A standoff biothreat detection and identification system for scanning large areas was designed, built and tested. The sensor is based on two wavelength ultraviolet light induced fluorescence (UVLIF) measured from a distance. The concept calls for multiple sensor modalities, fused to give the required overall performance. It makes use of multiple cameras, ambient light reflectance, high optical power and wavelength modulated UV LED illumination and synchronized fluorescence detection. A two-step operational mode is described along with results from independent demonstrations for each step. The first step is screening of the scene to recognize the surfaces that maximize the chances of biothreat detection and classification. This step used computer vision and artificial intelligence (semantic segmentation) for automation. The material constituting the surface is identified from color images. A second monochrome camera gives total “fluorescence” images excited with an intensity modulated 368nm UV illuminator. The second demonstration is scanning of slides (the “scene” in this case) from 1.2m away, threat detection (the spots on the slides) and classification via active multispectral fluorescence imaging at two different excitation wavelengths (280 and 368nm) and ambient light reflectance at up to 0.5m2/min. It is primarily the surface characteristics that drive the difficulty of the detection and classification of biological warfare agents (BWAs) on surfaces, along with the amount of BWA present on the surface. This presentation details the results obtained, the lessons learned and the envisioned way ahead.
A software application, SIST, has been developed for the simulation of the video at the output of a thermal imager. The approach offers a more suitable representation than current identification (ID) range predictors do: the end user can
evaluate the adequacy of a virtual camera as if he was using it in real operating conditions. In particular, the ambiguity in the interpretation of ID range is cancelled. The application also allows for a cost-efficient determination of the optimal design of an imager and of its subsystems without over- or under-specification: the performances are known early in the development cycle, for targets, scene and environmental conditions of interest. The simulated image is also a powerful method for testing processing algorithms. Finally, the display, which can be a severe system limitation, is also fully
considered in the system by the use of real hardware components. The application consists in Matlabtm routines that
simulate the effect of the subsystems atmosphere, optical lens, detector, and image processing algorithms. Calls to
MODTRAN® for the atmosphere modeling and to Zemax for the optical modeling have been implemented. The realism of the simulation depends on the adequacy of the input scene for the application and on the accuracy of the subsystem
parameters. For high accuracy results, measured imager characteristics such as noise can be used with SIST instead of
less accurate models. The ID ranges of potential imagers were assessed for various targets, backgrounds and atmospheric conditions. The optimal specifications for an optical design were determined by varying the Seidel aberration coefficients to find the worst MTF that still respects the desired ID range.
Automated video monitoring of mobile objects is a growing trend in many sectors, especially in surveillance applications. Many research groups are addressing substantial efforts to develop autonomous applications able to recognize specified events. Reliability of such a system is mainly defined by its ability to extract and to track features of interest in image sequences. Since the capacity to perform this basic task is strongly related to changes in image
contrast, video monitoring units made up mono-spectral sensors offer limited performances in many situations. The best example of such a limitation is the uselessness of a visible CCD camera in low brightness scene. To overcome these shortcomings, we developed an acquisition unit including an uncooled VOx thermal camera (8-12 mm)and a High-Dynamic-Range-CMOS(R) color camera more suitable for outdoor applications. Unlike similar systems, we perform image registration at hardware level rather than at software level. Advantageous characteristics of such a design are presented in this paper. A simple framework is also proposed in order to achieve context-independent event extraction from color and thermal information.