Options
2000
Conference Paper
Titel
ATR components for aerial image analysis
Abstract
The ultimate goal in the analysis of aerial imagery is the interpretation of a scene subject to reconnaissance in the context of a specific task. Target cueing is one of the most prominent examples but also change detection is a common procedure among many others. The information extracted by these procedures as well as the context may vary according to the specific objectives. ATR algorithms were developed to automatically detect, recognise and analyse objects and structures mainly in order to assist in the image interpretation process. The algorithms use mathematical models to represent the a priori known context and the objectives of the specific task. These models are used to recognise spatial and temporal patterns which represent the objects under consideration. They may vary from simple (but very general) "man made object" features to highly complex CAD models. Dynamic models are used to discriminate objects with respect to their temporal behaviour. Different such ATR algorithms for screening, classification, geocoding and image registration, recognition, and trajectory analysis within the framework of military applications will be presented showing different complexity of the underlying models and the use of explicit and implicit models.