Enhancing active vision system categorization capability through uniform local binary patterns

Previous research in Neuro-Evolution controlled Active Vision Systems has shown its potential to solve various shape categorization and discrimination problems. However, minimal investigation has been done in using this kind of evolved system in solving more complex vision problems. This is partly due to variability in lighting conditions, reflection, shadowing etc., which may be inherent to these kinds of problems. It could also be due to the fact that building an evolved system for these kinds of problems may be too computationally expensive. We present an Active Vision System controlled Neural Network trained by a Genetic Algorithm that can autonomously scan through an image pre-processed by Uniform Local Binary Patterns [8]. We demonstrate the ability of this system to categorize more complex images taken from the camera of a Humanoid (iCub) robot. Preliminary investigation results show that the proposed Uniform Local Binary Pattern [8] method performed better than the gray-scale averaging method of [1] in the categorization tasks. This approach provides a framework that could be used for further research in using this kind of system for more complex image problems.

The final publication is available at Springer via http://dx.doi.org/10.1007/978-3-319-18084-7_3

Paper published as:
Lanihun O., Tiddeman B., Tuci E., Shaw P. (2015) Enhancing Active Vision System Categorization Capability Through Uniform Local Binary Patterns. In: Headleand C., Teahan W., Ap Cenydd L. (eds) Artificial Life and Intelligent Agents. ALIA 2014. Communications in Computer and Information Science, vol 519. Springer, Cham