GEOG 566






         Advanced spatial statistics and GIScience

April 5, 2017

My Spatial Problem

Filed under: 2017,My Spatial Problem 2017 @ 8:01 pm

This is the blog page for Spring 2017 – My Spatial Problem posts

Print Friendly, PDF & Email


1 Comment

  1.   garmsc — April 5, 2017 @ 1:36 pm    

    My Spatial Problem Blog Post
    Integrating Mobile and UAV-based Lidar Point Clouds to Estimate Forest Inventory
    Cory Garms
    April 2017

    1.) The research question I’m exploring is: “Can point cloud data from combining UAV-based and mobile lidar systems be used to reliably estimate forest inventory of thinned Douglas fir stands?”

    2.) My data consist of lidar point clouds from two different sensors. The first one is part of OSU’s mobile lidar system (MLS), a Toyota Tacoma pickup truck with the laser scanner in the bed that takes scans while the truck moves down the road. The second is a laser scanner that is mounted on the bottom of an unmanned aerial vehicle (UAV). The idea is that the MLS scanner and UAV based lidar will provide different vantages and, when combined in a single point cloud, will provide sufficiently resolute reconstructions of tree geometry to make reliable measurements of parameters like standing timber volume. Typical point cloud data from preliminary scans suggest that resolutions <5cm can be expected from the mobile scanner. The UAV based lidar is currently being mounted and flight tested, so less is known about the resolution of those point clouds. The manufacturer specifications of the sensor suggest that similar resolutions (<5cm) are attainable with good flight planning. Both lidar scans will be conducted on the same day or on consecutive days minimize temporal error.

    3.) I expect that there will be some edge effects that I will need to correct for and am under the assumption that these effects will be systematic. The hope is that the MLS laser will penetrate the stand past the extent of the edge effect and will allow detection of tree geometry within the stand. If this is the case, extrapolation of these geometries onto the rest of the stand should be possible using the points acquired from the UAV. My hypothesis is that the combination of these two lidar sensors will provide better estimates of forest inventory than timber cruises, which represent the current industry standard.

    4.) There will be some significant processing involved in joining the two point clouds and evaluating the accuracy of the registration. Assuming this works, a deliberate step-by-step cloud segmentation approach will be developed. I will need to first extract all the ground points from the cloud. Once the ground is extracted and modeled, I can begin to remove trees that fall into the ‘edge zone’ and then extract individual tree stems and estimate their dimensions. Ideally, I can develop an automated system of doing this extraction and filtering process such as a computer learning algorithm. In the end, I want to extrapolate tree diameter/height relationships onto the entire stand and produce estimates of forest inventory.

    5.) Although there will be many opportunities along the processing workflow to generate tables, figures, and images to help tell the story of this study, my main objective is to produce standing volume estimates for the stand and compare them to timber cruise estimates. Statistical relationships between cruise estimates, lidar estimates, and harvest value could be compared after the fact to determine the accuracy of economic predictions.

    6.) If this method works, it would be significant because foresters can take inventory without ever entering a stand. Furthermore, the MLS is also a powerful tool for monitoring infrastructure, especially roads. In the case of forest roads, they would be measured in great detail while simultaneously measuring the forest. It is my hope that my method will be cheaper, safer, and more accurate than the current status quo, providing an opportunity to revolutionize a vital part of the forestry industry.

    7.) I have a strong foundation in ArcGIS and point cloud manipulation software programs like Leica Cyclone, FUSION, and CloudCompare. In the past 9 months, I have become familiar with R for a variety of applications and plan to use it extensively in my analysis. One of the major issues I expect to deal with is related to the immense size of the datasets I’m working with. The pointclouds can be 50-100GB in the raw form and tend to swell during processing. As a result, I would also like to use this class as an opportunity to think about how I can make more manageable files out of the point clouds by extracting only the information that I’m interested in using and using geometric models to simplify highly detailed 3D representations.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

© 2019 GEOG 566   Powered by WordPress MU    Hosted by blogs.oregonstate.edu