Segmentation and Matching in infra-red airborne images using a binary neural network.

This paper describes the application a binary neural network, the Advanced Distributed Associative Memory (ADAM), to the recognition of features in images. In particular, it describes how infra-red images of the ground can be scanned for extraction of road features, urban and field areas. The paper shows how a hierarchical organization of such networks can be used to quickly locate specific points in such images, which can be used for way point identification. The paper describes how the network can be trained rapidly for subsequent application in such tasks. 1 Introduction. This paper describes the results of applying a neural network based associative memory to the task of analysis of infra red images of the ground. The work has formed part of a project to investigate the problems of matching ordnance survey maps to infra-red images (Austin, Brown, Kelly, et.al, 1993). The research described here has investigated the use of an Advanced Distributed Associative Memory (ADAM) (Austin,..