAM_CU83 sends the RAW RGB-IR data from the sensor to the FPGA, separating the RGB and IR data using our proprietary RGB-IR separation algorithm. The RGB data is sent to the ISP, which outputs the Y12, while the IR data is directly streamed.

It is perfect for application use cases, such as biometric access control, in-cabin monitoring, crop health monitoring, image-guided surgeries, and patient monitoring.