Can some of the professionals help me with an explanation about the following results in an object base classification in the program eCognition. I did an initial classification of a segmented image using the standard nearest neighbor only. All the 5 classes were correctly classified. As a result it can be said that a clear classification were obtained by selecting only the "Nearest Neighbor" feature space. The following step were to extend the feature space and optimise it by adding the following features: • Mean Layer (1-4) • Standard Deviation Layer (1-4) • Ratio Layer (1-4) • Average Mean difference to neighbors of sub-objects Layer (1-4) • Area of sub-objects: Mean (1) The result after the new classification was now the class urban areas / settlements is totally misclassified, and eliminated. Can somebody please explain this to me?