Decision Support for Object Recognition from Multi-Sensor Data

Recognition of objects by remote sensing is a crucial task for reconnaissance and surveillance in the defence and security domain. Beyond an effective processing of the sensor data in order to support the perceptibility of objects of interest for human interpreters, those interpreters also need support to cope with the huge amount of relevant objects and their appearances with respect to different sensors. An object database in conjunction with an efficient way to search objects given their recognition features can obviously improve recognition performance. To recognize an object in optimal time and with maximum accuracy, it becomes moreover crucial for the interpreter to select features which are easily recognized and provide a powerful discrimination at the same time. Under the title RecceMan® (Reconnaissance Manual) the Fraunhofer IITB has developed a software concept to guide mainly image interpreters in the military reconnaissance domain through the complex world of recognition features using optimal feature selection. RecceMan® is right on the way to become the basic decision support tool for the image interpreters in the airborne IMINT (imagery intelligence) units of the German Bundeswehr. The concept can be translated easily to non-imaging sensors and other application domains such as civil security, especially forensic analysis.