Camera Assisted Roadside Monitoring for Invasive Alien Plant Species Using Deep Learning

Abstract

Invasive alien plant species (IAPS) pose a threat to biodiversity as they propagate and outcompete natural vegetation. In this study, a system for monitoring IAPS on the roadside is presented. The system consists of a camera that acquires images at high speed mounted on a vehicle that follows the traffic. Images of seven IAPS (Cytisus scoparius, Heracleum, Lupinus polyphyllus, Pastinaca sativa, Reynoutria, Rosa rugosa, and Solidago) were collected on Danish motorways. Three deep convolutional neural networks for classification (ResNet50V2 and MobileNetV2) and object detection (YOLOv3) were trained and evaluated at different image sizes. The results showed that the performance of the networks varied with the input image size and also the size of the IAPS in the images. Binary classification of IAPS vs. non-IAPS showed an increased performance, compared to the classification of individual IAPS. This study shows that automatic detection and mapping of invasive plants along the roadside is possible at high speeds.

Deep learning‐based training data augmentation combined with post‐classification improves the classification accuracy for dominant and scattered invasive forest tree species

“…This option supports a more efficient identification of this invasive species and may facilitate timely intervention to avoid its spread in areas of interest. With the support of aerial imaging, tree composition can be mapped with high accuracy and monitored easily, quickly and more accurately in order to detect the changes of invasive tree species both spatially and temporally (da Silva et al., 2023; Dyrmann et al., 2021). This also provides a valuable practical option in the mapping of forest trees where there are stands mixed with invasive tree species.…”

“…Gathering stand‐level data, therefore, is important for implementing sustainable forest management practices in order to preserve biodiversity and multifunctional ecosystem services (Vo et al., 2013). Knowledge on the species composition is also important in quantifying timber volume and above ground biomass (Brahma et al., 2021; Nagy et al., 2020; Vo et al., 2013; Vorster et al., 2020), in nature conservation and landscape protection (Blackman, 2013; Zhang & Xie, 2012), and in tracking the spread of invasive tree species (Asner et al., 2008; Wilfong et al., 2009; Baron et al., 2018; Dyderski & Jagodziński, 2020; Dyderski & Pawlik, 2020; Dyrmann et al., 2021; Lake et al., 2022; da Silva et al., 2023).…”

The use of photos to investigate ecological change

“…Similarly, Pardo‐Primoy and Fagúndez (2019) used GSV to assess the distribution and recent spread of an invasive grass species in industrial sites in Galicia (Spain). Dyrmann et al (2021) also showed that mapping of invasive plants along roadsides is possible based on images taken from a driving vehicle, although they did not use GSV, but used their own equipment to obtain such images. An important limitation of the GSV system is that images are not always taken on the same date, or even the same season.…”

“…This reduces the probability of plants being detected (e.g. for species that are less distinct when not flowering) and makes GSV less suitable for, for example, phenology studies (Dyrmann et al, 2021; Kotowska et al, 2021).…”

“…Rousselet et al, 2013) and plants (e.g. Dyrmann et al, 2021; Kotowska et al, 2021). Photos are a type of legacy remote sensing data that allow scientists to test for global environmental change effects (sensu Vellend et al, 2013), just as herbarium specimens, resurveyed vegetation plots and land survey records (De Frenne, 2015; reviewed by Vellend et al, 2013; Willis et al, 2017).…”