Application of sensor data fusion techniques to the light armoured vehicle reconnaissance (LAV Recce).

摘要

The ability to employ dissimilar sensors to collect information concerning the behavior of a wide variety of potential threats is critical to the survivability of high priority military assets. To this end, accurately locating and identifying enemy targets forms the basis of an effective response. Consequently, research into surveillance and tracking systems that best utilize available resources is a high priority.; In this thesis, an efficient, robust and effective target tracking architecture is developed that combines active information from a Moving Target Indicator (MTI) radar and passive information from a Forward Looking Infrared (FLIR) imager. The final design, a decentralized architecture, is selected out of a field of three candidates on the basis of performance in Monte Carlo simulation. The other candidates are examples of sequential and passive decentralized fusion. The design is intended as a potential enhancement to the surveillance suite aboard the Canadian Forces' newly-acquired Light Armored Vehicle Reconnaissance (LAV Recce).; The design effort includes (i) evaluating various data fusion architectures; (ii) selecting appropriate tracking filters; (iii) choosing maneuver compensation and track initiation schemes; and, (iv) developing a method to extract a targets angular position from analog infrared images. The scope of this thesis is limited to a simulation of the single target case; however, an extension to the multiple target scenario is also discussed in terms of selecting an appropriate data association algorithm.; No tracking system design is complete without a thorough field test. To judge the effectiveness of the proposed architecture, the LAV Recce was tested in a realistic surveillance exercise that produced a complete set of pre-processed data with which to validate this and future prototypes. In the trial, the radar and FUR sensors were assigned a common line-of-sight and both covered 200 milliradian horizontal fields of view. The target, a moving vehicle, was observed from approximately one kilometer away. An example of the infrared data is presented, and the image processing techniques adopted in this thesis are demonstrated successfully. The complete test results, which include a test of the proposed tracking architecture with both radar and FUR data, will follow in a departmental technical report by the author at a later date.