The different techniques for segmentation in eCognition can be used to construct a hierarchical network of image objects which represents image information in different spatial resolutions simultaneously. The image objects are networked, so that each image object “knows” its context (neighborhood), its super-object and its sub-objects. Thus, it is possible to define relations between objects, e.g., “Rel. Border to Forest,” and to utilize this kind of local context information.
Starting at the level of pixels the levels are consecutively numbered.
This hierarchical network is topologically definite, i.e., the border of a super-object is consistent with the borders of its sub-objects. The area represented by a specific image object is defined by the sum of its sub-objects' areas. Technically this is carried into effect relatively simply, since all segmentation techniques used in eCognition are region merging algorithms. Each level is constructed based on its direct sub-objects, i.e., the sub-objects are merged into larger image objects on the next level. Merging is limited by the borders of super-objects; adjacent image objects cannot be merged when they are sub-objects of different super-objects. In eCognition, image objects are defined as being spatially self-consistent.
The hierarchical network of image objects provides possibilities for innovative techniques:
• Structures of different scales can be represented simultaneously and thus classified in relation to each other.
• Different hierarchical levels can be segmented based on different data; an upper layer, for instance can be built based on thematic land register information, whereas a lower layer is segmented using remote sensing data. Classifying the upper level, each land register object can be analyzed based on the composition of its classified sub-objects. By means of this technique different data types can be analyzed in relation to each other.
• The shape of image objects can be corrected based on a regrouping of sub-objects.
Note! It is obvious that in this case the sequence in which the levels are segmented plays an important role. It makes a difference which level is constructed first. In this case, for instance, it would make sense to first build the cadastral level and then to create sub-objects. Again, adjacent sub-objects cannot be merged if they are not sub-objects of the same super-object.
A powerful technique is the analysis of image objects based on sub-objects. For this task you have the following possibilities:
• Texture analysis based on sub-objects, classifying attributes of all sub-objects of an image object on average. Attributes can for instance be contrast or shape.
• Line analysis based on sub-objects.
• Class-related features: relationships to classified sub-objects, such as the relative area of other image objects assigned to a certain class.
Another application of the hierarchical network of image objects is to classify image objects in relation to their respective super-object.
All segmentation procedures provided by eCognition operate on arbitrary levels in this hierarchical network. Since the level of pixels and the level of the whole image always exist by definition, each segmentation of a new level is a construction in between a lower and an upper level. To guarantee a definite hierarchy over the spatial shape of all objects the segmentation procedures follow two rules:
• Object borders must follow borders of objects on the next lower level.
• Segmentation is constrained by the border of the object on the next upper level.
(Excerpts from pages 86-88 of "eCognition Professional - User Guide 4")
*********************************************************************
During image analysis multiple image object levels can be created and layered above the basic pixel level. Two or more image object levels build the image object hierarchy.
Put simply: The image object hierarchy serves as a storage rack for all image objects levels which represent the different shelves storing the image objects. Thus the image object hierarchy provides the working environment for the extraction of image information.
The entirety of image objects is organized into a hierarchical network of image objects. Such a network is called image object hierarchy. It consists of one or more image object levels, from fine resolution on the lowest image object level to coarse resolution on the highest image object level.
Image objects within an image object level are linked horizontally. Similarly, image objects are linked vertically in the image object hierarchy. The image objects are networked in a manner that each image object knows its context, that are its neighbors, its superobject on a higher image object level and its subobjects on a lower image object level. It should be emphasized in this context that even single pixels or single-pixel objects are a special case of image objects. They represent the smallest possible processing scale.
To assure definite relations between image object levels, no image object may have more than one superobject but it can have multiple subobjects. The border of a superobject is consistent with the border of its subobjects.
(Excerpts from pages 26-27 of "Definiens Developer 7 - User Guide")
