Future weapon systems will have to meet increasing demands regarding precision, performance and autonomy, which in turn leads to a need of advanced control by sensor feedback, paying due respect to the geometrical structure of the sensor signals. Sensor data may consist of a picture or a transformation. Computer vision provides rich examples of this. The methods studied in the present report should be useful for several modern sensor systems. This report treats object recognition and position determination with a method using differential invariants. The method is developed for possible use in a control loop for vehicle guidance. The method consists in calculating of the invariants for a curve under some transformation group, and obtaining from these invariants a unique signature curve, a sort of identification card. The signature is independent of the position from which the object is seen, which is an advantage in comparison with other possible methods. The original contour can be extracted with image processing methods from a digital image. The signature curve can then be compared to a library of predefined signature curves. The algorithms are investigated under idealized conditions, and then perform well. A modified version of this method would be used in a lot of different military computer vision problems.