The sensory perception of other vehicles in cluttered environments is an essential component of situational awareness for a mobile vehicle. However, vehicle detection is normally applied to visible imagery sequences, while in this paper we investigate how polarized, infrared imagery can add additional discriminatory power. Using knowledge about the properties of the objects of interest and the scene environment, we have developed a polarimetric cluster-based descriptor to detect vehicles using long-wave infrared radiation in the range of 8-12 μm. Our approach outperforms both intensity and polarimetric image histogram descriptors applied to the infrared data. For example, at a false positive rate of 0.01 per detection window, our cluster approach results in a true positive rate of 0.63 compared to a rate of 0.05 for a histogram of gradient descriptor trained and tested on the same dataset. In conclusion, we discuss the potential of this new approach in comparison with state-of-the-art infrared and conventional video detection.
We perform a systematic study of the factors governing the optical properties of type II Ga 1−x In x Sb/InAs superlattice structures. We map the parameter space corresponding to the layer widths, alloy concentrations and interface bonding types, and identify those structures for which the fundamental gap lies in the desired range for device application. In addition, we examine the higher lying miniband energies to assess the structures for favourable Auger recombination limits. The microscopic interface bonding configuration is shown to have a significant impact upon the magnitude of the fundamental gap, and confirms the requirement for full-bandstructure calculations in the evaluation of possible structures. We study the features of the optical spectra for those structures whose bandstructures are recognized as most suitable for detector applications.
This paper presents a new, practical infrared video based surveillance system, consisting of a resolution-enhanced, automatic target detection/recognition (ATD/R) system that is widely applicable in civilian and military applications. To deal with the issue of small numbers of pixel on target in the developed ATD/R system, as are encountered in long range imagery, a super-resolution method is employed to increase target signature resolution and optimise the baseline quality of inputs for object recognition. To tackle the challenge of detecting extremely low-resolution targets, we train a sophisticated and powerful convolutional neural network (CNN) based faster-RCNN using long wave infrared imagery datasets that were prepared and marked in-house. The system was tested under different weather conditions, using two datasets featuring target types comprising pedestrians and 6 different types of ground vehicles. The developed ATD/R system can detect extremely low-resolution targets with superior performance by effectively addressing the low small number of pixels on target, encountered in long range applications. A comparison with traditional methods confirms this superiority both qualitatively and quantitatively.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.