Principles of hierarchical neural networks for analysis of multi-images

To construction the control systems of mobile robots are considered principles of hierarchical neural networks for solving the problems of analysis large volume of video information, in particular for the analysis of stereo-and multi-images in real time. To create such control systems is proposed using neural network technology, and to achieve practically acceptable time learning of neural networks is proposed decomposition of a single (global) neural network to the local subnet based on the authors introduced the principle of “cognitive” neurons, and hierarchical organization of the architecture. Such systems can automatically detect and avoid obstacles in the path of the mobile robot through the analysis of multi-image, coming from a single camera.