UAV-BASED MULTI-SENSOR DATA FUSION FOR TIME-CRITICAL DISASTER RESPONSE Oktay Baysal, Guoqing Zhou * Batten College of Engineering and Technology, Old Dominion University Norfolk, Virginia, 23529 Tel: (757) 683-3619; E-mail: gzhou@odu.edu KEY WORDS: UAV, Data Fusion, Disaster Response, Video Flow, Real-time ABSTRACT: We present a mathematical model for a UAV-based, multi-sensor data integration. As a background, we first discuss the design and the implementation of a low-cost civilian UAV system, including its field flight validation, system calibration, and mapping accuracy evaluation. Then, this photogrammetry-based mathematical model is developed. The field flight tests demonstrate that the designed low-cost UAV is capable of collecting clear and high-resolution video. The UAV can be controlled and navigated remotely and video stream and navigation data strings, including position and attitude, can be downlinked to the ground control station in real-time. The present multi-sensor, mathematical model reveal that the boresight matrix in a low-cost UAV system does not remain constant. This contradicts the practice in traditional airborne mapping systems where the boresight matrix is assumed to be a constant over an entire mission. Thus, to achieve a high-accurate mapping, EOPs of each video frame in a low-cost UAV mapping system have to be estimated. The present model can achieve a planimetric mapping accuracy using 1-2 pixels when compared with the USGS DOQ orthophoto. * Corresponding author. gzhou@odu.edu 1. INTRODUCTION Forest fires may adversely impact more people in U.S.A than any other natural disaster. Nearly 1,000 structures and over 4 million acres of land are burned by wildfires annually (Kimball, 2003). In addition to the human and financial costs to the society, wildfires cause tremendous physical damages, and have notable environmental impacts. A reduction of wildfires demands large amounts of federal resources, costing up to $1.6 billion per year, along with the lives of ten to twenty firefighters (Kimball, 2003). Therefore, the efforts of improving wildfire surveillance technology for mitigating disasters must continue. Spaceborne-based and remotely sensed imagery has been a major data source for forest fire reconnaissance. Although the revisit cycle of useable satellites can be as good as one to three days, data collection at a revisit cycle of hours or decade minutes is increasingly required due to the requirement of fast- response to disasters (Zhou et al., 2009, 2002). Therefore, Unmanned Aerial Vehicles (UAVs) equipped with thermal infrared imaging technology and data telemetry to collect high- resolution video data, have been employed for forest fire surveillance in recent years (Wegener et al. 2002). The high- resolution video image, on the one hand, brings us clarity and details of the behaviour and characteristics of wildfire, on the other hand, presents new challenges in data processing. For example, how do we generate orthoimage from high-resolution UAV video images in forest fire areas at (near) real time? The orthoimage is critical geospatial data for wildfire experts, because: (1) it serves as a base map on which wildfire experts can add, register and compile other geospatial data; (2) it can be easily displayed as mosaiced products, quickly exploited to derive high-precision 3-D geolocations of objects within each video frame; (3) geometric measurement of wildfires (e.g., wildfire scopes, disaster areas), tree parameters (e.g., crown diameters, canopy closure) from orthoimages are more reliable than those from original perspective photographs since orthophotos theoretically are free of perspective distortion. The orthoimage will therefore be able to provide firefighters, wildfire analysts and decision-makers with greater situational awareness for wildfire behaviors, characteristics, and effects. The present paper reports the results of UAV-based multi- sensor data fusion for wild fire reconnaissance. 2. DESIGN AND IMPLEMENTATION OF UAV SYSTEM In contrast to tactical or strategic ones used for military missions, these UAVs must carry highly reliable, accurate, but very expensive, and complicated instruments (Henri, 2004). The civilian UAV users have a strong demand for a low-cost, moderately functional, small airborne platform, varying in size, computerization and levels of autonomy (Moore et al., 2003). Therefore, how to develop such an economical UAV system, including hardware and software, for small private sectors and non-military government agencies to meet geospatial needs focusing on small areas of interest and that can be used for a broader array of mapping purposes, is key in designing and implementing our UAV platform. For this reason, an end-to-end development for low-cost civilian UAV system including hardware and software has been implemented. The present paper reports only the research results pertaining to design and implementation of a small, lower-cost UAV system, and the multi-sensor data fusion. The real-time processing of the UAV-based video data and its evaluation for civilian applications in fast-response to time- critical disaster environments, such as wildfire surveillance, 1