MAPPING INVASIVE PLANTS USING A HELMET BASED VIDEO SYSTEM

BY Corey V. Ransom*, Heather E. Olsen; Utah State University, Logan, UT. Presented at the Western Society of Weed Science conference in Portland, Oregon. March 9-12, 2015.

ABSTRACT

Conducting invasive plant inventories is a critical component of an integrated approach to invasive plant management.  Inventory data often provides the information necessary to evaluate the extent of weed invasion allowing land managers to prioritize management efforts.  Invasive plant inventory data is expensive to collect.  Aerial approaches to invasive plant mapping can be more efficient for highly visible species, but are limited to plants visible from the air.  Recent advances in video technology allow collection of high definition video with compact, relatively inexpensive cameras.  Research was conducted to compare two ground-based methods of weed mapping for infestation estimate accuracy and time required to conduct the inventories.  The first inventory method involved mappers on foot inputting infestation data into a handheld GPS.  The second approach utilized a person riding a mountain bike wearing two helmet mounted video cameras (GoPro Hero2, GoPro Inc.) and later using the video to generate inventory polygons on a desktop computer in the office.  A GPS or smart phone was used to collect tracklog data to accompany the video footage.  The helmet mounted cameras were placed facing forward and focused approximately 70 degrees apart to give wide perspective to the right and left of the rider.  Five trails were mapped using both approaches in mid May 2014 while dyer’s woad was in full bloom.  Dyer’s woad was selected as the target since its bright yellow flowers are easily distinguishable from surrounding green vegetation.  The videos from both cameras were blended into a single video (Premiere CS6, Adobe) and then imported along with the corresponding tracklog into a software (VIRB Edit, Garmin Ltd.) that allows the video and the tracklog to play simultaneously.  Using a second computer monitor, infestation shapes were drawn onto a GIS map (ArcPad 10, ESRI) as they were observed in the video and the location was identified on the corresponding map.  The time spent mapping on the computer was recorded and was added to the time required to ride each trail section to determine total time required for mapping.  Time required to stitch videos together or to sync tracklogs with the video was not included in calculations as the process could likely be automated in the future.  Comparison of the two mapping methods included total time, total number of points, polygons, and lines, as well as total infested acres.  Time efficiency as well as total infested acreage estimated varied widely between the two techniques.  Time savings using the helmet mounted video approach ranged from 17 to 25% for a very steep trail and a small parcel to 60 to 73% for trails that were relatively flat to mostly downhill.  The video mapping approach had lower estimates (70 to 83%) than the on-foot approach for 2 of the trails, but infestation estimate was almost 35% higher for another trail.  Unfortunately there was no way to determine which method is more accurate since there was no actual infestation measurement for comparison.  Future studies will need to include such a comparison.  In some instances, both mapping methods identified small patches or single plants in the exact same location and in most cases, while infestation polygons differed in size, the location of plants and patches were similar between the methods.  Many discrepancies were due to the method each mapper selected to represent any given infestation (individual patches vs. large polygons or line features).  The video approach did allow fairly clear differentiation between dyer’s woad and other yellow-flowered species which were in bloom.  Newer video cameras offers even higher resolutions and video frame capture rates that could increase the ease of identifying specific species.  Approaches to stabilize the camera during data collection are currently being investigated and have potential to improve video clarity.  This research shows that helmet mounted video cameras can be used to map easily detected weed patches, with potential time savings compared to mapping on foot.

Read more from USU>