Artificial Neural Networks for Plant Classification with Image Processing

Abstract Some investigations about the use and the design of artificial neural networks in the area of plant classification were carried out. For this purpose leaves of 13 different de-cidous trees were scanned and selected features were evaluated with backpropagation networks, Kohonen feature maps and Hopfield networks. Building classification models with backpropagation networks gives best results, but in the light of generalization the classification accuracy was not sufficient. Therefore artificial neural networks have to be used in combination with knowledge-based systems.