Effect of Textural Features for Landcover Classification of Uav Multispectral Imagery of a Salt Marsh Restoration Site (original) (raw)
Related papers
Remote Sensing, 2024
Monitoring salt marshes with remote sensing is necessary to evaluate their state and restoration. Determining appropriate techniques for this can be overwhelming. Our study provides insight into whether a pixel-or object-based Random Forest classification approach is best for mapping vegetation in north temperate salt marshes. We used input variables from drone images (raw reflectances, vegetation indices, and textural features) acquired in June, July, and August 2021 of a salt marsh restoration and reference site in Aulac, New Brunswick, Canada. We also investigated the importance of input variables and whether using landcover classes representing areas of change was a practical way to evaluate variation in the monthly images. Our results indicated that (1) the classifiers achieved overall validation accuracies of 91.1-95.2%; (2) pixel-based classifiers outperformed object-based classifiers by 1.3-2.0%; (3) input variables extracted from the August images were more important than those extracted from the June and July images; (4) certain raw reflectances, vegetation indices, and textural features were among the most important variables; and (5) classes that changed temporally were mapped with user's and producer's validation accuracies of 86.7-100.0%. Knowledge gained during this study will inform assessments of salt marsh restoration trajectories spanning multiple years.
The purpose of this project was to evaluate and compare methodologies for using remotely sensed hyperspatial imagery from both Unmanned Aircraft Systems (UAS) and nextgeneration satellite technology to calculate species composition and ecosystem service metrics in a coastal intermediate marsh. Such methods would be important steps towards comprehensive monitoring of wetland landscapes and would provide useful metrics to study wetland condition and response to ecosystem restoration and disturbance events. BACKGROUND: Plant species composition, cover, density, and biomass are structural components of coastal marshes that are commonly used to quantify vegetative characteristics and often serve as indicators of wetland condition (Chamberlain and Ingram 2012; Cretini et al. 2012). Historically, regional and coastwide surveys to map coastal vegetation have consisted of laborious traversing of wetland landscapes, including time consuming and subjective ocular estimates of species type and cover (O'Neil 1949; Chabreck and Linscombe 1978; Sasser et al. 2014). Although recent surveys, like those in coastal Louisiana, enlist the use of helicopters for transport between, and hovering over, sampling sites, they continue to rely on ocular estimates of dominant species and abundance (i.e., Braun-Blanquet cover scale), they are costly, and are typically reproduced approximately every ten years. Rapid species classification (even of dominant plants) using remotely sensed data would provide many advantages over traditional field techniques.
Eelgrass Bed Mapping with Multispectral Uav Imagery in Atlantic Canada
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Eelgrass (Zostera marina L.) is a marine angiosperm that grows throughout coastal regions in Atlantic Canada. This study aimed to assess the capability of UAV multispectral imagery to map the presence of eelgrass beds within two estuaries in Atlantic Canada (Souris River and Richibucto River). The images were mosaicked using Agisoft and calibrated in reflectance. The corrected images were classified using a non-parametric supervised classifier (Random Forests). The input features of the classification were the UAV band reflectance and associated bathymetric ratios and vegetation indices. The resulting maps were compared with sonar data. The overall validation accuracy for presence/absence was 91.30% with the Souris image and 86.92%% with the Richibucto images. The limitations of the study are also presented.
Fine scale plant community assessment in coastal meadows using UAV based multispectral data
Ecological Indicators, 2020
Highlights:-Plant communities in coastal wetlands are at risk due to the impacts of global change-Knowing the distribution of plant communities is essential for nature conservation-Communities distribution maps were produced using a UAV-based multispectral sensor-The Random Forest classifier yielded the highest classification accuracy-Species diversity and aboveground biomass affect the classification performance ABSTRACT Coastal meadows worldwide are subjected to habitat degradation due to abandonment, intensification and the impacts of global change. In order to protect and restore these habitats and ensure the supply of valuable ecosystem services, it is necessary to know the extent and location of plant communities in coastal meadows. In this study, five plant communities were mapped at very high resolution in three different study sites in West Estonia. A fixed wing UAV was used to obtain multispectral images and derive a set of vegetation indices. Two different image classification techniques were used to cluster the vegetation indices maps and produce plant community distribution maps. The highest classification accuracy was obtained using a Random Forest classifier and 13 vegetation indices. Additionally, the spectral characteristics of the training samples were correlated with aboveground biomass and species diversity. Both biomass and species diversity were positively correlated with the spectral diversity of training samples and are thus likely to have an effect on the classification accuracy. The results of this study highlight the need to utilize a wide array of vegetation indices and assess the spectral characteristics of training samples in order to obtain high classification accuracies and understand the nature of misclassification errors. The resulting maps provide a solid foundation for global change impact assessment and habitat management and restoration in coastal meadows.
2018
Une analyse des systèmes aériens multispectraux téléguidés pour la classification de la couverture terrestre et la génération numérique de modèles d'élévation dans le cas de marais estuariens Par Logan D. Horrocks Les progrès récents dans les systèmes aériens téléguidés (UAS) et leur accessibilité croissance ont favorisé leur utilisation au sein de la communauté scientifique. Malgré ces innovations, les UAS tentent de cartographier l'élévation de la surface d'un site obstrué par la végétation à partir du logiciel « Motion Multi-View Stereo » (SFM-MVS); aboutissant à la création d'un Modèle Numérique d'Élévation (MNE) au lieu du Modèle Numérique de Terrain (MNT) souhaité. Ce projet vise à quantifier les différentes hauteurs des communautés végétales dans le marais salé de Masstown Est, en produisant des MNE pour les paysages de marais/vasières avec une précision comparable à celle des MNE. La génération des MNT a été réalisée à partir de deux étapes distinctes. La première étape repose sur des classifications de la couverture terrestre à partir de données UAS dérivées, corrigées radiométriquement et géométriquement. Les classifications de la couverture terrestre sont évaluées à partir de matrices de confusion. Dans un second temps, les hauteurs de la canopée étudiée et les hauteurs dérivées de la fonction sont soustraites de leurs classes respectives, générant les MNT. La validation des MNT a été réalisée en comparant les valeurs des relevés topographiques avec les valeurs modélisées, en utilisant la mesure de l'erreur quadratique moyenne (EQM). Le projet compare ensuite les différents paramètres mis en place pour les classifications de la couverture terrestre et la précision du MNT. Les méthodes de génération de MNT ont ensuite été couplées pour produire un MNT final avec une EQM de 6 cm. Les résultats indiquent que les UAS multispectraux « Grand public » peuvent produire des MNT avec des précisions comparables aux MNE initiaux générés, et mériteraient ainsi d'autres études quant à leur valeur scientifique. Le 11 Avril, 2018 iv ACKNOWLEDGEMENTS This project owes many thanks to many people, without whom this pursuit would not have been feasible. I would like to thank my supervisor Dr. Danika van Proosdij for her insightful feedback on a weekly basis which crafted the project to its current state, and my second reader Dr. Philip Giles for his comprehensive comments and advice. The completion of this project owes many thanks to Greg Baker for his consultations, without whom the flights would have been impossible to complete. A big thank-you to Jennie Graham for her guidance in the vegetation survey, and Graham Matheson for his guidance in the topographic survey. A big thanks to Sam, Reyhan, and Larissa as well for their help completing the vegetation surveys, and to Freddie Jacks for all her help with edits. I owe countless thanks to my Dad and Mom for everything that's allowed for me to reach this point. A final thanks to Carl, Dylan, and all my friends for keeping me sane in this period. v
International Journal of Remote Sensing, 2014
Mapping of patterns and spatial distribution of land-use/cover (LULC) has long been based on remotely sensed data. In the recent past, efforts to improve the reliability of LULC maps have seen a proliferation of image classification techniques. Despite these efforts, derived LULC maps are still often judged to be of insufficient quality for operational applications, due to disagreement between generated maps and reference data. In this study we sought to pursue two objectives: first, to test the new-generation multispectral RapidEye imagery classification output using machine-learning random forest (RF) and support vector machines (SVM) classifiers in a heterogeneous coastal landscape; and second, to determine the importance of different RapidEye bands on classification output. Accuracy of the derived thematic maps was assessed by computing confusion matrices of the classifiers’ cover maps with respective independent validation data sets. An overall classification accuracy of 93.07% with a kappa value of 0.92, and 91.80 with a kappa value of 0.92 was achieved using RF and SVM, respectively. In this study, RF and SVM classifiers performed comparatively similarly as demonstrated by the results of McNemer’s test (Z = 1.15). An evaluation of different RapidEye bands using the two classifiers showed that incorporation of the red-edge band has a significant effect on the overall classification accuracy in vegetation cover types. Consequently, pursuit of high classification accuracy using high-spatial resolution imagery on complex landscapes remains paramount.
Remote Sensing, 2018
Efforts are increasingly being made to classify the world's wetland resources, an important ecosystem and habitat that is diminishing in abundance. There are multiple remote sensing classification methods, including a suite of nonparametric classifiers such as decision-tree (DT), rule-based (RB), and random forest (RF). High-resolution satellite imagery can provide more specificity to the classified end product, and ancillary data layers such as the Normalized Difference Vegetation Index, and hydrogeomorphic layers such as distance-to-a-stream can be coupled to improve overall accuracy (OA) in wetland studies. In this paper, we contrast three nonparametric machine-learning algorithms (DT, RB, and RF) using a large field-based dataset (n = 228) from the Selenga River Delta of Lake Baikal, Russia. We also explore the use of ancillary data layers selected to improve OA, with a goal of providing end users with a recommended classifier to use and the most parsimonious suite of input parameters for classifying wetland-dominated landscapes. Though all classifiers appeared suitable, the RF classification outperformed both the DT and RB methods, achieving OA >81%. Including a texture metric (homogeneity) substantially improved the classification OA. However, including vegetation/soil/water metrics (based on WorldView-2 band combinations), hydrogeomorphic data layers, and elevation data layers to increase the descriptive content of the input parameters surprisingly did not markedly improve the OA. We conclude that, in most cases, RF should be the classifier of choice. The potential exception to this recommendation is under the circumstance where the end user requires narrative rules to best manage his or her resource. Though not useful in this study, continuously increasing satellite imagery resolution and band availability suggests the inclusion of ancillary contextual data layers such as soil metrics or elevation data, the granularity of which may define its utility in subsequent wetland classifications.
Billions of dollars are projected to be spent on restoration projects along the northern Gulf Coast which will require efficient monitoring at both landscape and project-specific scales. Recent developments in unmanned aircraft systems (UAS) have sparked interest in the ability of these "drones" to capture hyperspatial imagery (pixel resolution < 10 cm) that resolves individual species and produces accurate data for monitoring programs in coastal landscapes. We present a case study conducted at Coastwide Reference Monitoring System (CRMS) station 0392, a Spartina patens-dominated, oligohaline coastal marsh in Terrebonne Parish, Louisiana. Results demonstrate the ability of UAS technology to collect hyperspatial, multispectral aerial images in a coastal wetland, and to produce very-high-resolution orthomosaics and digital elevation models. We then used object-based image analysis (OBIA) techniques to (1) delineate the land-water interface, (2) classify composition by dominant species, and (3) quantify average plant height by species. Model results were validated with traditional on-the-ground CRMS vegetation surveys. Results suggest that OBIA methods can overcome the spectral variability of hyperspatial datasets, quantify uncertainties in conventional techniques, and provide improved estimates of wetland vegetation cover and species composition. These methods scale conventional plot-level coverage values to data-rich landscape-level models and provide useful tools to monitor restoration performance, landscape changes, and ecosystem services in coastal wetland systems.
ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Vegetation mapping, identifying the distribution of plant species, is important for analysing vegetation dynamics, quantifying spatial patterns of vegetation evolution, analysing the effects of environment changes on vegetation, and predicting spatial patterns of species diversity. Such analysis can contribute to the development of targeted land management actions that maintain biodiversity and ecological functions. This paper represents a methodology for 3D vegetation mapping of a coastal dune complex using a multispectral camera mounted on an Unmanned Aerial System (UAS) with particular reference to the Buckroney dune complex in Co. Wicklow, Ireland. UAS, also known as Unmanned Aerial Vehicles (UAV's) or drones, have enabled high-resolution and highaccuracy ground-based data to be gathered quickly and easily on-site. The Sequoia multispectral camera used in this study has green, red, red-edge and near infrared wavebands, and a normal RGB camera, to capture both visible and NIR images of the land surface. The workflow of 3D vegetation mapping of the study site included establishing ground control points, planning the flight mission and camera parameters, acquiring the imagery, processing the image data and performing features classification. The data processing outcomes include an orthomosiac model, a 3D surface model and multispectral images of the study site, in the Irish Transverse Mercator coordinate system, with a planimetric resolution of 0.024 m and a georeferenced Root-Mean-Square (RMS) error of 0.111 m. There were 235 sample area (1 m × 1 m) used for the accuracy assessment of the classification of the vegetation mapping. Feature classification was conducted using three different classification strategies to examine the efficiency of multispectral sensor data for vegetation mapping. Vegetation type classification accuracies ranged from 60% to 70%. This research illustrates the efficiency of data collection at Buckroney dune complex and the high-accuracy and high-resolution of the vegetation mapping of the site using a multispectral sensor mounted on UAS.
Aerial Image Classification for the Mapping of Riparian Vegetation Habitats
Acta Silvatica et Lignaria Hungarica, 2013
In the current study, aerial image analysis has been applied to map vegetation communities in a riparian wetland ecosystem, Szigetköz (Hungary). Remote sensing offers an objective and timeeffective method for the detection of detailed vegetation habitats with the use of high resolution aerial photos combined with ancillary botanical and silvicultural data. Three images of the same test site, acquired in three different years have been analysed by sample-based semi-automated image classification technique. Due to the heterogeneous nature of the target vegetation classes, besides using spectral features (e.g. vegetation indices) textural descriptors were also involved in the classification procedure. The most appropriate parameters have been chosen using a statistical feature selection method based on the Jeffries-Matusita distance. The accuracy assessment proved for each scene that the combined use of spectral and textural features gave the best classification results in comparison t...