Many security & defense systems need to capture their environment in one, two or even three dimensions. Therefore adequate measurement sensors are required that provide fast, accurate and reliable 3D data. With the upcoming range imaging cameras, like the SwissRanger^TM introduced by CSEM Switzerland, new cheap sensors with such ability and high performance are available on the market. Because of the measurement concept these sensors long for a special calibration approach. Due to the implementation of several thousand distance measurement systems as pixels, a standard photogrammetric camera calibration is not sufficient. This paper will present results of investigations on the accuracy of the range imaging camera SwissRanger. A systematic calibration method is presented which takes into consideration the different influencing parameters, like reflectivity, integration time, temperature and distance itself. The analyzed parameters with respect to their impact on the distance measuring pixels and their output data were determined. The investigations were mainly done on the high precision calibration track line in the calibration laboratory at ETH Zurich, which provides a relative accuracy of several microns. In this paper it will be shown, under which circumstances the goal accuracy of the sub centimeter level can be reached. The results of this work can be very helpful for users of range imaging systems to increase their accuracy and thus the reliability of their systems. As an example, the usefulness of a range imaging camera in security systems for room surveillance is presented.

Sensors used for Security purposes have to cover the non-invasive inspection of persons, baggage and letters with the aim to detect weapons, explosives and chemical or biological threat material. Currently, emphasis is placed on system concepts and technologies for this type of applications, employing millimeterwave-, submillimeterwave- and terahertz sensors. This is based on the capability of these frequency bands to look through textiles and the possibility to achieve a geometric resolution which is sufficient to resolve critical items within the necessary range. Using multiple frequencies promises to give more detailed information about the structure of the observed objects. Furthermore, to overcome the limitations of passive millimeter- and submillimeterwave sensors which depend on indirect illumination, systems using miniaturized mmw-radar modules are applied as well. This paper describes two approaches for the detection of concealed weapons, the first using a millimeterwave radiometer on a scanner and the second employing a miniaturized radar module based on a synthetic aperture method.

Range-gated or burst illumination systems have recently drawn a great deal of attention concerning the use for target classification. The development of eye safe lasers and detectors will make these systems ideal to be combined with thermal imagers for long range targeting at night but also for short range security applications like reading of signs and licence plates, looking into cars and buildings etc. Examples of imagery collected for different range and atmospheric conditions will be presented and discussed with respect to image quality and processing techniques.

In this paper, we present techniques related to registration and change detection using 3D laser radar data. First, an experimental evaluation of a number of registration techniques based on the Iterative Closest Point algorithm is presented. As an extension, an approach for removing noisy points prior to the registration process by keypoint detection is also proposed. Since the success of accurate registration is typically dependent on a satisfactorily accurate starting estimate, coarse registration is an important functionality. We address this problem by proposing an approach for coarse 2D registration, which is based on detecting vertical structures (e.g. trees) in the point sets and then finding the transformation that gives the best alignment. Furthermore, a change detection approach based on voxelization of the registered data sets is presented. The 3D space is partitioned into a cell grid and a number of features for each cell are computed. Cells for which features have changed significantly (statistical outliers) then correspond to significant changes.

A compact laboratory demonstrator providing both active polarimetric and multispectral images is designed. Its buildings blocks include, at emission part, a multi-wavelength optical parametric oscillator and, at the reception part, a polarimetric hyperspectral imager. Some of the results obtained with this system are illustrated and discussed. In particular, we show that a multispectral polarimetric image brings additional information on the scene, especially when interpreted in conjunction with its counterpart intensity image, since these two images are complementary in most cases. Moreover, although hyperspectral imaging might be mandatory for recognition of small targets, we evidence that the number of channels can be limited to a set of few wavelengths as far as target detection is considered.

Highly efficient target detection algorithms in hyperspectral remote sensing technology, particularly for the long range detection of very low observable objects which exhibit extremely small detection cross sections, are in great demand. This is more so for a near or real time application. This paper is concerned with global anomaly detections (GAD), and conventional methods to achieve better detection using multiple approach fusion (MAF), which fuses detection outputs from various detectors using either logical operators, or, via a model based estimation of the joint detection statistics from all detectors, is found to be not good enough. This work emphasises the need to integrate a more comprehensive background modelling into the GAD to develop a robust anomaly detector (AD). Then, the detection output from this detector is fused with other detectors via MAF for a further improvement of detection performance. The MUF2 algorithm is formulated exactly using this 2-level fusion mechanism, in which mixture modelling and spectral unmixing fusion have been employed. The significance of background modelling in GAD has been highlighted in this work using real data. The result has shown a factor of 2-5 reduction in detection performance when a very small amount of target pixels (~0.1%) is misclassified as background. This is because anomalies are defined with reference to a model of the background, and subsequently two new background classification techniques have been proposed in this work. The effectiveness of the MUF2 has been assessed using three representative data sets which contain various different types of targets, ranging from vehicles to small plates embedded in backgrounds with various degrees of homogeneity. The performance of MUF2 has been shown to be more superior than the conventional GAD frequently in orders of magnitude, regardless of the background homogeneity and target types. The current version of the MUF2 is run under Matlab and it takes ~2 minutes to process a 20K pixel imagery. This work forms part of the research programme supported by the EMRS DTC established by the UK MOD.

For certain applications the angular distribution of IR radiation from materials and surfaces could be interesting. Various cases have to be taken into account, such as passive scattering, active radiation from material at a given temperature or the response to a thermal excitation. In further development of a full hemispherical scatterometrical device based on an elliptical mirror, an angular resolved IR radiometer has been developed. In contradiction to traditional goniometer designes the set-up allows to measure a hemispherical radiation distribution without moving parts. One application could be the fast acquisition of large amounts of BRDF data in IR range to be compiled into libraries. The use of imaging sensors reduces the measurement time, therefore even time dependence could be addressed. The paper deals with the instrument design, the calibration, and gives some measurement results.

Hyperspectral thermal IR remote sensing is an effective tool for the detection and identification of gas plumes and solid materials. Virtually all remotely sensed thermal IR pixels are mixtures of different materials or temperatures. As sensors improve and hyperspectral thermal IR remote sensing becomes more quantitative, the concept of homogeneous pixels becomes inadequate. The contributions of the constituents to the pixel spectral ground leaving radiance are weighted by their spectral emissivity as well as their temperature, or more correctly, temperature distributions, because real pixels are rarely thermally homogeneous. Planck's Law defines a relationship between temperature and radiance that is strongly wavelength dependent, even for blackbodies. Spectral ground leaving radiance (GLR) from mixed pixels is temperature and wavelength dependent and the relationship between observed radiance spectra from mixed pixels and library emissivity spectra of mixtures of 'pure' materials is indirect. This paper presents results from a simple model of linear mixing of pixel spectral GLR. A pixel consists of one or more materials each with a temperature distribution and an emissivity spectrum. Temperature distributions consistent with high resolution thermal images are used as inputs to the model. The impact of spatial-temporal fluctuation of skin temperature on skin temperature variability will be discussed. The results show the strong sensitivity of spectral GLR at shorter wavelengths to temperature and significant variation of radiance mixture proportions with wavelength in the mid-infrared (3-5μm). Spectral GLR of mixtures in the 8-12μm domain are more modestly impacted but the impact of subpixel mixing and variability is still significant. A demonstration of the effects of linear mixing on linear un-mixing is also presented.

The application of scattering matrix theory in optics has usually been limited to multilayer propagation of plane waves but the analysis of spatial diffraction, the most characteristic effect of any optical element, has had to be considered through techniques much less competent. This paper describes the use of a generalized modal scattering matrix theory as a fast, efficient approach to the analysis of optical systems. In contrast with other methods, the new technique uses a type of optical vortices, called Bessel beams. This rigorous modelling technique has interest in areas of optics as diverse as optical communications, spectrometry, and remote sensing systems. The tactic allows solving both multilayered reflections problems and spatial diffraction phenomena using scattering parameters associated with the transmitted and reflected vortical spectrum.

A numerical technique for gradient-type interface in the inverse scattering problems is presented in this paper. Such an interface is a point at which the velocity profile suffers a jump in first derivative. The time domain approach to scattering from gradient-type interface leads to an integro-differential equation. Using Legendre-Gauss-Lobatto nodes we construct the N^th polynomial interpolation to solve the integro-differential equation. An illustrative example is included to demonstrate the accuracy of the proposed method.