Lumidigm multispectral whole hand imaging system

As biometric systems are increasingly being used in very large-scale and increasingly critical applications, such as border crossing and container control, the need for extremely accurate systems has arisen. These types of applications require little or no false accepts to increase underlying security and little or no false rejects to increase user convenience and acceptance and prevent the loss of time required for additional checks. Multimodal biometric systems are the best candidates for meeting security requirements because each modality adds assurance of the user’s identity. However, the extra time and actions required for collecting the additional biometrics can decrease user convenience and increase the time needed for user verification. Therefore, a single system which can capture multiple biometrics in a single touch would meet both the security and convenience requirements.

To meet this significant need, Lumidigm developed a multispectral whole-hand imaging system that captures hand shape, fingerprints, and palmprint modalities of a user’s hand by a single user interaction and has the advantages of fast acquisition time and high quality images. The initial development system collects 6 images from the hand, each using different illumination angles, polarization conditions, and wavelengths. Images are collecting using a single 14 megapixel CMOS color image which allows the system to achieve 500 dpi resolution over the entire 9” x 6” platen.

Example of image captures: full hand and metacarpal joint

Because the system is based on Lumidigm’s patented multispectral technology, high quality images are captured in adverse conditions, such as dry or wet hands, dirt, and inconsistent pressure. Software and algorithms have been developed which are able to fuse the multiple hand images together to create a single, high-quality, high-contrast image. This is then segmented for use by the multiple modalities’ matching algorithms. The matching scores are then fused together.

Performance of multispectral whole hand sensor

Initial studies were conducted by Lumidigm to determine the performance of the complete system using 134 unique hands and over 800 images. These studies have shown this to be a very promising technology for fast data collection and extremely high accuracy. On the data set collected, perfect separation was achieved when combining all modalities (Figure 3). This was not achievable based on any of the modalities alone.

Concept for compact multispectral whole hand product

The complete system includes hardware for collecting high quality images of the hand and software to collect images, process the images to the individual biometrics, and perform matching.

Prototype Non-Contact Fingerprint Reader

A non-contact fingerprint sensor is the ultimate for a convenient, secure biometric. Lumidigm is developing such a fingerprint reader that allows you to move your hand in front of the sensor while it automatically captures the fingerprint image and verifies your identity. This sensor would be as easy to use as a barcode reader with the security of a fingerprint reader.

The Lumidigm development system for non-contact fingerprinting uses a new poly-imaging system. The system is based on a combination of photometric stereo and stereo imaging technologies. A hardware development system has been built by customizing and combining two imaging sensors currently used in the Lumidigm Venus fingerprint sensors. The development system is comprised of two synchronized color CMOS imagers that are aligned to view the same region of space located approximately 10 cm above the front surface of each imager’s lens. This region of space is simultaneously illuminated by three LEDs of different colors (red, green, and blue) from three different angles. When a finger is placed in the field of view, both imagers collect images of the finger from two different view points and illuminated from three distinct angles.

Non-contact collection GUI

The interface provides an ability to acquire the synchronized video streams at approximately 10 FPS. The streams go through a circular buffer so that when an acquisition is triggered, the current frame as well as 29 prior frames from both video streams are available for subsequent processing. This capability was incorporated to facilitate a temporal analysis of the data as well as the static analysis.

A graphical user interface for image collection is also a part of the development system. The GUI shows the two separate images as well as a focus metric. The focus metric gives the user visual cues to aid the user in good placement of the finger to capture properly focused images. This metric is based on the vertical and horizontal Sobel gradients that are calculated for both images.

Image capture is initialized by the user. Once initialized, the system detects the presence of a finger in the imaging area and triggers the capture sequence. A series of images is acquired for which a focus profile is computed. At the end of the process the system retains the image pair with the highest focus number for further processing. The focused images can be saved for off-line analysis.

Fingerprint image with minutiae points overlayed

The key aspect of the stereo imaging system is to combine the raw image data into a single quasi-3D representation from which biometric data can be extracted. The algorithmic approach is based on concepts from stereo imaging and shape from color.

Concept non-contact system

Initial biometric performance studies using a small data set of finger images has been conducted by Lumidigm. The matching between the samples lead to a good separation between genuine and impostor scores suggesting strong biometric capabilities of the prototype. Further development will lead to hardware in a very small form factor useful in a wide range of applications.

Genuine and impostor match scores