Dickmanns 1987b : A method for interpreting high frequency image sequences is presented that confines image data processing to the last image of the sequence and yet - by using smoothing integration-operations -  allows to determine velocity components in space explicitly. This is achieved by simultaneously exploiting 3D object and motion models together with the laws of perspective projection. These integral models of the process observed in the world and of the (very flexible) measurement process using vision are utilized in the sense of model-based feedback control theory (Kalman filter, observer) to estimate the state variables in space and time directly. The concept has been tested in three application areas: A planar docking maneuver between two 3D-objects, autonomous road vehicle guidance, and autonomous aircraft landing.