Autonomous Detection of Mouse-Ear Hawkweed Using Drones, Multispectral Imagery and Supervised Machine Learning

Abstract

Hawkweeds (Pilosella spp.) have become a severe and rapidly invading weed in pasture lands and forest meadows of New Zealand. Detection of hawkweed infestations is essential for eradication and resource management at private and government levels. This study explores the potential of machine learning (ML) algorithms for detecting mouse-ear hawkweed (Pilosella officinarum) foliage and flowers from Unmanned Aerial Vehicle (UAV)-acquired multispectral (MS) images at various spatial resolutions. The performances of different ML algorithms, namely eXtreme Gradient Boosting (XGB), Support Vector Machine (SVM), Random Forest (RF), and K-nearest neighbours (KNN), were analysed in their capacity to detect hawkweed foliage and flowers using MS imagery. The imagery was obtained at numerous spatial resolutions from a highly infested study site located in the McKenzie Region of the South Island of New Zealand in January 2021. The spatial resolution of 0.65 cm/pixel (acquired at a flying height of 15 m above ground level) produced the highest overall testing and validation accuracy of 100% using the RF, KNN, and XGB models for detecting hawkweed flowers. In hawkweed foliage detection at the same resolution, the RF and XGB models achieved highest testing accuracy of 97%, while other models (KNN and SVM) achieved an overall model testing accuracy of 96% and 72%, respectively. The XGB model achieved the highest overall validation accuracy of 98%, while the other models (RF, KNN, and SVM) produced validation accuracies of 97%, 97%, and 80%, respectively. This proposed methodology may facilitate non-invasive detection efforts of mouse-ear hawkweed flowers and foliage in other naturalised areas, enabling land managers to optimise the use of UAV remote sensing technologies for better resource allocation.

Monitoring of Antarctica’s Fragile Vegetation Using Drone-Based Remote Sensing, Multispectral Imagery and AI

“…It is known for its high execution speed and good performance. XGBoost was chosen for this project as it is a popular ML framework that is used widely in industry and academia, especially for detecting different vegetation types [ 61 , 62 ]. The XGBoost classifier script was fed the ROI multispectral and mask files described in the XGBoost ROI extraction section.…”

Deep Learning-Based Weed–Crop Recognition for Smart Agricultural Equipment: A Review

“…In contrast to RGB cameras, multispectral cameras have additional spectral bands and are capable of sensing radiation in both the invisible (red-edge and near-infrared) and visible segments of the spectrum, typically spanning four to six bands. The inclusion of a reflectance calibration panel makes multispectral cameras less susceptible to environmental variation [39,40]. A multispectral image is essentially a collection of grayscale images, with each image corresponding to a specific wavelength or band of wavelengths in the electromagnetic spectrum.…”

“…Modern agricultural equipment integrates advanced technologies, such as artificial intelligence, navigation, sensing systems and communication, to increase agricultural productivity and promote smart agriculture [22,122,123]. Among the information, navigation data, image recognition data, etc., re- Hile Narmilan Amarasingam et al [40] studied the potential of machine learning (ML) algorithms for the detection of mouse-ear grass leaves and flowers from multispectral (MS) images acquired by unmanned aerial vehicles (UAVs) at different spatial resolutions and compared different machine learning. The highest machine learning recognition was achieved with 100% accuracy.…”

Advancements of UAV and Deep Learning Technologies for Weed Management in Farmland

“…The above shortcomings will reduce the performance of the algorithm and lead to weed detection errors. The traditional approach to weed detection management tasks from aerial images is based on classical machine learning algorithms such as Support Vector Machine algorithms (SVMs), Random Forest algorithms (RFs), Extreme Gradient Boosting (XGB), and k-means algorithms [17][18][19]. The above technologies manually extract features from images through different feature extraction methods.…”