Unmanned Aerial Vehicle Data Acquisition for Damage Assessment in Hurricane Events Stuart M. Adams a Carol J. Friedland b and Marc L. Levitan c ABSTRACT This paper examines techniques for data collection using an unmanned aerial vehicle to capture building damage resulting from hurricane events. An investigation into an improved procedure and proper parameterization associated with unmanned aerial data collection is explored. This paper aims to develop a methodology to accurately and efficiently capture aerial imagery depicting wind and flood/storm surge induced building damage, an important initiative in determining infrastructure response to hurricane events. Remote damage assessment after hurricane events is often performed using optical imagery, acquired through satellite, aerial, and ground based platforms. These images are then analyzed, often manually, using visual and photogrammetric principals. For analysis, each source has its advantages such as breadth for satellite imagery, resolution for aerial imagery, and obliqueness for ground based imagery; however, limitations often exist with these mediums. Satellite and aerial imagery can lack sufficient obliqueness for use in façade investigations, and traditional aerial imagery can be costly. Additionally, ground imagery can be discontinuous and onerous to collect, thus inhibiting rapid assessments of damage. The aforementioned limitations create difficulties when assessing damage for extreme events with multiple damage mechanisms, such as combined wind and flood/storm surge found in hurricanes. Assessing flood/storm surge damage from a nadir perspective presents significant challenges as flooding/storm surge damage beneath an intact roof can be unaccounted for when image sources lack obliqueness. Furthermore, images with high degrees of obliqueness can hinder full roof damage inspections. The need for a hybrid image collection system to assess this type of damage is evident and can be achieved using unmanned aerial vehicles. Unmanned aerial vehicles have undergone significant advances in equipment capabilities and now have the capacity to acquire high resolution imagery from many angles in a cost effective, efficient manner. This suggests that the use of unmanned aerial vehicles to collect post disaster imagery should be explored. In addition to requirements for remote sensing damage assessment, an understanding of the abilities of unmanned aerial vehicles is needed to determine proper techniques for unmanned aerial vehicle data collection of damage resulting from hurricane events. This paper addresses how an unmanned aerial vehicle s equipment setup as well as flight conditions can be optimized a Department of Civil and Environmental Engineering, Louisiana State University, 3418 Patrick F. Taylor Hall, Baton Rouge, LA 70803, phone: 1+601-278-2459, fax: 1+225-578-4945,sadam15@lsu.edu. b Department of Construction Management & Industrial Engineering, Louisiana State University, 3128 Patrick F. Taylor Hall, Baton Rouge, LA 70803, phone: 1+225-578-1155, fax: 1+225-578-5109 friedland@lsu.edu. c Department of Civil & Environmental Engineering, Louisiana State University, 3214A Patrick F. Taylor Hall, Baton Rouge, LA 70803, phone: 1+225-578-4445, fax: 1+225-578-4945, levitan@lsu.edu.
in order to acquire useful damage data for hurricane events. This work provides a foundation for data collection methodologies using unmanned aerial vehicles that will ultimately be integrated with damage assessment techniques for this type of aerial imagery. INTRODUCTION The concept of using of remote-sensing applications to acquire post disaster infrastructure response imagery has been presented as appropriate for many types of data needs; however, the methodologies used to capture the images have different advantages and limitations. This paper explores what current methods are used to collect post hurricane event imagery and why advances in unmanned aerial vehicle (UAV) capabilities, coupled with a decrease in its costs provide a strong basis for an increased use of UAVs for post hurricane event reconnaissance needs. Once the appropriate imagery is acquired, visual and photogrammetric principals can used to analyze the data. THE HURRICANE HAZARD Hurricane events often are comprised of multiple hazardous conditions. These conditions include high winds, storm surge, and flooding. Roof damage is a common indicator of high winds and can be seen using traditional aerial imagery; however, damage beneath an intact roof caused by flooding or storm surge is often ignored if the images are taken at nadir. UAVs have the ability to take pictures at an angle oblique enough to show both roof and façade damage while keeping costs low and efficiency high. PHOTOGRAMMETRY Photogrammetry is a method used to extract information from imagery. Metric photogrammetry, the science of determining object characteristics from rectified stereo-photographs, will be used when analyzing post hurricane imagery. Distances, angles, areas, volumes, elevations, object sizes, and object shape within overlapping images are some of the many characteristics that can be determined using photogrammetry. In order to perform photogrammetry, multiple photographs of an object must be taken at different locations (preferably at an angle >30 Degrees) using a calibrated camera. User inputted and/or computer calculated control points present in multiple photographs are established in order to determine the 3D coordinate system for the combined images. Once the 3D coordinate system has been established and the scale and orientation of each photograph are determined, a 3D point cloud can be created using commercial software that matches a point in one picture with that of another picture. Points within the point cloud have a calculated position in the 3D field and the accuracy of these points is based on the accuracy of the camera calibration and control points. Using the point cloud, object characteristics such as length and elevation can be determined. A point mesh or textured view of the photographs can be created by overlaying the images on the point cloud. This technique is used to create topographic maps and calculate volume of areas. Volume, areas and length of debris lines are of particular interest in the proposed post hurricane analyzes.
Figure 1 - Stereo-Photogrammetry Diagram (Photomodeler Scanner User Manual, Eos Systems Inc.) CAMERA CALIBRATION Camera Calibration is important in order to determine the characteristics of the images taken by the device. Knowing the focal length, principal distance, radial distortion, decentering distortion, and lens distortion enables users to rectify and orient images properly. Camera Calibration can be done before or after image collection using manual and automated techniques. All cameras placed on UAVs should be properly calibrated. CURRENT OPTICAL IMAGERY COLLECTION METHODS Optical imagery is collected using satellite, aerial and ground based platforms. As illustrated in Figure 1, each methodology of acquiring imagery has certain limitations that can hinder post disaster infrastructure damage assessments.
Concern Satellite Aerial Ground Pre-Storm images Spatial resolution Availability typically superior to aerial and ground Typically more limited than satellite, greater than ground Limited Low Medium High Planning for image collection Minimal planning necessary for consumer Detailed planning necessary Very detailed planning needed Post-storm image availability Access to desired site Possibility for stereo images (overlapping scenes) Collection Areas Orbits allow access to all areas within 3 days Not subject to damage-area restrictions; Limited by clouds, haze, smoke, daylight, and orbits; Limited by revisit times Not presently available for most scenes Much larger areas obtained in a single image, due to higher altitude Subject to availability of equipment and personnel. Proximity of target site to usable airport may influence cost. May be subject to post disaster flight restrictions; Sometimes able to fly beneath cloud cover that prohibits satellite data collection; Flexible scheduling, not subject to revisit times Standard procedure More images needed to cover same area as satellite Subject to availability of equipment and personnel. May be subject to road closures; Limited to readily accessible areas User controlled Can be discontinues; Many images needed to cover same area as aerial Figure 2 Comparison of Satellite, Aerial, and Ground Platforms as Sources of Remote-Sensing Data (Adapted from J.A. Womble) UNMANNED AERIAL VEHICLES Unmanned Aerial Vehicles (UAVs) have become increasingly popular as a means to collect aerial photography. The ability to gather high resolution imagery from UAVs in a low cost manner can now be realized do to advances in consumer digital cameras, remote control devices and global positioning systems (GPS). Proper collection techniques can facilitate the use of photogrammetry on the imagery and thus enabling accurate characterization of objects within stereo photographs. Utilizing UAV-based imagery for storm damage collection would help solve with many of the limitations associated with ground-based damage assessment. UAVs have the ability to cover more area than traditional ground-base assessment teams. Further, images can be to quickly gather and sent remotely to personnel away from the damage to be processes. This aids with safety risks and can significantly lower travel costs.
UAV MODELS The current UAV available for use by Louisiana State University researchers for storm damage image collection is a remote controlled, Styrofoam-based, model airplane equipped with a single digital camera mounted to the side of the plane s body. The plane can fly to elevations above 1,500 feet and can be remotely controlled from over a mile away. A pivot tube is mounted to the right wing in order to determine the flight speed. On the left wing, a forward-facing camera is mounted to help steer the plane when it is out of the user s sight. Screens showing the in-flight and image capturing camera views are available to assist in flight and image quality. In addition to the current UAV plane, an easier to operate Draganflyer E4 Helicopter will be purchased to explore the use of an umanned aerial vehicle image collection methodology. This UAV will be equipped with a high resolution digital camera for use in acquiring post hurricane event images. An open communication API will allow access to telemetry, flight control, altitude, and roll, pitch, and yaw angles. Preprogrammed flight plans can be achieved using the UAV s GPS positioning capabilities. Because this UAV is based on a helicopter platform, the possible take-off and landing locations are significantly increased. Boat based launchings and landings are now a distinct possibility, thus allowing for marsh and swampland deployments. Figure 3 - Draganflyer E4 Helicopter Tech Specs (Draganfly Innovations Inc., 2009)
Figure 4 - Dragan View Software Screenshots (Draganfly Innovations Inc., 2009) UAV IMAGE ACQUISITION PARAMETERS AND ASSESSMENT Determining the best elevation, flight path, optical axis angle, flight speed, and picture overlap will aid in quicker image processing and faster storm damage assessment. Limitations arising from shadows and trees must be accounted for. An analysis of image processing results for differing acquisition parameters will be explored and on the basis of visual inspection, pixel change, and edge detection. Ground trothing will be employed to validate results. DISCUSSION AND FUTURE WORKS The use of UAV-based imagery to conduct post hurricane damage assessments is a realistic endeavor due to advances in technology capabilities coupled with a decrease in cost. Understanding imagery needs and the principles of photogrammetry will help create the optimal UAV image gathering technique. Refining the data collection techniques will lead to quicker image processing and faster storm damage assessment. Future research involves the use of the Draganflyer E4 for trial image acquisition. This UAV has the ability to record important flight data which will be used to decipher what the proper flight path, elevation, and speed as well as the picture angle and overlapping imagery characteristics. Image collection trials involving differing ground conditions, building densities, and foliage degrees will be conducted to explore which data acquisition parameters work best in for certain terrians.
REFERENCES Adams, B.J., J.A. Womble, S. Ghosh and C. Friedland. Deployment of remote sensing technology for multi-hazard post-katrina damage assessment within a spatially tiered reconnaissance framework. Proceedings, Fourth International Workshop on Remote Sensing for Post Disaster Response. 2006, Cambridge, UK. "Draganflyer X4 Helicopter Technical Specifications - Draganfly Innovations Inc." Draganfly.com Industrial Aerial Video Systems & UAVs. Web. 17 July 2010. <http://www.draganfly.com/uav-helicopter/draganflyer-e4/specifications/>. Friedland, C., B. J. Adams and M. Levitan. Results of Neighborhood Level Analysis of Structural Storm Surge Damage to Residential Structures. Proceedings, Sixth International Workshop on Remote Sensing for Post Disaster Response. 2008, University of Pavia, Italy. Friedland, C., Residential Building Damage from Hurricane Storm Surge: Proposed Methodologies to Describe, Assess and Model Building Damage, Department of Civil and Environmental Engineering at Louisiana State University. Wolf, P and Dewitt, B.; Elements of Photogrammetry with Applications in GIS, The McGraw- Hill Companies, Inc., 3 rd Edition 2000. Womble, J. A., Remote-Sensing Applications to Windstorm Damage Assessment. A Dissertation in Civil Engineering. December 2005, Texas Tech University, Lubbock, Texas USA. Zongjian, L.; UAV for Mapping- Low Altitude Photogrammetric Survey, Chinese Academy of Surveying and Mapping. 16 Beitaiping Road, China 2008 ACKNOWLEDGEMENTS Travel funds for the presenting author have been provided by ImageCat, Inc. Aerial Photography was provided by Research Associate Eddie Weeks of Louisiana State University.