Weed control is one of the biggest challenges in organic farms or nature reserve areas where mass spraying is prohibited. Recent advancements in remote sensing and airborne technologies provide a fast and efficient means to support environmental monitoring and management, allowing early detection of invasive species. However, in order to perform weed classification, current studies mostly relied on object-based image analysis (OBIA) and proprietary software which required substantial human inputs. This paper proposes an open-source workflow for automated weed mapping using a commercially available unmanned aerial vehicle (UAV). The UAV was flown at a low altitude between 10 m and 20 m, and collected truecolour RGB imagery over a weed-infested nature reserve. The aim of this study is to develop a repeatable and robust system for early weed detection, with minimum human intervention, for classification of Rumex obtusifolius (R. obtusifolius). Preliminary results of the proposed workflow achieved an overall accuracy of 92.1% with an F1 score of 78.7%. The approach also demonstrated the capability to map R. obtusifolius in datasets collected at various flight altitudes, camera settings and light conditions. This shows the potential to perform semi-or fully automated early weed detection system in grasslands using UAV-imagery.
This paper presents the results of a study undertaken to classify lowland native grassland communities in the Tasmanian Midlands region. Data was collected using the 20 band hyperspectral snapshot PhotonFocus sensor mounted on an unmanned aerial vehicle. The spectral range of the sensor is 600 to 875 nm. Four vegetation classes were identified for analysis including Themeda triandra grassland, Wilsonia rotundifolia, Danthonia/Poa grassland, and Acacia dealbata. In addition to the hyperspectral UAS dataset, a Digital Surface Model (DSM) was derived using a structure-from-motion (SfM). Classification was undertaken using an object-based Random Forest (RF) classification model. Variable importance measures from the training model indicated that the DSM was the most significant variable. Key spectral variables included bands two (620.9 nm), four (651.1 nm), and 11 (763.2 nm) from the hyperspectral UAS imagery. Classification validation was performed using both the reference segments and the two transects. For the reference object validation, mean accuracies were between 70% and 72%. Classification accuracies based on the validation transects achieved a maximum overall classification accuracy of 93.
This paper presents a case study for the analysis of endangered lowland native grassland communities in the Tasmanian Midlands region using field spectroscopy and spectral convolution techniques. The aim of the study was to determine whether there was significant improvement in classification accuracy for lowland native grasslands and other vegetation communities based on hyperspectral resolution datasets over multispectral equivalents. A spectral dataset was collected using an ASD Handheld-2 spectroradiometer at Tunbridge Township Lagoon. The study then employed a k-fold cross-validation approach for repeated classification of a full hyperspectral dataset, a reduced hyperspectral dataset, and two convoluted multispectral datasets. Classification was performed on each of the four datasets a total of 30 times, based on two different class configurations. The classes analysed were Themeda triandra grassland, Danthonia/Poa grassland, Wilsonia rotundifolia/Selliera radicans, saltpan, and a simplified C 3 vegetation class. The results of the classifications were then tested for statistically significant differences using ANOVA and Tukey's post-hoc comparisons. The results of the study indicated that hyperspectral resolution provides small but statistically significant increases in classification accuracy for Themeda and Danthonia grasslands. For other classes, differences in classification accuracy for all datasets were not statistically significant. The results obtained here indicate that there is some potential for enhanced detection of major lowland native grassland community types using hyperspectral resolution datasets, and that future analysis should prioritise good performance in these classes over others. This study presents a method for identification of optimal spectral resolution across multiple datasets, and constitutes an important case study for lowland native grassland mapping in Tasmania.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.