Many of different descriptors of spatial properties of natural terrain and objects, in particular different texture descriptors, have been implemented. Using results from detection theory and image quality studies a set of texture measures has been selected by investigation of the amount of necessary uncorrelated measures. Using these we are able to measure the statistical multidimensional difference between terrain areas and object areas in a way that correlates with target acquisition performance.
Developments in the area of signature suppression make it progressively more difficult to recognize targets. Due to the high resolution of modern sensors, it is necessary to take into account small structures as well as the whole target. Measures describing the difference between targets and background are crucial when assessing signature reduction efforts. These measures should be associated with the process of detection of targets in background. Two approaches are feasible, trying to simulate human performance or using an autonomous sensor. In both cases, we have to rely on a set of features discriminating targets from the background. In the spatial domain, we need filters on different scales. The smallest filter will not be able to catch statistical features but has to be based on the use of small image parts like blobs and lines. Larger filter will give statistically relevant feature values. In addition, spectral properties can be used in a multi-dimensional approach investigating targets on different scales, i.e. from very low-resolution to wellresolved objects. Experiments with a new set of featuresand the use of linear discriminant analysis to get overall signature assessment values are described.
Developments in the area of signature suppression make it progressively more difficult to recognize targets. The emphasis has been on the reduction of distinct features, like hot spots in the infrared band. Thus, in order to obtain a low false alarm rate, threat sensors have to utilize more information, i.e. spatial and spectral properties. The purpose of our work is to develop a general tool for camouflage assessment.The approach proposed in this paper is to apply texture descriptors to quantify the similarity between different parts of an image. In addition, other descriptors are used to distinguish man-made object characteristics. The selection of an appropriate set of features is discussed. The assumption is that an area containing observable targets has different statistics than other areas. Statistical properties along with detected target specific features have to be combined with methods used in data fusion. An experiment with a data set of 44 reference images has been carried out, using a recently developed computer program called Terrtex. High correlation with perception experiments was achieved using only one or two texture features. The importance of a careful selection of background area size is finally discussed.
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.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.