The Effect of Positioning Error on the Repeatability of Small-Footprint Waveform Lidar Signals
Repeated observations in remote sensing,e.g.,change detection,constitute an important technique for the evaluation of ecological system change over time,such as deforestation,land use change,invasive species mapping,etc.However,though change detection protocols are well established for multi-temporal spectral data and discrete lidar data,protocols have not yet been established for performing temporal analysis of small-footprint waveform lidar data.This study evaluated the feasibility of multi-temporal waveform-to-waveform signal comparison for change detection in a broadleaf forest.Specifically,we analyzed the effect of system positioning errors (e.g.,GPS/IMU) on generating repeatable backscattered waveform signals,given a lack of any structural change,such as branch movement due to wind.In order to guarantee that the scene remained constant at different levels of positioning error,we used the digital imaging and remote sensing image generation (DIRSIG) model to simulate lidar waveforms at known locations within a realistic virtual forest scene.The forest scene was based on a forest inventory plot within Harvard Forest,MA,USA.Two experiments were performed for this study: (i) random position (x,y) jitter or pointing (roll,pitch,and yaw) were added to the sensor to simulate possible uncertainties in position knowledge and (ii) a simulated profiling lidar was used with a large overlap in adjacent footprints to compute tree height.The first experiment was designed to evaluate the impact of GPS/IMU errors on the backscattered waveform signal,while the second was used to describe the evolution of the error over distance in terms of a simple waveform product,namely height.Results showed that even small shifts (<0.10 m) in the x-y position of the lidar system,or 0.005° in angular x-,y-,and z-axis rotations were likely to result in dramatically different waveform lidar signals.The passive terms were more similar than the lidar signals over all uncertainties simulated.In addition,there is a significant falloff in the ability to correlate true height from lidar-derived height at only a few centimeters of positional uncertainty.We conclude that the structural heterogeneity in broadleaf forests at the small-footprint scale (0.5m) is too large to allow time-series comparisons at the waveform-to-waveform level given realistic positioning errors.We suggest that small-footprint waveform lidar time-series studies focus on comparison of higher-level products or even the “object- or tree-level,as opposed to the waveforms themselves.
Paul Romanczyk Jan van Aardt Dave Kelbe Kerry Cawse-Nicholson Tanya Ramond
Chester F. Carlson Center for Imaging Science, Rochester Institute of Technology, Rochester, NY, USA National Ecological Observatory Network, Boulder, CO, USA
国际会议
北京
英文
202-209
2013-10-09（万方平台首次上网日期，不代表论文的发表时间）