Stand-off detection of biological hazards on common surfaces by laser-induced fluorescence (LIF): Improved localization and classification routines

Kurzfassung

Detecting hot-spots of potential biological hazards rapidly is desired during a CBRNe event in order to optimize the sample collection for identification of the potentially hazardous materials. The detection is ideally done from distance to avoid direct human contact. Spectrally and time-resolved laser-induced fluorescence (LIF) with multiple excitation wavelengths is a powerful method for such a biological stand-off challenge. On a path towards small scale, transportable systems for in situ detection, we developed a compact and transportable LIF-based technology demonstrating system. With this system we design, use and evaluate surface-scanning techniques for applied studies on mapping biologically contaminated surfaces. The apparatus returns a highly resolved map of target surfaces, that shows areas of contamination. To this end we use simultaneously acquired spectral and time resolved fluorescence responses from dual wavelength excitation. The recognition is based on machine-learning algorithms and can be used for detection at distances from 3 to 10 m. In our presentation, we address the challenge of detecting biologically hazardous materials on fluorescing and unknown surface materials in realistic scenarios: we discuss the results of our algorithms for an optimized detection on surface materials unknown to the evaluation unit, where the algorithm is designed to separate between background signals and fluorescence responses from contaminations. In addition, we present possible training methods to improve the robustness of the evaluation unit against varying environmental conditions. These influences include e.g. changing daylight conditions, varying laser intensities and others. We also discuss the use of additional, complementary, accompanying information, on the detection efficiency, such as absorption information from monochrome filtered CCD-camera images.