Navigation Sensor Accuracy

The paper draws attention to the emerging military capabilities of laser radar by describing state-oftechnology sensor processing results for applications involving helicopter obstacle warning, automatic target identification at far range and automatic target detection and classification at close range. In all applications sensor processing requires accurate motion compensation using navigation data, the accuracy of the navigation sensor needed depending on the application and the laser radar being used. The paper analyses how navigation and sensor errors affect sensor processing results, specifies maximal velocity and angular velocity errors of the navigation system and examines to what extent an INS with low-cost micromachined gyros and accelerometers can be expected to meet these specifications. 1.0 INTRODUCTION Eyesafe solid state lasers with pulse powers of several kilowatts and pulse rates of 10’s of kilohertz have only been available since the early 90’s; their military potential has only begun to be explored. Used in scanning or staring-array laser range finders (also called laser radar or ladar for short) with high resolution and high range accuracy, they provide manned and unmanned military vehicles with previously unattained scene understanding capabilities. Applications include helicopter obstacle warning and automatic obstacle avoidance, automatic target classification and identification at high range for gunner support applications, target detection and selection for submunition dispensing UAVs and target identification and aim point selection for guided missiles. Algorithms for each of the above applications have been developed, tested and evaluated at the FOM, using real sensor data and realistic military scenarios. The main problem to be solved in all the above applications is the real-time automatic recognition of man-made objects in natural, outdoor scenes. If only passive or non-range sensor data is available (infrared, visual, SAR, etc.), this problem is far from being solved. However, if range imagery is available, fast, robust solutions can be obtained as follows: 1) Object/ground segmentation: This step separates arbitrary objects from the terrain surface on which they are located. It requires no object models. 2) Object classification: This step performs a shape classification of objects extracted in step 1. It calculates the class probability, position and orientation of objects given a catalogue of object class features. 3) Object identification: The final step matches a geometric surface model of the expected objects with the object range data from one or more range frames. RTO-MP-0 Paper presented at the RTO SET Symposium on “Emerging Military Capabilities Enabled by Advances in Navigation Sensors”, held in Istanbul, Turkey, 14-16 October 2002, and published in RTO-MP-093.