Detection of alligator weed using an unmanned aerial vehicle

• Published in: Plant protection quarterly Vol. 29; no. 3; pp. 84 - 89
• Main Authors: Daniel Clements, Tony Dugdale, Trevor Hunt, Robert Fitch, Calvin Hung, Salah Sukkarieh, Zhe Xu
• Format: Journal Article
• Published: 01.01.2014
• Subjects: Alternanthera phylloxeroides; Control; Drone aircraft; Growth; Noxious weeds; Weeds; Victoria
• ISSN: 0815-2195

A key impediment to the successful eradication of high priority aquatic weeds (State Prohibited Weeds in Victoria, Australia) is the ability to detect infestations so that control programs can be enacted. Currently, the sole method used to detect State Prohibited Weeds (SPWs) is on-ground human surveillance. Advances in unmanned aerial vehicle (UAV) technology offer an opportunity to detect SPWs using high resolution aerial images of areas known, or suspected, to contain SPWs. This proof of concept field trial used a UAV coupled with a camera to gain aerial imagery of an urban creek and wetlands to detect alligator weed (Alternanthera philoxeroides (Mart.) Griseb.), a SPW that is currently being targeted for eradication from Victoria. The ability of three methods to detect patches of alligator weed was compared: intensive on-ground surveys; visual assessment of images collected by the UAV; and an automated algorithm to scan images for the spectral signature of alligator weed. The automated algorithm was able to detect and delineate patches of alligator weed growing between >2.5 to 4 m2 (area cover metric) along the urban creek, while visual assessment of the images collected with the UAV could detect patches of alligator weed >0.06 m2. Intensive on-ground surveys detected patches >0.002 m2. Using a UAV to collect images provides a potential tool to detect patches of alligator weed at a scale useful for alligator weed eradication programs. Refinement of the algorithm is required before it is useful for detection of alligator weed patches for use in eradication programs, where a high proportion of patches must be detected. At a landscape scale, the algorithm could be used to detect large infestations of alligator weed (or other weed species) from existing high altitude aerial photography (orthophotos).