Automated detection of koalas using low-level aerial surveillance and machine learning

Heating decoys to mimic thermal signatures of live animals for drones

Thermal sensors mounted on drones (unoccupied aircraft systems) are popular and effective tools for monitoring cryptic animal species, although few studies have quantified sampling error of animal counts from thermal images. Using decoys is one effective strategy to quantify bias and count accuracy; however, plastic decoys do not mimic thermal signatures of representative species. Our objective was to produce heat signatures in animal decoys to realistically match thermal images of live animals obtained from a drone-based sensor. We tested commercially available methods to heat plastic decoys of three different size classes, including chemical foot warmers, manually heated water, electric socks, pad, or blanket, and mini and small electric space heaters. We used criteria in two categories, 1) external temperature differences from ambient temperatures (ambient difference) and 2) color bins from a palette in thermal images obtained from a drone near the ground and in the air, to determine if heated decoys adequately matched respective live animals in four body regions. Three methods achieved similar thermal signatures to live animals for three to four body regions in external temperatures and predominantly matched the corresponding yellow color bins in thermal drone images from the ground and in the air. Pigeon decoys were best and most consistently heated with three-foot warmers. Goose and deer decoys were best heated by mini and small space heaters, respectively, in their body cavities, with a heated sock in the head of the goose decoy. The materials and equipment for our best heating methods were relatively inexpensive, commercially available items that provide sustained heat and could be adapted to various shapes and sizes for a wide range of avian and mammalian species. Our heating methods could be used in future studies to quantify bias and validate methodologies for drone surveys of animals with thermal sensors.•We determined optimal heating methods for plastic animal decoys with inexpensive and commercially available equipment to mimic thermal signatures of live animals.•Methods could be used to quantify bias and improve thermal surveys of animals with drones in future studies.

Evaluating Thermal Infrared Drone Flight Parameters on Spider Monkey Detection in Tropical Forests

Geoffroy's spider monkeys, an endangered, fast-moving arboreal primate species with a large home range and a high degree of fission-fusion dynamics, are challenging to survey in their natural habitats. Our objective was to evaluate how different flight parameters affect the detectability of spider monkeys in videos recorded by a drone equipped with a thermal infrared camera and examine the level of agreement between coders. We used generalized linear mixed models to evaluate the impact of flight speed (2, 4, 6 m/s), flight height (40, 50 m above ground level), and camera angle (-45°, -90°) on spider monkey counts in a closed-canopy forest in the Yucatan Peninsula, Mexico. Our results indicate that none of the three flight parameters affected the number of detected spider monkeys. Agreement between coders was "substantial" (Fleiss' kappa coefficient = 0.61-0.80) in most cases for high thermal-contrast zones. Our study contributes to the development of standardized flight protocols, which are essential to obtain accurate data on the presence and abundance of wild populations. Based on our results, we recommend performing drone surveys for spider monkeys and other medium-sized arboreal mammals with a small commercial drone at a 4 m/s speed, 15 m above canopy height, and with a -90° camera angle. However, these recommendations may vary depending on the size and noise level produced by the drone model.

Aerial Wildlife Image Repository for animal monitoring with drones in the age of artificial intelligence

Drones (unoccupied aircraft systems) have become effective tools for wildlife monitoring and conservation. Automated animal detection and classification using artificial intelligence (AI) can substantially reduce logistical and financial costs and improve drone surveys. However, the lack of annotated animal imagery for training AI is a critical bottleneck in achieving accurate performance of AI algorithms compared to other fields. To bridge this gap for drone imagery and help advance and standardize automated animal classification, we have created the Aerial Wildlife Image Repository (AWIR), which is a dynamic, interactive database with annotated images captured from drone platforms using visible and thermal cameras. The AWIR provides the first open-access repository for users to upload, annotate, and curate images of animals acquired from drones. The AWIR also provides annotated imagery and benchmark datasets that users can download to train AI algorithms to automatically detect and classify animals, and compare algorithm performance. The AWIR contains 6587 animal objects in 1325 visible and thermal drone images of predominantly large birds and mammals of 13 species in open areas of North America. As contributors increase the taxonomic and geographic diversity of available images, the AWIR will open future avenues for AI research to improve animal surveys using drones for conservation applications. Database URL: https://projectportal.gri.msstate.edu/awir/.

Luminescence-Based Infrared Thermal Sensors: Comprehensive Insights

Recent chronological breakthroughs in materials innovation, their fabrication, and structural designs for disparate applications have paved transformational ways to subversively digitalize infrared (IR) thermal imaging sensors from traditional to smart. The noninvasive IR thermal imaging sensors are at the cutting edge of developments, exploiting the abilities of nanomaterials to acquire arbitrary, targeted, and tunable responses suitable for integration with host materials and devices, intimately disintegrate variegated signals from the target onto depiction without any discomfort, eliminating motional artifacts and collects precise physiological and physiochemical information in natural contexts. Highlighting several typical examples from recent literature, this review article summarizes an accessible, critical, and authoritative summary of an emerging class of advancement in the modalities of nano and micro-scale materials and devices, their fabrication designs and applications in infrared thermal sensors. Introduction is begun covering the importance of IR sensors, followed by a survey on sensing capabilities of various nano and micro structural materials, their design architects, and then culminating an overview of their diverse application swaths. The review concludes with a stimulating frontier debate on the opportunities, difficulties, and future approaches in the vibrant sector of infrared thermal imaging sensors.

Counting manatee aggregations using deep neural networks and Anisotropic Gaussian Kernel

Manatees are aquatic mammals with voracious appetites. They rely on sea grass as the main food source, and often spend up to eight hours a day grazing. They move slow and frequently stay in groups (i.e. aggregations) in shallow water to search for food, making them vulnerable to environment change and other risks. Accurate counting manatee aggregations within a region is not only biologically meaningful in observing their habit, but also crucial for designing safety rules for boaters, divers, etc., as well as scheduling nursing, intervention, and other plans. In this paper, we propose a deep learning based crowd counting approach to automatically count number of manatees within a region, by using low quality images as input. Because manatees have unique shape and they often stay in shallow water in groups, water surface reflection, occlusion, camouflage etc. making it difficult to accurately count manatee numbers. To address the challenges, we propose to use Anisotropic Gaussian Kernel (AGK), with tunable rotation and variances, to ensure that density functions can maximally capture shapes of individual manatees in different aggregations. After that, we apply AGK kernel to different types of deep neural networks primarily designed for crowd counting, including VGG, SANet, Congested Scene Recognition network (CSRNet), MARUNet etc. to learn manatee densities and calculate number of manatees in the scene. By using generic low quality images extracted from surveillance videos, our experiment results and comparison show that AGK kernel based manatee counting achieves minimum Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). The proposed method works particularly well for counting manatee aggregations in environments with complex background.

Fusion of visible and thermal images improves automated detection and classification of animals for drone surveys

Visible and thermal images acquired from drones (unoccupied aircraft systems) have substantially improved animal monitoring. Combining complementary information from both image types provides a powerful approach for automating detection and classification of multiple animal species to augment drone surveys. We compared eight image fusion methods using thermal and visible drone images combined with two supervised deep learning models, to evaluate the detection and classification of white-tailed deer (Odocoileus virginianus), domestic cow (Bos taurus), and domestic horse (Equus caballus). We classified visible and thermal images separately and compared them with the results of image fusion. Fused images provided minimal improvement for cows and horses compared to visible images alone, likely because the size, shape, and color of these species made them conspicuous against the background. For white-tailed deer, which were typically cryptic against their backgrounds and often in shadows in visible images, the added information from thermal images improved detection and classification in fusion methods from 15 to 85%. Our results suggest that image fusion is ideal for surveying animals inconspicuous from their backgrounds, and our approach uses few image pairs to train compared to typical machine-learning methods. We discuss computational and field considerations to improve drone surveys using our fusion approach.

Thermal aerial culling for the control of vertebrate pest populations

Helicopter-based shooting is an effective management tool for large vertebrate pest animals. However, animals in low-density populations and/or dense habitat can be difficult to locate visually. Thermal-imaging technology can increase detections in these conditions. We used thermal-imaging equipment with a specific helicopter crew configuration to assist in aerial culling for feral pigs (Sus scrofa) and fallow deer (Dama dama) in South Australia in 2021. Seventy-two percent of pigs and 53% of deer were first detected in dense canopy/tall forest habitat. Median time from the first impact shot to incapacitation was < 12 s. The culling rate (animals hour-1) doubled compared to visual shoots over the same populations and the wounding rate was zero resulting in a incapacitation efficiency of 100%. The crew configuration gave the shooter a wide field of view and the thermal operator behind the shooter provided essential support to find new and escaping animals, and to confirm species identification and successful removal. The crew configuration allowed for successful target acquisition and tracking, with reduced target escape. The approach can increase the efficiency of aerial culling, has the potential to increase the success of programs where eradication is a viable option, and can improve animal welfare outcomes by reducing wounding rates and the escape of target animals.

Machine Learning-Based Sensor Data Fusion for Animal Monitoring: Scoping Review

The development of technology, such as the Internet of Things and artificial intelligence, has significantly advanced many fields of study. Animal research is no exception, as these technologies have enabled data collection through various sensing devices. Advanced computer systems equipped with artificial intelligence capabilities can process these data, allowing researchers to identify significant behaviors related to the detection of illnesses, discerning the emotional state of the animals, and even recognizing individual animal identities. This review includes articles in the English language published between 2011 and 2022. A total of 263 articles were retrieved, and after applying inclusion criteria, only 23 were deemed eligible for analysis. Sensor fusion algorithms were categorized into three levels: Raw or low (26%), Feature or medium (39%), and Decision or high (34%). Most articles focused on posture and activity detection, and the target species were primarily cows (32%) and horses (12%) in the three levels of fusion. The accelerometer was present at all levels. The findings indicate that the study of sensor fusion applied to animals is still in its early stages and has yet to be fully explored. There is an opportunity to research the use of sensor fusion for combining movement data with biometric sensors to develop animal welfare applications. Overall, the integration of sensor fusion and machine learning algorithms can provide a more in-depth understanding of animal behavior and contribute to better animal welfare, production efficiency, and conservation efforts.

Counting animals in aerial images with a density map estimation model

Animal abundance estimation is increasingly based on drone or aerial survey photography. Manual postprocessing has been used extensively; however, volumes of such data are increasing, necessitating some level of automation, either for complete counting, or as a labour-saving tool. Any automated processing can be challenging when using such tools on species that nest in close formation such as Pygoscelis penguins. We present here a customized CNN-based density map estimation method for counting of penguins from low-resolution aerial photography. Our model, an indirect regression algorithm, performed significantly better in terms of counting accuracy than standard detection algorithm (Faster-RCNN) when counting small objects from low-resolution images and gave an error rate of only 0.8 percent. Density map estimation methods as demonstrated here can vastly improve our ability to count animals in tight aggregations and demonstrably improve monitoring efforts from aerial imagery.

The broad scale impact of climate change on planning aerial wildlife surveys with drone-based thermal cameras

Helicopters used for aerial wildlife surveys are expensive, dangerous and time consuming. Drones and thermal infrared cameras can detect wildlife, though the ability to detect individuals is dependent on weather conditions. While we have a good understanding of local weather conditions, we do not have a broad-scale assessment of ambient temperature to plan drone wildlife surveys. Climate change will affect our ability to conduct thermal surveys in the future. Our objective was to determine optimal annual and daily time periods to conduct surveys. We present a case study in Texas, (United States of America [USA]) where we acquired and compared average monthly temperature data from 1990 to 2019, hourly temperature data from 2010 to 2019 and projected monthly temperature data from 2021 to 2040 to identify areas where surveys would detect a commonly studied ungulate (white-tailed deer [Odocoileus virginianus]) during sunny or cloudy conditions. Mean temperatures increased when comparing the 1990-2019 to 2010-2019 periods. Mean temperatures above the maximum ambient temperature in which white-tailed deer can be detected increased in 72, 10, 10, and 24 of the 254 Texas counties in June, July, August, and September, respectively. Future climate projections indicate that temperatures above the maximum ambient temperature in which white-tailed deer can be detected will increase in 32, 12, 15, and 47 counties in June, July, August, and September, respectively when comparing 2010-2019 with 2021-2040. This analysis can assist planning, and scheduling thermal drone wildlife surveys across the year and combined with daily data can be efficient to plan drone flights.

Artificial intelligence for automated detection of large mammals creates path to upscale drone surveys

Imagery from drones is becoming common in wildlife research and management, but processing data efficiently remains a challenge. We developed a methodology for training a convolutional neural network model on large-scale mosaic imagery to detect and count caribou (Rangifer tarandus), compare model performance with an experienced observer and a group of naïve observers, and discuss the use of aerial imagery and automated methods for large mammal surveys. Combining images taken at 75 m and 120 m above ground level, a faster region-based convolutional neural network (Faster-RCNN) model was trained in using annotated imagery with the labels: "adult caribou", "calf caribou", and "ghost caribou" (animals moving between images, producing blurring individuals during the photogrammetry processing). Accuracy, precision, and recall of the model were 80%, 90%, and 88%, respectively. Detections between the model and experienced observer were highly correlated (Pearson: 0.96-0.99, P value < 0.05). The model was generally more effective in detecting adults, calves, and ghosts than naïve observers at both altitudes. We also discuss the need to improve consistency of observers' annotations if manual review will be used to train models accurately. Generalization of automated methods for large mammal detections will be necessary for large-scale studies with diverse platforms, airspace restrictions, and sensor capabilities.

Towards the fully automated monitoring of ecological communities

High-resolution monitoring is fundamental to understand ecosystems dynamics in an era of global change and biodiversity declines. While real-time and automated monitoring of abiotic components has been possible for some time, monitoring biotic components-for example, individual behaviours and traits, and species abundance and distribution-is far more challenging. Recent technological advancements offer potential solutions to achieve this through: (i) increasingly affordable high-throughput recording hardware, which can collect rich multidimensional data, and (ii) increasingly accessible artificial intelligence approaches, which can extract ecological knowledge from large datasets. However, automating the monitoring of facets of ecological communities via such technologies has primarily been achieved at low spatiotemporal resolutions within limited steps of the monitoring workflow. Here, we review existing technologies for data recording and processing that enable automated monitoring of ecological communities. We then present novel frameworks that combine such technologies, forming fully automated pipelines to detect, track, classify and count multiple species, and record behavioural and morphological traits, at resolutions which have previously been impossible to achieve. Based on these rapidly developing technologies, we illustrate a solution to one of the greatest challenges in ecology: the ability to rapidly generate high-resolution, multidimensional and standardised data across complex ecologies.

Robust Algorithms for Drone-Assisted Monitoring of Big Animals in Harsh Conditions of Siberian Winter Forests: Recovery of European elk (Alces alces) in Salair Mountains

Simple Summary

Forest animals can be used as a sensitive indicator of the real state of biodiversity. The research objective was to study the potential of drone planes equipped with thermal infrared imaging cameras for large animal monitoring in the conditions of Siberian winter forests with snow background at temperatures of −5 °C to −30 °C. The surveyed territory included the Salair State Nature Reserve in the Kemerovo Region, Russia. Drone planes were effective in covering large areas, while thermal infrared cameras provided accurate information in the harsh winter conditions of Siberia. The research featured the population of the European elk (Alces alces), which is gradually deteriorating due to poaching and deforestation. The designed technical methods and analytic algorithms are cost-efficient and they can be applied for monitoring large areas of Siberian, Canadian and Alaskan winter forests.

Abstract

There are two main reasons for monitoring the population of forest animals. First, regular surveys reveal the real state of biodiversity. Second, they guarantee a prompt response to any negative environmental factor that affects the animal population and make it possible to eliminate the threat before any permanent damage is done. The research objective was to study the potential of drone planes equipped with thermal infrared imaging cameras for large animal monitoring in the conditions of Siberian winter forests with snow background at temperatures −5 °C to −30 °C. The surveyed territory included the Salair State Nature Reserve in the Kemerovo Region, Russia. Drone planes were effective in covering large areas, while thermal infrared cameras provided accurate statistics in the harsh winter conditions of Siberia. The research featured the population of the European elk (Alces alces), which is gradually deteriorating due to poaching and deforestation. The authors developed an effective methodology for processing the data obtained from drone-mounted thermal infrared cameras. The research provided reliable results concerning the changes in the elk population on the territory in question. The use of drone planes proved an effective means of ungulate animal surveying in snow-covered winter forests. The designed technical methods and analytic algorithms are cost-efficient and they can be applied for monitoring large areas of Siberian and Canadian winter forests.

Multiyear monitoring of threatened iconic arboreal mammals in a mid‐elevation conservation reserve in eastern Australia

Multiyear investigations of population dynamics are fundamental to threatened species conservation. We used multiseason occupancy based on spotlight surveys to investigate dynamic occupancy of the koala and the greater glider over an 8‐year period that encompassed a severe drought in year 6. We combined our occupancy estimates with literature estimates of density to estimate the population sizes of these species within the focal conservation reserve. Both species showed substantial yearly variation in the probability of detection (koala: 0.13–0.24; greater glider: 0.12–0.36). Detection of the koala did not follow any obvious pattern. Low detection of the greater glider coincided with the drought and two subsequent years. We suggest the low detection reflected a decline in abundance. The probability of occupancy of the koala was estimated to be 0.88 (95% CI: 0.75–1.0) in year 8. Autonomous recording units were also used in year 8, enabling an independent occupancy estimate of 0.80 (0.64–0.90). We found no evidence of a drought‐induced decline in the koala. Habitat variables had a weak influence on koala occupancy probabilities. The probability of occupancy of the greater glider changed little over time, from 0.52 (95% CI: 0.24–0.81) to 0.63 (0.42–0.85) in year 8. Modeling suggested that the probability of colonization was positively influenced by the percentage cover of rainforest. Increased cover of these nonbrowse trees may reflect thermal buffering, site productivity, or soil moisture. We estimate that our study reserve is likely to contain >900 adult koalas and >2400 adult greater gliders. These are among some of the first reserve‐wide estimates for these species. Our study reserve can play an important role in the conservation of both species.

Automated Detection of Koalas with Deep Learning Ensembles

Effective management of threatened and invasive species requires regular and reliable population estimates. Drones are increasingly utilised by ecologists for this purpose as they are relatively inexpensive. They enable larger areas to be surveyed than traditional methods for many species, particularly cryptic species such as koalas, with less disturbance. The development of robust and accurate methods for species detection is required to effectively use the large volumes of data generated by this survey method. The enhanced predictive and computational power of deep learning ensembles represents a considerable opportunity to the ecological community. In this study, we investigate the potential of deep learning ensembles built from multiple convolutional neural networks (CNNs) to detect koalas from low-altitude, drone-derived thermal data. The approach uses ensembles of detectors built from combinations of YOLOv5 and models from Detectron2. The ensembles achieved a strong balance between probability of detection and precision when tested on ground-truth data from radio-collared koalas. Our results also showed that greater diversity in ensemble composition can enhance overall performance. We found the main impediment to higher precision was false positives but expect these will continue to reduce as tools for geolocating detections are improved. The ability to construct ensembles of different sizes will allow for improved alignment between the algorithms used and the characteristics of different ecological problems. Ensembles are efficient and accurate and can be scaled to suit different settings, platforms and hardware availability, making them capable of adaption for novel applications.

Individual Identification of Large Felids in Field Studies: Common Methods, Challenges, and Implications for Conservation Science

Large felids represent some of the most threatened large mammals on Earth, critical for both tourism economies and ecosystem function. Most populations are in a state of decline, and their monitoring and enumeration is therefore critical for conservation. This typically rests on the accurate identification of individuals within their populations. We review the most common and current survey methods used in individual identification studies of large felid ecology (body mass &gt; 25 kg). Remote camera trap photography is the most extensively used method to identify leopards, snow leopards, jaguars, tigers, and cheetahs which feature conspicuous and easily identifiable coat patterning. Direct photographic surveys and genetic sampling are commonly used for species that do not feature easily identifiable coat patterning such as lions. We also discuss the accompanying challenges encountered in several field studies, best practices that can help increase the precision and accuracy of identification and provide generalised ratings for the common survey methods used for individual identification.

Drone thermal imaging technology provides a cost-effective tool for landscape-scale monitoring of a cryptic forest-dwelling species across all population densities

Abstract Context Drones, or remotely piloted aircraft systems, equipped with thermal imaging technology (RPAS thermal imaging) have recently emerged as a powerful monitoring tool for koala populations. Before wide uptake of novel technologies by government, conservation practitioners and researchers, evidence of greater efficiency and cost-effectiveness than with other available methods is required. Aims We aimed to provide the first comprehensive analysis of the cost-effectiveness of RPAS thermal imaging for koala detection against two field-based methods, systematic spotlighting (Spotlight) and the refined diurnal radial search component of the spot-assessment technique (SAT). Methods We conducted various economic comparisons, particularly comparative cost-effectiveness of RPAS thermal imaging, Spotlight and SAT for repeat surveys of a low-density koala population. We compared methods on cost-effectiveness as well as long-term costs by using accumulating cost models. We also compared detection costs across population density using a predictive cost model. Key results Despite substantial hardware, training and licensing costs at the outset (>A$49 900), RPAS thermal imaging surveys were cost-effective, detecting the highest number of koalas per dollar spent. Modelling also suggested that RPAS thermal imaging requires the lowest survey effort to detect koalas within the range of publicly available koala population densities (~0.006–18 koalas ha−1) and would provide long-term cost reductions across longitudinal monitoring programs. RPAS thermal imaging would also require the lowest average survey effort costs at a landscape scale (A$3.84 ha−1), providing a cost-effective tool across large spatial areas. Conclusions Our analyses demonstrated drone thermal imaging technology as a cost-effective tool for conservation practitioners monitoring koala populations. Our analyses may also form the basis of decision-making tools to estimate survey effort or total program costs across any koala population density. Implications Our novel approach offers a means to perform various economic comparisons of available survey techniques and guide investment decisions towards developing standardised koala monitoring approaches. Our results may assist stakeholders and policymakers to confidently invest in RPAS thermal imaging technology and achieve optimal conservation outcomes for koala populations, with standardised data collection delivered through evidence-based and cost-effective monitoring programs.

Greening the Artificial Intelligence for a Sustainable Planet: An Editorial Commentary

Artificial intelligence (AI) is one of the most popular and promising technologies of our time [...]

Counting using deep learning regression gives value to ecological surveys

Many ecological studies rely on count data and involve manual counting of objects of interest, which is time-consuming and especially disadvantageous when time in the field or lab is limited. However, an increasing number of works uses digital imagery, which opens opportunities to automatise counting tasks. In this study, we use machine learning to automate counting objects of interest without the need to label individual objects. By leveraging already existing image-level annotations, this approach can also give value to historical data that were collected and annotated over longer time series (typical for many ecological studies), without the aim of deep learning applications. We demonstrate deep learning regression on two fundamentally different counting tasks: (i) daily growth rings from microscopic images of fish otolith (i.e., hearing stone) and (ii) hauled out seals from highly variable aerial imagery. In the otolith images, our deep learning-based regressor yields an RMSE of 3.40 day-rings and an [Formula: see text] of 0.92. Initial performance in the seal images is lower (RMSE of 23.46 seals and [Formula: see text] of 0.72), which can be attributed to a lack of images with a high number of seals in the initial training set, compared to the test set. We then show how to improve performance substantially (RMSE of 19.03 seals and [Formula: see text] of 0.77) by carefully selecting and relabelling just 100 additional training images based on initial model prediction discrepancy. The regression-based approach used here returns accurate counts ([Formula: see text] of 0.92 and 0.77 for the rings and seals, respectively), directly usable in ecological research.

A Comprehensive Overview of Technologies for Species and Habitat Monitoring and Conservation

The range of technologies currently used in biodiversity conservation is staggering, with innovative uses often adopted from other disciplines and being trialed in the field. We provide the first comprehensive overview of the current (2020) landscape of conservation technology, encompassing technologies for monitoring wildlife and habitats, as well as for on-the-ground conservation management (e.g., fighting illegal activities). We cover both established technologies (routinely deployed in conservation, backed by substantial field experience and scientific literature) and novel technologies or technology applications (typically at trial stage, only recently used in conservation), providing examples of conservation applications for both types. We describe technologies that deploy sensors that are fixed or portable, attached to vehicles (terrestrial, aquatic, or airborne) or to animals (biologging), complemented with a section on wildlife tracking. The last two sections cover actuators and computing (including web platforms, algorithms, and artificial intelligence).

Estimating koala density from incidental koala sightings in South-East Queensland, Australia (1997-2013), using a self-exciting spatio-temporal point process model

The koala, Phascolarctos cinereus, is an iconic Australian wildlife species facing a rapid decline in South-East Queensland (SEQLD). For conservation planning, the ability to estimate the size of koala populations is crucial. Systematic surveys are the most common approach to estimate koala populations but because of their cost they are often restricted to small geographic areas and are conducted infrequently. Public interest and participation in the collection of koala sighting data is increasing in popularity, but such data are generally not used for population estimation. We modeled monthly sightings of koalas reported by members of the public from 1997 to 2013 in SEQLD by developing a self-exciting spatio-temporal point process model. This allowed us to account for characteristics that are associated with koala presence (which vary over both space and time) while accounting for detection bias in the koala sighting process and addressing spatial clustering of observations. The density of koalas varied spatially due to the heterogeneous nature of koala habitat in SEQLD, with a mean density of 0.0019 koalas per km2 over the study period. The percentage of land areas with very low densities (0-0.0005 koalas per km2) remained similar throughout the study period representing, on average, 66% of the total study area. The approach described in this paper provides a useful starting point to allow greater use to be made of incidental koala sighting data. We propose that the model presented here could be used to combine systematic koala survey data (which is spatially restricted, but more precise) with koala sighting data (which is incidental and often biased by nature, but often collected over large geographical areas). Our approach could also be adopted for modeling the density of other wildlife species where data is collected in the same manner.

A comparison of manual and automated detection of rusa deer (

Abstract Context Monitoring is an essential part of managing invasive species; however, accurate, cost-effective detection techniques are necessary for it to be routinely undertaken. Current detection techniques for invasive deer are time consuming, expensive and have associated biases, which may be overcome by exploiting new technologies. Aims We assessed the accuracy and cost effectiveness of automated detection methods in comparison to manual detection of thermal footage of deer captured by remotely piloted aircraft systems. Methods Thermal footage captured by RPAS was assessed using an algorithm combining two object-detection techniques, namely, YOLO and Faster-RCNN. The number of deer found using manual review on each sampling day was compared with the number of deer found on each day using machine learning. Detection rates were compared across survey areas and sampling occasions. Key results Overall, there was no difference in the mean number of deer detected using manual and that detected by automated review (P = 0.057). The automated-detection algorithm identified between 66.7% and 100% of deer detected using manual review of thermal imagery on all but one of the sampling days. There was no difference in the mean proportion of deer detected using either manual or automated review at three repeated sampling events (P = 0.174). However, identifying deer using the automated review algorithm was 84% cheaper than the cost of manual review. Low cloud cover appeared to affect detectability using the automated review algorithm. Conclusions Automated methods provide a fast and effective way to detect deer. For maximum effectiveness, imagery that encompasses a range of environments should be used as part of the training dataset, as well as large groups for herding species. Adequate sensing conditions are essential to gain accurate counts of deer by automated detection. Implications Machine learning in combination with RPAS may decrease the cost and improve the detection and monitoring of invasive species.

Automated Detection of Animals in Low-Resolution Airborne Thermal Imagery

Detecting animals to estimate abundance can be difficult, particularly when the habitat is dense or the target animals are fossorial. The recent surge in the use of thermal imagers in ecology and their use in animal detections can increase the accuracy of population estimates and improve the subsequent implementation of management programs. However, the use of thermal imagers results in many hours of captured flight videos which require manual review for confirmation of species detection and identification. Therefore, the perceived cost and efficiency trade-off often restricts the use of these systems. Additionally, for many off-the-shelf systems, the exported imagery can be quite low resolution (<9 Hz), increasing the difficulty of using automated detections algorithms to streamline the review process. This paper presents an animal species detection system that utilises the cost-effectiveness of these lower resolution thermal imagers while harnessing the power of transfer learning and an enhanced small object detection algorithm. We have proposed a distant object detection algorithm named Distant-YOLO (D-YOLO) that utilises YOLO (You Only Look Once) and improves its training and structure for the automated detection of target objects in thermal imagery. We trained our system on thermal imaging data of rabbits, their active warrens, feral pigs, and kangaroos collected by thermal imaging researchers in New South Wales and Western Australia. This work will enhance the visual analysis of animal species while performing well on low, medium and high-resolution thermal imagery.

Hot stuff in the bushes: Thermal imagers and the detection of burrows in vegetated sites

Thermal imaging technology is a developing field in wildlife management. Most thermal imaging work in wildlife science has been limited to larger ungulates and surface-dwelling mammals. Little work has been undertaken on the use of thermal imagers to detect fossorial animals and/or their burrows. Survey methods such as white-light spotlighting can fail to detect the presence of burrows (and therefore the animals within), particularly in areas where vegetation obscures burrows. Thermal imagers offer an opportunity to detect the radiant heat from these burrows, and therefore the presence of the animal, particularly in vegetated areas. Thermal imaging technology has become increasingly available through the provision of smaller, more cost-effective units. Their integration with drone technology provides opportunities for researchers and land managers to utilize this technology in their research/management practices.We investigated the ability of both consumer (AUD$65,000) mounted on drones to detect rabbit burrows (warrens) and entrances in the landscape as compared to visual assessment.Thermal imagery and visual inspection detected active rabbit warrens when vegetation was scarce. The presence of vegetation was a significant factor in detecting entrances (p < .001, α = 0.05). The consumer imager did not detect as many warren entrances as either the professional imager or visual inspection (p = .009, α = 0.05). Active warren entrances obscured by vegetation could not be accurately identified on exported imagery from the consumer imager and several false-positive detections occurred when reviewing this footage.We suggest that the exportable frame rate (Hz) was the key factor in image quality and subsequent false-positive detections. This feature should be considered when selecting imagers and suggest that a minimum export rate of 30 Hz is required. Thermal imagers are a useful additional tool to aid in identification of entrances for active warrens and professional imagers detected more warrens and entrances than either consumer imagers or visual inspection.

Evaluating new technology for biodiversity monitoring: Are drone surveys biased?

Drones and machine learning-based automated detection methods are being used by ecologists to conduct wildlife surveys with increasing frequency. When traditional survey methods have been evaluated, a range of factors have been found to influence detection probabilities, including individual differences among conspecific animals, which can thus introduce biases into survey counts. There has been no such evaluation of drone-based surveys using automated detection in a natural setting. This is important to establish since any biases in counts made using these methods will need to be accounted for, to provide accurate data and improve decision-making for threatened species. In this study, a rare opportunity to survey a ground-truthed, individually marked population of 48 koalas in their natural habitat allowed for direct comparison of the factors impacting detection probability in both ground observation and drone surveys with manual and automated detection. We found that sex and host tree preferences impacted detection in ground surveys and in manual analysis of drone imagery with female koalas likely to be under-represented, and koalas higher in taller trees detected less frequently when present. Tree species composition of a forest stand also impacted on detections. In contrast, none of these factors impacted on automated detection. This suggests that the combination of drone-captured imagery and machine learning does not suffer from the same biases that affect conventional ground surveys. This provides further evidence that drones and machine learning are promising tools for gathering reliable detection data to better inform the management of threatened populations.

Animal Welfare Implications of Digital Tools for Monitoring and Management of Cattle and Sheep on Pasture

The opportunities for natural animal behaviours in pastures imply animal welfare benefits. Nevertheless, monitoring the animals can be challenging. The use of sensors, cameras, positioning equipment and unmanned aerial vehicles in large pastures has the potential to improve animal welfare surveillance. Directly or indirectly, sensors measure environmental factors together with the behaviour and physiological state of the animal, and deviations can trigger alarms for, e.g., disease, heat stress and imminent calving. Electronic positioning includes Radio Frequency Identification (RFID) for the recording of animals at fixed points. Positioning units (GPS) mounted on collars can determine animal movements over large areas, determine their habitat and, somewhat, health and welfare. In combination with other sensors, such units can give information that helps to evaluate the welfare of free-ranging animals. Drones equipped with cameras can also locate and count the animals, as well as herd them. Digitally defined virtual fences can keep animals within a predefined area without the use of physical barriers, relying on acoustic signals and weak electric shocks. Due to individual variations in learning ability, some individuals may be exposed to numerous electric shocks, which might compromise their welfare. More research and development are required, especially regarding the use of drones and virtual fences.

Evaluating Alternative Flight Plans in Thermal Drone Wildlife Surveys—Simulation Study

The rapidly developing technology of unmanned aerial vehicles (drones) extends to the availability of aerial surveys for wildlife research and management. However, regulations limiting drone operations to visual line of sight (VLOS) seriously affect the design of surveys, as flight paths must be concentrated within small sampling blocks. Such a design is inferior to spatially unrestricted randomized designs available if operations beyond visual line of sight (BVLOS) are allowed. We used computer simulations to assess whether the VLOS rule affects the accuracy and precision of wildlife density estimates derived from drone collected data. We tested two alternative flight plans (VLOS vs. BVLOS) in simulated surveys of low-, medium- and high-density populations of a hypothetical ungulate species with three levels of effort (one to three repetitions). The population density was estimated using the ratio estimate and distance sampling method. The observed differences in the accuracy and precision of estimates from the VLOS and BVLOS surveys were relatively small and negligible. Only in the case of the low-density population (2 ind./100 ha) surveyed once was the VLOS design inferior to BVLOS, delivering biased and less precise estimates. These results show that while the VLOS regulations complicate survey logistics and interfere with random survey design, the quality of derived estimates does not have to be compromised. We advise testing alternative survey variants with the aid of computer simulations to achieve reliable estimates while minimizing survey costs.

Trialling a real-time drone detection and validation protocol for the koala (Phascolarctos cinereus)

Remotely piloted aircraft system (RPAS), or drone, technology has emerged as a promising survey method for the cryptic koala. We demonstrate an in-field protocol for wild koala RPAS surveys which provides real-time validation of thermal signatures. During 15 trial flights using a quadcopter drone (DJI Matrice 200 v2) we successfully detected and validated koala thermal signatures (n=12) using two in-field approaches: validation by on-ground observer (n=10) and validation using 4K footage captured and reviewed directly after the survey (n=2). We also provide detectability considerations relative to survey time, temperature, wildlife–RPAS interactions and detection of non-target species, which can be used to further inform RPAS survey protocols.

Real-time drone derived thermal imagery outperforms traditional survey methods for an arboreal forest mammal

Koalas (Phascolarctos cinereus) are cryptic and currently face regional extinction. The direct detection (physical sighting) of individuals is required to improve conservation management strategies. We provide a comparative assessment of three survey methods for the direct detection of koalas: systematic spotlighting (Spotlight), remotely piloted aircraft system thermal imaging (RPAS), and the refined diurnal radial search component of the spot assessment technique (SAT). Each survey method was repeated on the same morning with independent observers (03:00-12:00 hrs) for a total of 10 survey occasions at sites with fixed boundaries (28-76 ha) in Port Stephens (n = 6) and Gilead (n = 1) in New South Wales between May and July 2019. Koalas were directly detected on 22 occasions during 7 of 10 comparative surveys (Spotlight: n = 7; RPAS: n = 14; and SAT: n = 1), for a total of 12 unique individuals (Spotlight: n = 4; RPAS: n = 11; SAT: n = 1). In 3 of 10 comparative surveys no koalas were detected. Detection probability was 38.9 ± 20.03% for Spotlight, 83.3 ± 11.39% for RPAS and 4.2 ± 4.17% for SAT. Effective detectability per site was 1 ± 0.44 koalas per 6.75 ± 1.03 hrs for Spotlight (1 koala per 6.75 hrs), 2 ± 0.38 koalas per 4.35 ± 0.28 hrs for RPAS (1 koala per 2.18 hrs) and 0.14 ± 0.14 per 6.20 ± 0.93 hrs for SAT (1 koala per 43.39 hrs). RPAS thermal imaging technology appears to offer an efficient method to directly survey koalas comparative to Spotlight and SAT and has potential as a valuable conservation tool to inform on-ground management of declining koala populations.

New technologies in the mix: Assessing N-mixture models for abundance estimation using automated detection data from drone surveys

Reliable estimates of abundance are critical in effectively managing threatened species, but the feasibility of integrating data from wildlife surveys completed using advanced technologies such as remotely piloted aircraft systems (RPAS) and machine learning into abundance estimation methods such as N-mixture modeling is largely unknown due to the unique sources of detection errors associated with these technologies.We evaluated two modeling approaches for estimating the abundance of koalas detected automatically in RPAS imagery: (a) a generalized N-mixture model and (b) a modified Horvitz-Thompson (H-T) estimator method combining generalized linear models and generalized additive models for overall probability of detection, false detection, and duplicate detection. The final estimates from each model were compared to the true number of koalas present as determined by telemetry-assisted ground surveys.The modified H-T estimator approach performed best, with the true count of koalas captured within the 95% confidence intervals around the abundance estimates in all 4 surveys in the testing dataset (n = 138 detected objects), a particularly strong result given the difficulty in attaining accuracy found with previous methods.The results suggested that N-mixture models in their current form may not be the most appropriate approach to estimating the abundance of wildlife detected in RPAS surveys with automated detection, and accurate estimates could be made with approaches that account for spurious detections.

Integrating drone-borne thermal imaging with artificial intelligence to locate bird nests on agricultural land

In conservation, the use of unmanned aerial vehicles (drones) carrying various sensors and the use of deep learning are increasing, but they are typically used independently of each other. Untapping their large potential requires integrating these tools. We combine drone-borne thermal imaging with artificial intelligence to locate ground-nests of birds on agricultural land. We show, for the first time, that this semi-automated system can identify nests with a high performance. However, local weather, type of arable field and height of the drone can affect performance. The results' implications are particularly relevant to conservation practitioners working across sectors, such as biodiversity conservation and food production in farmland. Under a rapidly changing world, studies like this can help uncover the potential of technology for conservation and embrace cross-sectoral transformations from the onset; for example, by integrating nest detection within the precision agriculture system that heavily relies on drone-borne sensors.

Evaluating the Efficacy and Optimal Deployment of Thermal Infrared and True-Colour Imaging When Using Drones for Monitoring Kangaroos

Advances in drone technology have given rise to much interest in the use of drone-mounted thermal imagery in wildlife monitoring. This research tested the feasibility of monitoring large mammals in an urban environment and investigated the influence of drone flight parameters and environmental conditions on their successful detection using thermal infrared (TIR) and true-colour (RGB) imagery. We conducted 18 drone flights at different altitudes on the Sunshine Coast, Queensland, Australia. Eastern grey kangaroos (Macropus giganteus) were detected from TIR (n=39) and RGB orthomosaics (n=33) using manual image interpretation. Factors that predicted the detection of kangaroos from drone images were identified using unbiased recursive partitioning. Drone-mounted imagery achieved an overall 73.2% detection success rate using TIR imagery and 67.2% using RGB imagery when compared to on-ground counts of kangaroos. We showed that the successful detection of kangaroos using TIR images was influenced by vegetation type, whereas detection using RGB images was influenced by vegetation type, time of day that the drone was deployed, and weather conditions. Kangaroo detection was highest in grasslands, and kangaroos were not successfully detected in shrublands. Drone-mounted TIR and RGB imagery are effective at detecting large mammals in urban and peri-urban environments.

Evaluating the use of thermal imaging cameras to monitor the endangered greater bilby at Astrebla Downs National Park

Spotlight surveys are widely used to monitor arid-zone-dwelling species such as the greater bilby (Macrotis lagotis). These surveys require a sufficient sample size to adequately model detection probability. Adequate sample sizes can be difficult to obtain for low-density populations and for species that avoid light and or have poor eyeshine like the bilby. Abundance estimates based on burrow counts can be problematic because of the variable relationship between the number of burrows used and bilby abundance. In 2013, feral predators devastated a Queensland bilby population and a method was required that could locate and monitor the remaining bilbies. We report on a study that compared density estimates derived from spotlighting and thermal cameras. Bilbies were surveyed annually over three years, using spotlights and thermal cameras on different nights but using the same transects to compare the methods. On average, thermal cameras detected twice the number of bilbies per kilometre surveyed than spotlighting. Despite this difference in the number of bilbies detected, density estimates (bilbies km−2) were similar (thermal camera versus spotlight: 0.6 versus 0.2 (2014), 3.4 versus 3.4 (2015) and 4.8 versus 3.3 (2016)). Nevertheless, the larger sample size obtained using thermal cameras gave greater confidence in modelling detection probability.

An experimental approach to evaluate the potential of drones in terrestrial mammal research: a gregarious ungulate as a study model

Research on the use of unmanned aircraft systems (UAS) in wildlife has made remarkable progress recently. Few studies to date have experimentally evaluated the effect of UAS on animals and have usually focused primarily on aquatic fauna. In terrestrial open arid ecosystems, with relatively good visibility to detect animals but little environmental noise, there should be a trade-off between flying the UAS at high height above ground level (AGL) to limit the disturbance of animals and flying low enough to maintain count precision. In addition, body size or social aggregation of species can also affect the ability to detect animals from the air and their response to the UAS approach. To address this gap, we used a gregarious ungulate, the guanaco (Lama guanicoe), as a study model. Based on three types of experimental flights, we demonstrated that (i) the likelihood of miscounting guanacos in images increases with UAS height, but only for offspring and (ii) higher height AGL and lower UAS speed reduce disturbance, except for large groups, which always reacted. Our results call into question mostly indirect and observational previous evidence that terrestrial mammals are more tolerant to UAS than other species and highlight the need for experimental and species-specific studies before using UAS methods.

Using virtual reality and thermal imagery to improve statistical modelling of vulnerable and protected species

Biodiversity loss and sparse observational data mean that critical conservation decisions may be based on little to no information. Emerging technologies, such as airborne thermal imaging and virtual reality, may facilitate species monitoring and improve predictions of species distribution. Here we combined these two technologies to predict the distribution of koalas, specialized arboreal foliovores facing population declines in many parts of eastern Australia. For a study area in southeast Australia, we complemented ground-survey records with presence and absence observations from thermal-imagery obtained using Remotely-Piloted Aircraft Systems. These field observations were further complemented with information elicited from koala experts, who were immersed in 360-degree images of the study area. The experts were asked to state the probability of habitat suitability and koala presence at the sites they viewed and to assign each probability a confidence rating. We fit logistic regression models to the ground survey data and the ground plus thermal-imagery survey data and a Beta regression model to the expert elicitation data. We then combined parameter estimates from the expert-elicitation model with those from each of the survey models to predict koala presence and absence in the study area. The model that combined the ground, thermal-imagery and expert-elicitation data substantially reduced the uncertainty around parameter estimates and increased the accuracy of classifications (koala presence vs absence), relative to the model based on ground-survey data alone. Our findings suggest that data elicited from experts using virtual reality technology can be combined with data from other emerging technologies, such as airborne thermal-imagery, using traditional statistical models, to increase the information available for species distribution modelling and the conservation of vulnerable and protected species.