Mathematical Models for Camouflage Pattern Assessment
Abstract
During this work we have developed a first approach to evaluate military camouflage pattern using advanced mathematical and image processing tools, in order to create indices for evaluating artificial patterns over a given background. This development is an innovative method which lay the foundations for evaluate and compare military camouflage patterns, this new approach can help to select the most effective camouflage for a given environment and situation, improving the security of the soldiers during a military operation. From the technical point of view, the main achievement is the development of a computational system, which allow us to differentiate between artificially generated textures, as well as natural ones, with the detection of objects within the image. The main tool is the use of mathematical methods to represent an image in terms of texture and the cartoon area in 2D examples. We improve the classical methods by detecting objects that have the same average of intensity, or may have the same texture but slightly shifted, since the classical discrimination lies on the comparison of the statistical distribution of the local image intensities, while for the proposed method the comparison is between the patches of the surrounding area. As a consequence, we automatically detect the periodicity of the texture and obtain a representation of the texture elements, as a dictionary of Gabor functions. So far, the latter has been implemented in 1D and we are extending it to 2D. This issue is important since a library of textures (backgrounds or camouflage patterns) may be obtained from a database of images. Therefore, it is possible to generate a classification and a common basic texture for each particular scenario, which allow us to measure the performance of a particular camouflage within the different scenarios. All the codes, for the calculation of indices, were implemented in Python and C (Open Source Code platforms).
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 17, 2016
- Accession Number
- AD1006015
Entities
People
- Fernando Padilla
- Jaime Ortega
- Matias Godoy
- Takeshi Asahi
Organizations
- University of Chile