Drone Surveys Research Articles

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.

A Drone Study of Sociality in the Finless Porpoise Neophocaena asiaeorientalis in the Ariake Sound, Japan

The social structure of animal populations is a fundamental component of their biology, influencing gene flow, habitat use, competition and co-operation around resources, and communication. However, ecological and social relationships can be challenging to describe in most marine mammals, who spend the majority of their lives underwater. The finless porpoise (Neophocaena asiaeorientalis) is one such cetacean species with a largely unknown social structure. Recent advances in drone technology enable more systematic surveys, photogrammetry, and photo-identification for diverse animal species. The present study aimed to validate new survey methods and provide a preliminary description of the spatiotemporal distribution of free-ranging finless porpoises in the coastal open-sea area of Ariake Sound, Japan. A vertical take-off and landing (VTOL) drone equipped with an action camera yielded GPS location datasets through line and area surveys, covering a total sea area of 120 km2. The results suggest highly flexible and varied aggregation sizes in finless porpoises. Distance analysis across individuals and aggregations revealed a cohesive tendency among groups, compared to solitaries and in pairs. Therefore, the present VTOL drone surveys both elucidated some social aspects of the study population and confirmed the efficacy of these standardized research protocols involving automated, programmed, and repeatable flight missions.

Faster and Better: Comparison between Traditional and Drone Monitoring in a Cryptic Species, the Purple Heron Ardea purpurea

Monitoring waterbirds is fundamental to understand the health status of wetland habitats. However, this monitoring has to be conducted by means of reliable data collection that can provide accurate information on population trends. Usually, waterbird monitoring is difficult, as nesting grounds are usually located in inaccessible reedbed, and by eye detection of cryptic species is hard. Drones have the capacity to overcome most of these problems, as they can provide with an aerial view of places otherwise unreachable, while reducing the disturbance and time spent in the field. The present study aims to compare the accuracy, disturbance levels, and managerial efficiency between ground (traditional) and drone counts of a cryptic species, the Purple Heron Ardea purpurea. Traditional monitoring methods were only capable of detecting 35% of the nesting pairs detected by the drone surveys (8.0 ± 11.8 versus 22.9 ± 38.2 nesting pairs in ground and drone surveys, respectively). Consequently, colony size estimates between methods showed poor agreement, to the point that traditional methods missed colonies otherwise detected by the drone. No apparent negative effects on nesting pairs where found when flying the drone. In addition, mean time spent to survey breeding sites with a drone was far less than with the traditional approach, down to a six-fold time reduction. This reduction, together with a lack of disturbance observed when conducting the drone monitoring, and an increasing monitoring precision and accuracy, supports the use of drones as the least invasive option for studies on population monitoring on hardly accessible sites.

Application of Artificial Intelligence in Drones for the Analysis of Agricultural Land Use in the Mining Lease

The utilization of artificial intelligence (AI) has facilitated the automation of drone control, which includes the management of navigation and movement. This application can be accomplished through several methods, including GPS tracking, computer vision, and machine learning algorithms. Drones exhibit a distinctive combination of spatial coverage and resolution, rendering them indispensable for land survey and mapping. The incorporation of multiple ground-control points has the potential to yield high precision georeferencing for the Orthomosaic product.
 In conjunction with field observations, drones provide a prompt and precise means of recording land data and its use. A drone survey and mapping operation was conducted within a mining lease situated near the village of Kanthariya, in the Tehsil and District of Chittorgarh, covering an area of 64.75 hectares, for the analysis of agricultural land use in the mining lease.

Camera Trap Methods and Drone Thermal Surveillance Provide Reliable, Comparable Density Estimates of Large, Free-Ranging Ungulates.

Camera traps and drone surveys both leverage advancing technologies to study dynamic wildlife populations with little disturbance. Both techniques entail strengths and weaknesses, and common camera trap methods can be confounded by unrealistic assumptions and prerequisite conditions. We compared three methods to estimate the population density of white-tailed deer (Odocoileus virgnianus) in a section of Pilot Mountain State Park, NC, USA: (1) camera trapping using mark-resight ratios or (2) N-mixture modeling and (3) aerial thermal videography from a drone platform. All three methods yielded similar density estimates, suggesting that they converged on an accurate estimate. We also included environmental covariates in the N-mixture modeling to explore spatial habitat use, and we fit models for each season to understand temporal changes in population density. Deer occurred in greater densities on warmer, south-facing slopes in the autumn and winter and on cooler north-facing slopes and in areas with flatter terrain in the summer. Seasonal density estimates over two years suggested an annual cycle of higher densities in autumn and winter than in summer, indicating that the region may function as a refuge during the hunting season.

Patterns of overlapping habitat use of juvenile white shark and human recreational water users along southern California beaches.

Juvenile white sharks (JWS) of the Northeastern Pacific population are present in nearshore southern California waters and form mixed size class (~1.5-3 m) aggregations for weeks to months, often within 500 m of shore. These nearshore beach habitats are heavily used for human recreation (e.g., surfing, swimming, body boarding, wading, and standup paddleboarding) and the amount of spatio-temporal overlap between JWS and humans is currently unknown. Increases in human population and the Northeastern Pacific population of white sharks have raised concern over human beach safety. To determine spatio-temporal JWS-human overlap at various spatial scales (e.g., across the entire southern California coastline, across different distances from shore, and within specific beach locations), 26 beach locations across southern California were surveyed monthly resulting in 1644 aerial drone surveys between January 2019 to March 2021. Thirteen environmental variables were assessed to predict when spatio-temporal overlap between JWS and water users was highest. Coast-wide distribution of JWS was clumped, limiting human-shark co-occurrence to specific locations, with 1096 of 1204 JWS observations occurring at Carpinteria and Del Mar Beach locations. Nearshore distribution indicated JWS are often close enough to the wave break to interact with some water users (median = 101 m, range = 2-702 m), although JWS had the most spatial overlap with stand-up paddlers. Daily human-shark co-occurrence was 97% at beaches where JWS aggregations had formed, and human activity showed high spatial overlap at shark aggregation sites. Although there is higher seasonal human-shark spatio-temporal overlap where aggregations form in southern California, the number of unprovoked shark bites across southern California is extremely low. This study provides evidence that high human-shark spatio-temporal overlap does not lead to an increased bite frequency in southern California, and there are a number of possible explanations as to why JWS are not biting water users despite daily encounters.

IoT, machine learning and photogrammetry in small hydropower towards energy and digital transition: potential energy and viability analyses

This research aims to evaluate and put into practise the design of a small hydropower plant on a stream at São Vicente, in Madeira Island, supported by internet of things (IoT). The photogrammetry technique is also used with a comprehensive digital transformation, in which new concepts, methods and models, such as machine learning (ML), and big data analytics play an important role due to the huge availability time series that have to be exploited in hydropower design studies. Nowadays, digitalization and massive data availability are imposing new ways to address many of the current challenges associated with the energy and digital transition. This research is based on a simple small hydropower design, to present an integrated methodology using new methods assigned by an internet protocol system, which includes the development of different steps and components supported by GIS, photogrammetry and the use of advanced tools, with the support of a drone survey with internet communication (IoT) that allow the generation of experimentally-based estimates in situ characterization, the volumetric flow, the hydrological data treatment, the hydraulic calculations and economic estimations for a real hydro project. Therefore, hydrological variables, hydraulic analysis and topographical survey are carried out in the IoT application platform supported by new tools and methods to optimise the size of hydraulic structures, estimate the performance and potential of the hydropower plant towards the best solution for energy and digital transition. Firstly, the data-base for the all study and posterior sizing of the case study of hydropower plant are defined and then the corresponding analyses and results are presented. Then, the cost estimation for the construction, maintenance and operation of the selected elements that compose the hydropower topology are determined, as well as the respective economic balance, considering the annual energy production. In addition, both economic and environmental return on investment is discussed. Finally, an analysis to equate the cost estimates and the respective benefits of hydropower generation using this new approach applicability is stablished, taking into account some economic indicators to determine the profitability of the project.

Three thousand years of river channel engineering in the Nile Valley

AbstractAcross a 1000‐km stretch of the River Nile, from the 1st Cataract in southern Egypt to the 4th Cataract in Sudan, many hundreds of drystone walls are located within active channels, on seasonally inundated floodplains or in now‐dry Holocene palaeochannel belts. These walls (or river groynes) functioned as flood and flow control structures and are of a type now commonly in use worldwide. In the Nile Valley, the structures have been subject only to localised investigations, and none have been radiometrically dated. Some were built within living memory to trap nutrient‐rich Nile silts for agriculture, a practice already recorded in the early 19th century C.E. However, others situated within ancient palaeochannel belts indicate construction over much longer time frames. In this paper, we map the distribution of these river groynes using remote sensing and drone survey. We then establish their probable functions and a provisional chronology using ethnoarchaeological investigation and the ground survey, excavation and radiometric dating of the structures in northern Sudan, focusing on the Holocene riverine landscape surrounding the pharaonic settlement of Amara West (c. 1300–1000 B.C.E.). Finally, we consider the historical and economic implications of this form of hydraulic engineering in the Nile Valley over the past three millennia.

Unmanned aerial vehicles (UAVs) with thermal infrared (TIR) sensors are effective for monitoring and counting threatened Vietnamese primates.

Monitoring the population size of threatened primate species with minimal disturbance is becoming an outstanding requirement for conservation and wildlife management. Drones with thermal infrared (TIR) and visible spectrum (RGB) imaging are increasingly used to survey arboreal primates, but ground-truthing is still required to assess the effectiveness of drone-based count estimates. Our pilot study aims to assess the ability of a drone with both TIR and RGB sensors to detect, count, and identify semi-wild population of four endangered species of langurs and gibbon in the Endangered Primate Rescue Center (EPRC) in northern Vietnam. We found that TIR imagery enabled higher detection rates compared to RGB imagery and obtained an accurate count with the TIR only after four drone flights. We could identify langurs species based on thermal signature at a flight height of 50m from the ground level (max tree height = 15m), via size and shape of the body. With TIR imagery, we were able to record inconspicuous behaviors such as foraging and play. While some individuals initially showed flight or avoidance behaviors when the drone was sighted, these behaviors decreased or were absent on following drone surveys. Our study suggests that monitoring and precisely counting langur and gibbon species populations could be successful with the use of thermal drones only.

Spatial Analysis of Soil Moisture and Turfgrass Health to Determine Zones for Spatially Variable Irrigation Management

Irrigated turfgrass is a major crop in urban areas of the drought-stricken Western United States. A considerable proportion of irrigation water is wasted through the use of conventional sprinkler systems. While smart sprinkler systems have made progress in reducing temporal mis-applications, more research is needed to determine the most appropriate variables for accurately and cost-effectively determining spatial zones for irrigation application. This research uses data from ground and drone surveys of two large sports fields. Surveys were conducted pre-, within and towards the end of the irrigation season to determine spatial irrigation zones. Principal components analysis and k-means classification were used to develop zones using several variables individually and combined. The errors associated with uniform irrigation and different configurations of spatial zones are assessed to determine comparative improvements in irrigation efficiency afforded by spatial irrigation zones. A determination is also made as to whether the spatial zones can be temporally static or need to be re-determined periodically. Results suggest that zones based on spatial soil moisture surveys and simple observations of whether the grass felt wet or dry are better than those based on NDVI, other variables and several variables in combination. In addition, due to the temporal variations observed in spatial patterns, ideally zones should be re-evaluated periodically. However, a less labor-intensive solution is to determine temporally static zones based on patterns in soil moisture averaged from several surveys. Of particular importance are the spatial patterns observed prior to the start of the irrigation season as they reflect more temporally stable variation that relates to soil texture and topography rather than irrigation management.

Phenology of harbor seal pupping and the influence of weather on pup counts, investigated by UAV

AbstractSurveys of harbor seals to assess pup production in Denmark are conducted at the assumed breeding season peak between 9 and 15 hr at winds <10 m/s, with no precipitation at least 6 hr before observations and at low tide in areas with a significant tidal range. Harbor seal pups can swim immediately after birth, so the fraction available for counting likely varies according to conditions. In the Limfjord, a separate harbor seal population is living in a large inlet in northern Denmark where >95% of pups are born at two haul‐out sites situated 4 km apart. Drone surveys of these two haul‐outs were conducted throughout the breeding seasons during 2017–2019. The effects of weather conditions and timing of surveys on counts were analyzed by generalized additive mixed models. The best‐performing model included date and wind speed. Pup counts grew from the onset of the observation period and were predicted to peak on June 22. Wind speed had a linear relationship with pup counts and > 50% more pups were counted at 5 m/s than at 1 m/s. To achieve representative pup counts careful consideration of survey conditions and possibly correction for these effects are necessary.

Spatial variation on the abundance of a threatened South American large herbivore using spatiotemporally replicated drone surveys

Spatial variation on the abundance of a threatened South American large herbivore using spatiotemporally replicated drone surveys

Use of Audiovisual Methods and Documentary Film for the Preservation and Reappraisal of the Vernacular Architectural Heritage of the State of Michoacan, Mexico

The State of Michoacan has a wide and rich architectural heritage which is represented by different vernacular typologies which employ different materials and techniques according to the local resources. However, this wide knowledge and cultural heritage are disappearing, with several monuments and buildings being abandoned due to new dynamics and social changes. The traditional techniques have been substituted for modern and industrial construction systems since they are associated with obsolete ways of life. Through international cooperation for development and research projects, many activities were carried out to analyze the durability of local materials and vernacular constructions of the state. The main ones included the architectural survey and inspection of these buildings, photographic and drone surveys, characterization in the laboratory and oral interviews with local users. During the field work, these interviews with the local population, the creation of multimedia content and the use of audiovisual methods had crucial importance in the documentation of this local knowledge and culture, being great methods for scientific communication and dissemination. The main product is the documentary Xirangua, which means ‘roots’ in the Purépecha language, the main indigenous ethnicity in the state. This documentary explores the traditional architecture and construction of Michoacan, showing the reality within the loss of this vernacular heritage and portraying the importance of its conservation through the perspective of the local population, users and artisans. The use of these audiovisual techniques has demonstrated to be a suitable option to promote and preserve cultural heritage and educate, include the local actors in the process, and make the general public aware of the current issues, with this being one of the first works which explores this approach.

How radar technology will cut the cost and duration of aerial drone surveys

Paul Lindup, chief unmanned aerial vehicle pilot at Jacobs, says civil engineers will be able to capture large linear data sets much quicker and more cheaply when the Civil Aviation Authority permits its new radar-equipped survey drones to fly beyond their operators’ line of sight.

A pilot study to estimate the population size of endangered Galápagos marine iguanas using drones

BackgroundLarge-scale species monitoring remains a significant conservation challenge. Given the ongoing biodiversity crisis, the need for reliable and efficient methods has never been greater. Drone-based techniques have much to offer in this regard: they allow access to otherwise unreachable areas and enable the rapid collection of non-invasive field data. Herein, we describe the development of a drone-based method for the estimation of population size in Galápagos marine iguanas, Amblyrhynchus cristatus. As a large-bodied lizard that occurs in open coastal terrain, this endemic species is an ideal candidate for drone surveys. Almost all Amblyrhynchus subspecies are Endangered or Critically Endangered according to the IUCN yet since several colonies are inaccessible by foot, ground- based methods are unable to address the critical need for better census data. In order to establish a drone-based approach to estimate population size of marine iguanas, we surveyed in January 2021 four colonies on three focal islands (San Cristobal, Santa Fe and Espanola) using three techniques: simple counts (the standard method currently used by conservation managers), capture mark-resight (CMR), and drone-based counts. The surveys were performed within a 4-day window under similar ambient conditions. We then compared the approaches in terms of feasibility, outcome and effort.ResultsThe highest population-size estimates were obtained using CMR, and drone-based counts were on average 14% closer to CMR estimates—and 17–35% higher—than those obtained by simple counts. In terms of field-time, drone-surveys can be faster than simple counts, but image analyses were highly time consuming.ConclusionThough CMR likely produces superior estimates, it cannot be performed in most cases due to lack of access and knowledge regarding colonies. Drone-based surveys outperformed ground-based simple counts in terms of outcome and this approach is therefore suitable for use across the range of the species. Moreover, the aerial approach is currently the only credible solution for accessing and surveying marine iguanas at highly remote colonies. The application of citizen science and other aids such as machine learning will alleviate the issue regarding time needed to analyze the images.

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.

Optimally designing drone‐based surveys for wildlife abundance estimation with N‐mixture models

Abstract Hierarchical N‐mixture models have been suggested for abundance estimation from spatiotemporally replicated drone‐based count surveys, since they allow modeling abundance of unmarked individuals while accounting for detection errors. However, it is still necessary to understand how these models perform in the wide variety of contexts and species in which drone surveys are being used. This knowledge is fundamental to plan study designs with optimal allocation of scarce resources in ecology and conservation. We conduct a simulation study to address N‐mixture model (binomial and multinomial) performance and optimal survey effort allocation in different scenarios of local abundance and detectability of individuals, focusing on their application for drone‐based surveys. We also investigate the benefits of using a double‐observer protocol (either human or algorithm) in image review to decompose the detection process in availability and perception. Finally, we illustrate our simulation‐based survey design considerations by applying them to abundance estimation of marsh deer in the Pantanal wetland (Brazil). Accuracy of abundance estimation with N‐mixture models increases with local abundance in sites and especially with the availability of individuals. The optimal design requires more visits at fewer sites when the availability probability is lower, and the optimal design is more flexible as local abundance increases. Two observers checking images can increase the estimator performance even at very high perception probabilities. We quantified how much the use of a double‐observer protocol in image review can reduce fieldwork effort while achieving the same accuracy. N‐mixture models can deliver accurate abundance estimates from spatiotemporally replicated drone surveys in a wide variety of contexts while accounting for imperfect detection. The improvements achieved by a consciously planned design, rearranging survey efforts among sites and visits, as well as using a second observer in image review, can be crucial to detect trends when monitoring a population or to categorize a species as threatened or not.

Pedigree‐management‐flight interaction for temporal phenotype analysis and temporal phenomic prediction

AbstractUnoccupied aerial systems (UAS, aka drones) provide high dimensional temporal phenotype data for predictive plant breeding and genetic dissection. Methods to assess temporal phenotype data are an emerging need to predict temporal breeding values of genotypes. Here a novel interaction design was developed and evaluated to include drone flight dates as a component into the mixed model; allowing the temporal changes of drone image derived traits of maize hybrids across different flight dates as well as different management conditions to be monitored. Across 2017 and 2019 respectively, 228 and 100 maize hybrids were grown under two types of management (optimal and late plantings). Seven drone surveys were conducted over each management in 2017 while five drone surveys were conducted over each management in 2019. Temporal plant height (canopy height measurements, CHM) and normalized green‐red difference index (NGRDI) were extracted from each drone survey and used as phenotype data to evaluate the interaction design. Day of flight effects explained the highest amount of total variation for grain yield in the interaction model, meaning the majority of phenotypic variation of CHM and NGRDI occurred across growth with a unique temporal trajectory in each management system. Temporal repeatability values remained higher than 0.5 for CHM and NGRDI in each year. Temporal CHM and NGRDI breeding values of maize hybrids were combined in ridge and lasso regression prediction models. Yield prediction ability of untested genotypes in untested environments were predicted higher by using pedigree × management × flight (PMF) and pedigree× management (PM) interaction results (∼0.34 and 0.52 in 2017 and 2019). Combining environment specific phenomic data (PMF plus PM) gave a larger improvement in yield prediction when the tested and untested environments were less similar. Overall, combined temporal phenomic data could moderately predict grain yield under the most challenging predictive breeding scenario, untested and unrelated genotypes in untested environments.

Satellite Imaging of Methane Super-Emitters To Provide Data To Clean Up Supply Chains

_ The use of satellite imaging to verify self-reporting of methane emissions using empirical data gathered in near-real time by artificial intelligence, could cost the fossil fuel industry dearly in fines under the new methane provisions of the US Inflation Reduction Act (IRA). Signed into law in August, the IRA requires the US Environmental Protection Agency (EPA) to adopt within 2 years methods to monitor and collect empirical data on methane emissions. The act introduces the federal government’s first-ever tax on greenhouse gas emissions, but does not specify preferred technologies. Satellite monitoring is assumed to be at the top of the list, according to geo-analytics company Kayrros. Private industry can, and likely will, play the same game, as it has access to the very technologies the government uses and can invest to acquire empirical data to identify problems companies might not even know they have. Once emissions are quantified at their source, industry can clean up supply chains and thus blunt the financial impact of penalties set to debut in 2024. Third-Party Audits: Good for ESG Optics but Missing the Bigger Issue In May, ExxonMobil announced it had become the first oil and gas company to earn an “A” grade certification under the independent MiQ standard for managing methane emissions from associated gas at its Poker Lake, New Mexico, facilities which the company said in 2021 accounted for more than 10% of its Permian Basin gas production. Responsible Energy Solutions, a Texas-based independent auditor in the field of Responsibly Sourced Gas (RSG), conducted the study. One month later, Chevron announced it had earned top marks from Denver-based Project Canary, an SaaS-based (software-as-a-service) data analytics company organized as a public benefit corporation which markets itself as the gold standard for independent RSG audits. It uses its own monitoring devices and does not run data through the client’s SCADA system. Gas producers are turning to third-party audits to differentiate their production streams as RSG to comply with the investment community’s ESG (environment, social, and corporate governance) requirements. In a study issued a year ago, the global research and business consultancy Wood Mackenzie referred to RSG certification as a “nascent” industry (audit companies tend to be innovation-driven data analytic startups) and noted that “certification processes exist (currently) only for upstream assets and there is no universal or industry standard.” “Certification is based on standards such as air emissions, water stewardship, land use, and community impacts,” Wood Mackenzie reported. “RSG differs from normally produced natural gas in that producers take extra steps to reduce their carbon footprint, mitigate emissions, and minimize environmental and social impacts.” RSG certification monitoring might collect empirical data—which is a start—but it is largely at pad level, ignoring the bigger and more ominous picture that can be assessed only from space. That bigger picture is what the EPA will turn to when it begins to levy its methane tax in 2024 and it is why the industry is starting to look “up” as well. Drone surveys are widening the view. A paper presented at the 2021 SPE Annual Technical Conference and Exhibition (ATCE) detailed how BP conducted the first methane emissions survey of an offshore facility with a miniature methane spectrometer onboard a fixed-wing unmanned aerial vehicle (SPE 206181).

Use of UAVs and Deep Learning for Beach Litter Monitoring

Stranded beach litter is a ubiquitous issue. Manual monitoring and retrieval can be cost and labour intensive. Therefore, automatic litter monitoring and retrieval is an essential mitigation strategy. In this paper, we present important foundational blocks that can be expanded into an autonomous monitoring-and-retrieval pipeline based on drone surveys and object detection using deep learning. Drone footage collected on the islands of Malta and Gozo in Sicily (Italy) and the Red Sea coast was combined with publicly available litter datasets and used to train an object detection algorithm (YOLOv5) to detect litter objects in footage recorded during drone surveys. Across all classes of litter objects, the 50%–95% mean average precision (mAP50-95) was 0.252, with the performance on single well-represented classes reaching up to 0.674. We also present an approach to geolocate objects detected by the algorithm, assigning latitude and longitude coordinates to each detection. In combination with beach morphology information derived from digital elevation models (DEMs) for path finding and identifying inaccessible areas for an autonomous litter retrieval robot, this research provides important building blocks for an automated monitoring-and-retrieval pipeline.