Rodent hole detection in a typical steppe ecosystem using UAS and deep learning

Abstract

IntroductionRodent outbreak is the main biological disaster in grassland ecosystems. Traditional rodent damage monitoring approaches mainly depend on costly field surveys, e.g., rodent trapping or hole counting. Integrating an unmanned aircraft system (UAS) image acquisition platform and deep learning (DL) provides a great opportunity to realize efficient large-scale rodent damage monitoring and early-stage diagnosis. As the major rodent species in Inner Mongolia, Brandt’s voles (BV) (Lasiopodomys brandtii) have markedly small holes, which are difficult to identify regarding various seasonal noises in this typical steppe ecosystem.MethodsIn this study, we proposed a novel UAS-DL-based framework for BV hole detection in two representative seasons. We also established the first bi-seasonal UAS image datasets for rodent hole detection. Three two-stage (Faster R-CNN, R-FCN, and Cascade R-CNN) and three one-stage (SSD, RetinaNet, and YOLOv4) object detection DL models were investigated from three perspectives: accuracy, running speed, and generalizability.ResultsExperimental results revealed that: 1) Faster R-CNN and YOLOv4 are the most accurate models; 2) SSD and YOLOv4 are the fastest; 3) Faster R-CNN and YOLOv4 have the most consistent performance across two different seasons.DiscussionThe integration of UAS and DL techniques was demonstrated to utilize automatic, accurate, and efficient BV hole detection in a typical steppe ecosystem. The proposed method has a great potential for large-scale multi-seasonal rodent damage monitoring.

Brandt’s vole hole detection and counting method based on deep learning and unmanned aircraft system

“…Despite the varying required image resolution among different rodent species ( Cui et al., 2020 ; Ezzy et al., 2021 ; Zhou et al., 2021 ; Du et al., 2022 ), there has been no research on the method of determining the UAV flight altitude or image resolution. This study uses manual visual interpretation as a benchmark and conventional classification evaluation indexes for resolution evaluation.…”

“…Monitoring the population size of this species is essential for its scientific management. However, traditional methods are labor–intensive, such as using visual observation or traps to count voles, or using plugging and opening to count active holes, all of which are time–consuming due to the limitation of quadrat to small scales with 0.25 ~ 1 hectare ( Du et al., 2022 ). Therefore, an efficient and accurate technology for pest rodent monitoring is urgent.…”

“…This combination has been used to detect wildlife ( Kellenberger et al., 2018 ; Peng et al., 2020 ) and livestock population surveys ( Soares et al., 2021 ; Wang et al., 2023b ). Additionally, it can be used to survey rodents in grassland such as yellow stepped vole ( Eolagurus luteus ; Sun et al., 2019 ), great gerbil ( Rhombomys opimus ; Cui et al., 2020 ), plateau pika ( Ochotona curzoniae ; Zhou et al., 2021 ), the Levant vole ( Microtus guentheri ; Ezzy et al., 2021 ), Brandt’s vole ( Du et al., 2022 ). By identifying the rodent holes, researchers can estimate the density of rodents.…”

Detection and classification of Brandt’s vole burrow clusters utilizing GF-2 satellite imagery and faster R-CNN model

“…While developing countries have established relatively comprehensive theoretical knowledge regarding rodent agricultural pest management, rodents still pose a significant problem for global food security (Meerburg et al, 2009).From an economic standpoint, when we contrast the damage caused by rodents worldwide prior to harvest and following harvest, the annual toll stands at 10-15% (Meerburg et al, 2009;Belmain et al, 2015). Rodent population's outbreaks and spread not only threaten the lives and property of local populations (Addink et al, 2010;Sage et al, 2017;Ocampo-Chavira et al, 2020) but also cause certain degrees of damage to steppe ecosystems (Du et al, 2022). Brandt's vole (Lasiopodomys brandtii, BV) is widely distributed in the Russian Federation, central and eastern regions of Mongolia, and northeastern areas of China (Avirmed et al, 2016).…”

“…For example, the detection of Levant vole (Microtus guentheri) burrows in alfalfa (Medicago sativa) fields was achieved with high reliability using You Only Look Once (YOLO)v3, showcasing its accuracy (Ezzy et al, 2021). Another study conducted in the steppe of Xilingol League, Inner Mongolia, China, utilized UAV imagery and DL methods to extract BV burrows, and Faster R-CNN and YOLOv4 yielded the most accurate results (Du et al, 2022). Despite the advantages of UAV remote sensing, such as its flexibility and high spatial resolution (Lyu et al, 2022), it still falls short of meeting the requirements of large-scale pest detection or is too costly for extensive pest monitoring.…”