Long-range concealed object detection through active covert illumination

When capturing a scene for surveillance, the addition of rich 3D data can dramatically improve the accuracy of object detection or face recognition. Traditional 3D techniques, such as geometric stereo, only provide a coarse grained reconstruction of the scene and are ill-suited to fine analysis. Photometric stereo is a well established technique providing dense, high-resolution, reconstructions, using active artificial illumination of an object from multiple directions to gather surface information. It is typically used indoors, at short range (<1m), under highly controlled lighting conditions, and so has rarely found application in real-world, outdoor surveillance. One difficulty in such scenes is that the near infrared light generally used to illuminate a scene is overwhelmed by ambient illumination from the sun, and so it is not possible to reliably discriminate the artificial light from the ambient. This work aims to overcome this limitation and counteract the effect of sunlight by exploiting atmospheric absorption at a particular wavelength, 940 nm, in combination with image capture using state-of-the-art black silicon sensors. We first show that this combination, together with low-cost, focused LED illuminators, allows us to reliably differentiate artificial illumination from ambient lighting. We then extend our system to perform photometric stereo reconstruction at distances far beyond the current state of the art under direct sunlight, with a possibility for extending the range to 100 m and beyond. Finally, we provide some example use cases for the photometric stereo data in the form of detection of concealed objects and face recognition.