A. Genovese, Contactless and less-constrained palmprint recognition, Ph.D. Dissertation, Università degli Studi di Milano, Italy, March 2014.
Biometric systems
Touchless 3D Palmprint Recognition
The project investigates touchless and less-constrained palmprint recognition using a low-cost two-view acquisition system. Traditional palmprint devices generally require users to place their hands on a surface or against positioning pegs, which limits usability, hygiene, and user acceptance. The proposed approach instead captures the palm without physical contact and requires only that the hand be placed approximately within the shared field of view of the cameras. Users are not required to spread their fingers, adopt a precise pose, or keep the hand at a fixed distance from the device.
The acquisition setup combines two cameras with uniform LED illumination to rapidly capture multiple views of the hand. After camera calibration, the acquired images are processed to segment the palm region and reconstruct a metric three-dimensional model. Registration and reprojection techniques are then used to normalize variations in hand pose and acquisition distance, producing representations that are more consistent across different acquisitions.
Recognition combines enhanced palm textures with both two-dimensional and three-dimensional biometric information. Texture features are extracted and matched after geometric normalization, while 3D features provide additional information to refine the comparison. This integration improves robustness to unconstrained positioning and supports accurate identity recognition using a compact and affordable acquisition device.
The project focuses on recognition accuracy, computational efficiency, usability, hygiene, and social acceptance. Its main contribution is a faster and less expensive alternative to conventional contactless 3D palmprint systems, achieved through a simple two-view configuration rather than more complex and costly three-dimensional sensors.
The acquisition setup combines two cameras with uniform LED illumination to rapidly capture multiple views of the hand. After camera calibration, the acquired images are processed to segment the palm region and reconstruct a metric three-dimensional model. Registration and reprojection techniques are then used to normalize variations in hand pose and acquisition distance, producing representations that are more consistent across different acquisitions.
Recognition combines enhanced palm textures with both two-dimensional and three-dimensional biometric information. Texture features are extracted and matched after geometric normalization, while 3D features provide additional information to refine the comparison. This integration improves robustness to unconstrained positioning and supports accurate identity recognition using a compact and affordable acquisition device.
The project focuses on recognition accuracy, computational efficiency, usability, hygiene, and social acceptance. Its main contribution is a faster and less expensive alternative to conventional contactless 3D palmprint systems, achieved through a simple two-view configuration rather than more complex and costly three-dimensional sensors.