Range and Velocity Estimation of Objects at Long Ranges Using Multiocular Video Sequences

Automatic warning systems for protection against airborne threats require a high detection probability at long ranges. Infrared Search and Track (IRST) systems which are designed as monocular and passive systems are able to fulfil this requirement. However, due to clutter and distortions by other objects, like birds or clouds, these systems do not achieve the required low false alarm rate. For a reduction of the false alarm rate the exploitation of features like shape, size, texture and intensity is not reliable, because objects at long ranges appear as points. Range and velocity are more robust features of the point like objects. These features can be obtained by stereo vision from image sequences of multi-ocular systems. In this paper we describe investigations about the accuracy and reliability concerning the three-dimensional position and velocity of the objects. Unavoidable uncertainties in the measurement of the two-dimensional object position in the sensor focal plane lead to large errors in the estimated distance. We present a quantitative analysis of this issue, which results in fundamental restrictions for velocity estimation of objects. These considerations of accuracy and reliability are important for the design of multi-ocular IRST systems. The theoretical analysis is compared with the result of a processed IR stereo image sequence recorded at a measurement campaign with real objects. It is shown that data processing considering the fundamental restrictions lead to robust results for estimation of the spatial position and velocity. This information can be effectively used to reduce the false alarm rate.