Holger Steiner

Design of an Active Multispectral SWIR Camera System for Skin Detection and Face Verification

Biometric face recognition is becoming more frequently used in different application scenarios. However, spoofing attacks with facial disguises are still a serious problem for state of the art face recognition algorithms. This work proposes an approach to face verification based on spectral signatures of material surfaces in the short wave infrared (SWIR) range. They allow distinguishing authentic human skin reliably from other materials, independent of the skin type. We present the design of an active SWIR imaging system that acquires four-band multispectral image stacks in real-time. The system uses pulsed small band illumination, which allows for fast image acquisition and high spectral resolution and renders it widely independent of ambient light. After extracting the spectral signatures from the acquired images, detected faces can be verified or rejected by classifying the material as “skin” or “no-skin.” The approach is extensively evaluated with respect to both acquisition and classification performance. In addition, we present a database containing RGB and multispectral SWIR face images, as well as spectrometer measurements of a variety of subjects, which is used to evaluate our approach and will be made available to the research community by the time this work is published.

Reliable face anti-spoofing using multispectral SWIR imaging

Recent studies point out that spoofing attacks using facial masks still are a severe problem for current biometric face recognition (FR) systems. As such systems are becoming more frequently used, for example, for automated border crossing or access control to critical infrastructure, advanced anti-spoofing techniques are necessary to counter these attacks. This work presents a novel, cross-modal approach that enhances existing solutions for face verification and uses multispectral short wave infrared (SWIR) imaging to ensure the authenticity of a face even in the presence of partial disguises and masks. It is evaluated on a dataset containing 137 subjects and a variety of spoofing attacks. Using a commercial FR system, it successfully rejects all attempts to counterfeit a foreign face with a false acceptance rate FARcf = 0% and most attempts to disguise the own identity with FARdg = 1% at a false rejection rate of FRR <; 5% using SWIR images for verification.

Design of an LED-based sensor system to distinguish human skin from workpieces in safety applications

Commercial light curtains use a technique known as muting to differentiate between work pieces and other objects (e.g., human limbs) based on precise model knowledge of the process. At manually fed machinery (e.g., bench saws), such precise models cannot be derived due to the way the machinery is used. This paper presents a multispectral scanning sensor to classify an objects surface material as a new approach for the problem. The system is meant to detect the presence of limbs and therefore optimized for human skin detection. Evaluation on a test set of skin and (wet) wood samples showed a sufficiently high reliability with respect to safety standards.

FIVIS – A Bicycle Simulation System

This contribution describes the FIVIS project. The project’s goal is the development of an immersive bicycle simulation platform for several applications in the areas of biomechanics, sports, traffic education, road safety and entertainment. To take physical, optical and acoustical characteristics of cycling into account, FIVIS uses a special immersive visualization system, a motion platform and a standard bicycle with sensors and actuators, as well as a surround sound system. First experimental results have shown that the FIVIS simulator provides a realistic training and exercising environment for traffic education and stress research.

A Near-Infrared LED-based Material Classification Sensor System

In safety applications it is often desired that certain materials do not enter a dangerous area. This paper presents a near-infrared LED-based sensor system for robust material classification and ranging up to a distance of 1,000mm.

Material classification through distance aware multispectral data fusion

Safety applications require fast, precise and highly reliable sensors at low costs. This paper presents signal processing methods for an active multispectral optical point sensor instrumentation for which a first technical implementation exists. Due to the very demanding requirements for safeguarding equipment, these processing methods are targeted to run on a small embedded system with a guaranteed reaction time T < 2 ms and a sufficiently low failure rate according to applicable safety standards, e.g., ISO-13849. The proposed data processing concept includes a novel technique for distance-aided fusion of multispectral data in order to compensate for displacement-related alteration of the measured signal. The distance measuring is based on triangulation with precise results even for low-resolution detectors, thus strengthening the practical applicability. Furthermore, standard components, such as support vector machines (SVMs), are used for reliable material classification. All methods have been evaluated for variants of the underlying sensor principle. Therefore, the results of the evaluation are independent of any specific hardware.

Combined Measurement System for the Evaluation of Multi Causal Strain

This work addresses the problem of measuring psychological strain in humans by the use of physiological data. The aim of the work is the research, development and evaluation of a measurement system for the acquisition of such data from humans and the differentiation of psychological and physical strain with the help of machine learning algorithms. The developed system records and analyzes the ECG, the EMG, as well as the skin conductance, and combines these physiological parameters with the subjects physical activity. The main purpose of this measurement system is to assess both types of strain in employees at their workplaces.