Thu, Aug 05, 2021:On Demand
Background/Question/Methods
Native sagebrush steppe habitats are some of the most diminished ecosystems in the western United States due to agricultural conversion, fragmentation, degradation and conversion to invasive annual grasses. Accurate and comprehensive monitoring, where information can be collected across multiple scales and be spatially referenced on a continual basis, is needed to develop better models. By using remote sensing techniques like unoccupied aerial vehicles (UAV), researchers can collect equivalent data in an afternoon that would take days or weeks to accumulate on the ground. For sagebrush restoration efforts, models can be created to track survival of individual plants through time, detect new seedlings, and measure plant growth. In this study, we used Agisoft Metashape software to process images taken in 2020 at the Soda Common Garden in Boise Idaho, USA. Our objective was to test the impacts of processing parameters to determine if changing the number of key points or tie points would improve the quality of the structure from motion (SfM) dense point cloud.
Results/Conclusions When we qualitatively evaluated the differences between our fall 2020 point cloud and one collected in early summer 2019, we found that the 2020 point cloud models resulted in a lower quality point cloud. The software had difficulty detecting the sagebrush plants in the 2020 images due to longer shadows and the lack of contrast between the sagebrush plants and the ground. We found that increasing the tie point limit had no effect on the number of points (R2=0.01), but increasing the key point limit decreased the number of points in the dense cloud (R2=0.22). Based on these results, our next steps will be to test the parameter that removes erroneous points, which could be affecting the final dense point cloud. We will also continue testing other parameters in Agisoft Metashape to improve the final products generated from UAV flights in both early summer and fall. Our results reflect that the products were comparatively unaffected by the change in processing parameters. This is useful for researchers seeking to compare structure across collections of processed imagery.
Results/Conclusions When we qualitatively evaluated the differences between our fall 2020 point cloud and one collected in early summer 2019, we found that the 2020 point cloud models resulted in a lower quality point cloud. The software had difficulty detecting the sagebrush plants in the 2020 images due to longer shadows and the lack of contrast between the sagebrush plants and the ground. We found that increasing the tie point limit had no effect on the number of points (R2=0.01), but increasing the key point limit decreased the number of points in the dense cloud (R2=0.22). Based on these results, our next steps will be to test the parameter that removes erroneous points, which could be affecting the final dense point cloud. We will also continue testing other parameters in Agisoft Metashape to improve the final products generated from UAV flights in both early summer and fall. Our results reflect that the products were comparatively unaffected by the change in processing parameters. This is useful for researchers seeking to compare structure across collections of processed imagery.