EFITA/WCCA 2005 25-28 July 2005, Vila Real, Portugal 2005 EFITA/WCCA JOINT CONGRESS ON IT IN AGRICULTURE The Kohonen type neural network in the process of identification of orchard pests Boniecki P., Weres J., Krysztofiak A., Mueller W. Department of Agricultural Engineering, Agricultural University of Poznań, Poland Abstract One of advantages of suggest a procedure is the ability of the Kohonen neural network to determine the degree of similarity occurring between classes. The Kohonen network can be also used to detect regularities occurring in the obtained empirical data. If at the network input, a new unknown case appears which the network is unable to recognise, it means that it is different from all the classes known previously. The Kohonen network taught in this way can serve as a detector signalling the appearance of a widely understood novelty. Such a network can also look for similarities between the known data and the noisy data. In this way, it is able to identify fragments of images presenting photographs of e.g. orchard pests. Key words: Kohonen neural networks, recognition of an image. Introduction The Kohonen neural networks are modelled on the topological properties of the human brain. These networks are also known as self-organizing feature maps (SOFM - Self Organizing Feature Maps). The Kohonen networks are most frequently used for a widely understood classification (Tadeusiewicz R., 1990). They perform this task in a relatively untypical way. Thanks to the processing of output values carried out within the postprocessing, the result of the network activity is an output variable of nominal character. Each value of this variable represents one definite class. Appropriate neurons occurring in the output layer of the network correspond to particular classes. The connection of the neuron with a given class is indicated by the label with the class name attributed to it. During the action of the network, after the input signal has appeared, the winning neuron is indicated each time (that is the neuron of the lowest activation, which indicates the highest compatibility of weights and the given input pattern). The label of the neuron determines the class to which the input case is ascribed. This untypical structure enables the user to define the output layer of the Kohonen network as a specific two-dimensional "map" of the input multi-dimensional data set. It enables to place in it an optional number of neurons that are distinguished and established points in this map. The topology of the Kohonen network differs considerably from the structures of other neural networks. The structure discussed is basically a one-layer network. It consists of an input layer and an output layer which processes the data presented. The output layer is built of radial neurons. This layer is also defined as a layer forming a topological map. Neurons in the layer forming the topological map are taken into consideration as if they were placed in space according to some predetermined order - usually for convenience and better perception we imagine them to be nodes of a two-dimensional network. 1225