Semi Autonomous Vehicle Intelligence: Real Time Target Tracking For Vision Guided Autonomous Vehicles

摘要: SEMI AUTONOMOUS VEHICLE INTELLIGENCE: REAL TIME TARGET TRACKING FOR VISION GUIDED VEHICLES Jonathan D. Anderson Department of Electrical and Computer Engineering Master Science Unmanned vehicles (UVs) are seeing more widespread use in military, scientific, civil sectors recent years. These UVs range from unmanned air ground to surface underwater vehicles. Each these different has its own inherent strengths weaknesses, payload freedom movement. Research this field is growing primarily because the National Defense Act 2001 mandating that onethird all military be by 2015. using small UVs, particular, a can go places may too dangerous for humans. Because limitations including power consumption payload, selection light weight low sensors processors becomes critical. Low CMOS cameras real-time vision processing algorithms provide fast reliable information UVs. often require computational limits their traditional general purpose conventional software. The latest developments programmable gate arrays (FPGAs) an alternative hardware software co-design complicated realtime algorithms. By tracking features one frame another, it possible perform many high-level tasks, object following. This thesis describes guidance system FPGA implementation operates tasks real-time. uses following algorithm allows UV follow target. heart rank transform, which transforms image into robust maintains edges found original image. A minimum sum absolute differences used determine best correlation between frames, output update moving Control code move pursuit target such as another vehicle.