Detecting plant species in the field with deep learning and drone technology

James, Katherine; Bradshaw, Karen

doi:10.1111/2041-210x.13473

Cited by 41 publications

(23 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In principle, remote sensing of hyperspectral, LIDAR and/or high-resolution photogrammetric data provides information that enables the identification of tree species (e.g. Kattenborn et al 2019;Cao et al 2020;James and Bradshaw 2020). Yet the remote identification of orchard tree species is relatively challenging since most of these species are closely related (belonging to the Rosaceae family) and thus phenotypically rather similar.…”

Section: Research Perspectivesmentioning

confidence: 99%

The value of remotely sensed vs. field-surveyed habitat structure for predicting bird abundance: a case study in traditional orchards

2022

View full text Add to dashboard Cite

Understanding environmental effects on the distribution and abundance of species is central to ecology, biogeography and evolutionary biology. This led to the development of species distribution models (SDMs) that relate spatial variation in occurrence and abundance to environmental variables. So far, SDMs rarely considered habitat structure, as a major determinant of bird distributions. While remote sensing increasingly provides high-resolution measures of habitat structure, certain structural variables affecting bird abundance still need to be measured with field surveys. In this study, we compare the value of remotely sensed vs. field-surveyed habitat structure for predicting bird abundance. Specifically, we analysed abundance data for nine bird species of traditional orchards in South-Western Germany. ‘Remote sensing SDMs’ related abundance to structural variables obtained by aerial photogrammetry of individual orchard trees. Alternative ‘field survey SDMs’ related bird abundance to detailed field surveys of the species composition and pruning state of orchard trees. Additionally, both remote sensing and field survey SDMs included climate and land use variables. Accounting for detailed habitat structure improved abundance predictions for seven of nine study species compared to models only incorporating climate and land use. The impact on model performance differed between remotely sensed and field-surveyed variables: the former improved abundance models for most (n = 7) bird species, whereas the latter had more variable impact, decreasing model performance for five species. The remotely sensed variable with strongest effects was overall tree density, which positively affected abundance of seven species. In contrast, multiple field-surveyed variables had similar effect strength, with the overall strongest effect found for pear tree density, to which seven bird species showed a unimodal response. These analyses have conservation implications since they predict expected responses of bird species to ongoing changes in orchard structure. Moreover, they identify structural variables that will be most promising to measure via remote sensing data in the future.

show abstract

Section: Research Perspectivesmentioning

confidence: 99%

The value of remotely sensed vs. field-surveyed habitat structure for predicting bird abundance: a case study in traditional orchards

2022

View full text Add to dashboard Cite

show abstract

“…Carlier et al [18] presented an application of morphological image analysis to provide an objective method for detection and accurate cover assessment of an IAPS. They used top-down images captured using a hand-held digital camera with images that cover 1 m × 1 m. James and Bradshaw [19] based their work on images collected using UAVs, which allowed them to cover larger areas, but instead of using morphological image analysis, James and Bradshaw [19] used the U-net convolutional neural network to segment images semantically. The use of a convolutional neural network makes the method less affected by changes in light intensity and shadows and makes it applicable in images with plant occlusion.…”

Section: Introductionmentioning

confidence: 99%

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

Dyrmann

Mortensen

Linneberg³

et al. 2021

Sensors

View full text Add to dashboard Cite

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

show abstract

“…We emphasize our approach for building a deep learning model in an effort to provide a beginning road map to aid conservation researchers considering the use of AI for drone-based census of other plant species. Many studies have used drones along with deep learning models to collect data in agriculture [10][11][12][13], and there are publications describing this approach for a variety of wild organisms, including plant species in general [14][15][16][17][18][19] and especially invasive species [20][21][22][23]. Interest in using drone imagery as a tool in rare plant conservation is increasing [14], but to our knowledge no published studies to date have successfully applied a deep learning approach to drone-acquired imagery with the goal of enumerating individuals of a rare plant species.…”

Section: Introductionmentioning

confidence: 99%

Drones, Deep Learning, and Endangered Plants: A Method for Population-Level Census Using Image Analysis

Rominger

Meyer

2021

Drones

View full text Add to dashboard Cite

A census of endangered plant populations is critical to determining their size, spatial distribution, and geographical extent. Traditional, on-the-ground methods for collecting census data are labor-intensive, time-consuming, and expensive. Use of drone imagery coupled with application of rapidly advancing deep learning technology could greatly reduce the effort and cost of collecting and analyzing population-level data across relatively large areas. We used a customization of the YOLOv5 object detection model to identify and count individual dwarf bear poppy (Arctomecon humilis) plants in drone imagery obtained at 40 m altitude. We compared human-based and model-based detection at 40 m on n = 11 test plots for two areas that differed in image quality. The model out-performed human visual poppy detection for precision and recall, and was 1100× faster at inference/evaluation on the test plots. Model inference precision was 0.83, and recall was 0.74, while human evaluation resulted in precision of 0.67, and recall of 0.71. Both model and human performance were better in the area with higher-quality imagery, suggesting that image quality is a primary factor limiting model performance. Evaluation of drone-based census imagery from the 255 ha Webb Hill population with our customized YOLOv5 model was completed in <3 h and provided a reasonable estimate of population size (7414 poppies) with minimal investment of on-the-ground resources.

show abstract

Detecting plant species in the field with deep learning and drone technology

Abstract: 1. Aerial drones are providing a new source of high-resolution imagery for mapping of plant species of interest, amongst other applications. On-board detection al-How to cite this article: James K, Bradshaw K. Detecting plant species in the field with deep learning and drone technology.

Cited by 41 publications

References 31 publications

The value of remotely sensed vs. field-surveyed habitat structure for predicting bird abundance: a case study in traditional orchards

The value of remotely sensed vs. field-surveyed habitat structure for predicting bird abundance: a case study in traditional orchards

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

Drones, Deep Learning, and Endangered Plants: A Method for Population-Level Census Using Image Analysis

Contact Info

Product

Resources

About