Multi-temporal Unmanned Aerial Vehicle Research Articles

UAV-Derived Data Application for Environmental Monitoring of the Coastal Area of Lake Sevan, Armenia with a Changing Water Level

The paper presents the range and applications of thematic tasks for ultra-high spatial resolution data from small unmanned aerial vehicles (UAVs) in the integral system of environmental multi-platform and multi-scaled monitoring of Lake Sevan, which is one of the greatest freshwater lakes in Eurasia. From the 1930s, it had been subjected to human-driven changing of the water level with associated and currently exacerbated environmental issues. We elaborated the specific techniques of optical and thermal surveys for the different coastal sites and phenomena in study. UAV-derived optical imagery and thermal stream were processed by a Structure-from-Motion algorithm to create digital surface models (DSMs) and ortho-imagery for several key sites. UAV imagery were used as additional sources of detailed spatial data under large-scale mapping of current land-use and point sources of water pollution in the coastal zone, and a main data source on environmental violations, especially sewage discharge or illegal landfills. The revealed present-day coastal types were mapped at a large scale, and the net changes of shoreline position and rates of shore erosion were calculated on multi-temporal UAV data using modified Hausdorff’s distance. Based on highly-detailed DSMs, we revealed the areas and objects at risk of flooding under the projected water level rise to 1903.5 m along the west coasts of Minor Sevan being the most popular recreational area. We indicated that the structural and environmental state of marsh coasts and coastal wetlands as potential sources of lake eutrophication and associated algal blooms could be more efficiently studied under thermal UAV surveys than optical ones. We proposed to consider UAV surveys as a necessary intermediary between ground data and satellite imagery with different spatial resolutions for the complex environmental monitoring of the coastal area and water body of Lake Sevan as a whole.

UAV-based evaluation of morphological changes induced by extreme rainfall events in meandering rivers.

Morphological changes, caused by the erosion and deposition processes due to water discharge and sediment flux occur, in the banks along the river channels and in the estuaries. Flow rate is one of the most important factors that can change river morphology. The geometric shapes of the meanders and the river flow parameters are crucial components in the areas where erosion or deposition occurs in the meandering rivers. Extreme precipitation triggers erosion on the slopes, which causes significant morphological changes in large areas during and after the event. The flow and sediment amount observed in a river basin with extreme precipitation increases and exceeds the long-term average value. Hereby, erosion severity can be determined by performing spatial analyses on remotely sensed imagery acquired before and after an extreme precipitation event. Changes of erosion and deposition along the river channels and overspill channels can be examined by comparing multi-temporal Unmanned Aerial Vehicle (UAV) based Digital Surface Model (DSM) data. In this study, morphological changes in the Büyük Menderes River located in the western Turkey, were monitored with pre-flood (June 2018), during flood (January 2019), and post-flood (September 2019) UAV surveys, and the spatial and volumetric changes of eroded/deposited sediment were quantified. For this purpose, the DSAS (Digital Shoreline Analysis System) method and the DEM of Difference (DoD) method were used to determine the changes on the riverbank and to compare the periodic volumetric morphological changes. Hereby, Structure from Motion (SfM) photogrammetry technique was exploited to a low-cost UAV derived imagery to achieve riverbank, areal and volumetric changes following the extreme rainfall events extracted from the time series of Tropical Rainfall Measuring Mission (TRMM) satellite data. The change analyses were performed to figure out the periodic morphodynamic variations and the impact of the flood on the selected meandering structures. In conclusion, although the river water level increased by 0.4–5.9 meters with the flood occurred in January 2019, the sediment deposition areas reformed after the flood event, as the water level decreased. Two-year monitoring revealed that the sinuosity index (SI) values changed during the flood approached the pre-flood values over time. Moreover, it was observed that the amount of the deposited sediments in September 2019 approached that of June 2018.

Multi-temporal UAV based repeat monitoring of rivers sensitive to flood

ABSTRACT Multi-temporal repeat monitoring of flood-vulnerable rivers is crucial due to rapid alteration of morphological properties of in-channel landforms. Besides, the characteristics of the river crossing bridges may deteriorate due to flood induced scouring around bridge piles or due to flood loads. Thus, in this study high-resolution topography of the study region was acquired during two consecutive years by unmanned aerial vehicle (UAV) based surveys using Structure-from-Motion (SfM) processing. Following the extraction of digital elevation models (DEM), repeat data that were obtained at each UAV survey were compared using Geomorphic Change Detection (GCD) to calculate volumes of deposition and erosion via DEM of difference (DoD) algorithm. Thus, detailed high-resolution maps of the river channels can be rapidly and efficiently generated by low cost UAV based measurement methods in order for continuous tracking of stream channel morphology for the rivers sensitive to floods.

Characterization of displacement and internal structure of landslides from multitemporal UAV and ERT imaging

Preliminary knowledge of the surface movement and internal structure of a landslide is essential to improve the understanding of landslide behavior as an initial step to develop suitable mitigation measures. This paper characterizes displacement and internal structure of landslides using multitemporal unmanned aerial vehicles (UAV) and electrical resistivity tomography (ERT) imaging. Multitemporal orthophotos, digital elevation model (DEM) profile lines, DEMs of difference (DoD), and ERT images allowed us to measure the landslide morphometry and areal changes, detect the trend of surface displacement, calculate the surface movement rate and direction, characterize the spatial variability in surface movement rates, and relate spatial variation in displacement to the internal structure of the landslide. The Kalisari Landslide experienced retrogressive movement where the scarp of landslide moved backward 38 m from 2015 to 2019. The surface movement ranged from 0.7 to 8.1 m with the direction mostly to the northeast. The depletion and accumulation ranged from 0.1 to 5 m, with the most active change located northwest of the landslide. The variability in movement was influenced by the spatial variability in the subsurface material inferred from the ERT images, i.e., surface material, clay, weathered breccias (possibly altered), and breccias. The interpretation from multitemporal UAV and ERT images indicates that the Kalisari Landslide has a nonhomogeneous multiple retrogressive behavior with a curved slip surface located at the clay layer.

Multi-Temporal Unmanned Aerial Vehicle Remote Sensing for Vegetable Mapping Using an Attention-Based Recurrent Convolutional Neural Network

Vegetable mapping from remote sensing imagery is important for precision agricultural activities such as automated pesticide spraying. Multi-temporal unmanned aerial vehicle (UAV) data has the merits of both very high spatial resolution and useful phenological information, which shows great potential for accurate vegetable classification, especially under complex and fragmented agricultural landscapes. In this study, an attention-based recurrent convolutional neural network (ARCNN) has been proposed for accurate vegetable mapping from multi-temporal UAV red-green-blue (RGB) imagery. The proposed model firstly utilizes a multi-scale deformable CNN to learn and extract rich spatial features from UAV data. Afterwards, the extracted features are fed into an attention-based recurrent neural network (RNN), from which the sequential dependency between multi-temporal features could be established. Finally, the aggregated spatial-temporal features are used to predict the vegetable category. Experimental results show that the proposed ARCNN yields a high performance with an overall accuracy of 92.80%. When compared with mono-temporal classification, the incorporation of multi-temporal UAV imagery could significantly boost the accuracy by 24.49% on average, which justifies the hypothesis that the low spectral resolution of RGB imagery could be compensated by the inclusion of multi-temporal observations. In addition, the attention-based RNN in this study outperforms other feature fusion methods such as feature-stacking. The deformable convolution operation also yields higher classification accuracy than that of a standard convolution unit. Results demonstrate that the ARCNN could provide an effective way for extracting and aggregating discriminative spatial-temporal features for vegetable mapping from multi-temporal UAV RGB imagery.

Detection of Road Surface Changes from Multi-Temporal Unmanned Aerial Vehicle Images Using a Convolutional Siamese Network

Road quality commonly decreases due to aging and deterioration of road surfaces. As the number of roads that need to be surveyed increases, general maintenance—particularly surveillance—can be quite costly if carried out using traditional methods. Therefore, using unmanned aerial vehicles (UAVs) and deep learning to detect changes via surveys is a promising strategy. This study proposes a method for detecting changes on road surfaces using pairs of UAV images captured at different times. First, a convolutional Siamese network is introduced to extract the features of an image pair and a Euclidean distance function is applied to calculate the distance between two features. Then, a contrastive loss function is used to enlarge the distance between changed feature pairs and reduce the distance between unchanged feature pairs. Finally, the initial change map is improved based on the preliminary differences between the two input images. Our experimental results confirm the effectiveness of this approach.

Multi-temporal UAV data for assessing rapid rill erosion in typical gully heads on the largest tableland of the Loess Plateau, China

Quantitative estimation of rill erosion research plays a significant role in the understanding of rill formation and their development mechanisms, which is lacking presently. In this paper, rill erosion is quantified to analyze and evaluate the spatial and temporal distribution in typical gully heads of the Loess Plateau of China. Although several studies have carried rill erosion quantification, the variability analysis of rill erosion rates at different slope position is limited primarily because of the unavailability of high resolution elevation models. Using the low-altitude photography with Unmanned Aerial Vehicles (UAVs), digital surface models (DSMs) with centimeter accuracy were generated, and then geometrical parameters of a total of 1,129 rills at tableland-slope (T-S), hillslope (S), and slopes affected by artificial vegetation (SV) in the study were counted. Then, spatial and temporal development as well as erosion rates of the rills at the three sites were compared. Although many studies revealed the erosion law at the S position, few studies have linked the development of rills at the T-S position, and artificial afforestation at the SV position exacerbates the rill erosion. Furthermore, based on the centimeter-resolution DSMs obtained by the two-stage UAV photography, we found that the slope gradient which was the most susceptible to rill erosion were 5-10°, 35-50°, and 65-80° at T-S, S and SV sites, respectively. At the SV, rill incision was the most sensitive to slope changes, and rill incision potential proved to be the greatest. Generally, the higher the rill erosion depth, the smaller the frequency of occurrence, but this trend was mitigated at the SV position. Therefore, urgent action is needed to control rill erosion caused by artificial afforestation on hillslopes to alleviate and eliminate the adverse effects of vegetation planting on rill erosion on the Loess Plateau of China.

Coral Reef Mapping of UAV: A Comparison of Sun Glint Correction Methods

Although methods were proposed for eliminating sun glint effects from airborne and satellite images over coral reef environments, a method was not proposed previously for unmanned aerial vehicle (UAV) image data. De-glinting in UAV image analysis may improve coral distribution mapping accuracy result compared with an uncorrected image classification technique. The objective of this research was to determine accuracy of coral reef habitat classification maps based on glint correction methods proposed by Lyzenga et al., Joyce, Hedley et al., and Goodman et al. The UAV imagery collected from the coral-dominated Pulau Bidong (Peninsular Malaysia) on 20 April 2016 was analyzed in this study. Images were pre-processed with the following two strategies: Strategy-1 was the glint removal technique applied to the whole image, while Strategy-2 used only the regions impacted by glint instead of the whole image. Accuracy measures for the glint corrected images showed that the method proposed by Lyzenga et al. following Strategy-2 could eliminate glints over the branching coral—Acropora (BC), tabulate coral—Acropora + Montipora (TC), patch coral (PC), coral rubble (R), and sand (S) with greater accuracy than the other four methods using Strategy-1. Tested in two different coral environments (Site-1: Pantai Pasir Cina and Site-2: Pantai Vietnam), the glint-removed UAV imagery produced reliable maps of coral habitat distribution with finer details. The proposed strategies can potentially be used to remove glint from UAV imagery and may improve usability of glint-affected imagery, for analyzing spatiotemporal changes of coral habitats from multi-temporal UAV imagery.

High-Throughput Phenotyping of Bioethanol Potential in Cereals Using UAV-Based Multi-Spectral Imagery.

Bioethanol production obtained from cereal straw has aroused great interest in recent years, which has led to the development of breeding programs to improve the quality of lignocellulosic material in terms of the biomass and sugar content. This process requires the analysis of genotype–phenotype relationships, and although genotyping tools are very advanced, phenotypic tools are not usually capable of satisfying the massive evaluation that is required to identify potential characters for bioethanol production in field trials. However, unmanned aerial vehicle (UAV) platforms have demonstrated their capacity for efficient and non-destructive acquisition of crop data with an application in high-throughput phenotyping. This work shows the first evaluation of UAV-based multi-spectral images for estimating bioethanol-related variables (total biomass dry weight, sugar release, and theoretical ethanol yield) of several accessions of wheat, barley, and triticale (234 cereal plots). The full procedure involved several stages: (1) the acquisition of multi-temporal UAV images by a six-band camera along different crop phenology stages (94, 104, 119, 130, 143, 161, and 175 days after sowing), (2) the generation of ortho-mosaicked images of the full field experiment, (3) the image analysis with an object-based (OBIA) algorithm and the calculation of vegetation indices (VIs), (4) the statistical analysis of spectral data and bioethanol-related variables to predict a UAV-based ranking of cereal accessions in terms of theoretical ethanol yield. The UAV-based system captured the high variability observed in the field trials over time. Three VIs created with visible wavebands and four VIs that incorporated the near-infrared (NIR) waveband were studied, obtaining that the NIR-based VIs were the best at estimating the crop biomass, while the visible-based VIs were suitable for estimating crop sugar release. The temporal factor was very helpful in achieving better estimations. The results that were obtained from single dates [i.e., temporal scenario 1 (TS-1)] were always less accurate for estimating the sugar release than those obtained in TS-2 (i.e., averaging the values of each VI obtained during plant anthesis) and less accurate for estimating the crop biomass and theoretical ethanol yield than those obtained in TS-3 (i.e., averaging the values of each VI obtained during full crop development). The highest correlation to theoretical ethanol yield was obtained with the normalized difference vegetation index (R2 = 0.66), which allowed to rank the cereal accessions in terms of potential for bioethanol production.

Comparison of Vegetation Indices Derived from UAV Data for Differentiation of Tillage Effects in Agriculture

Unmanned aerial vehicle (UAV) platforms with sensors covering the red-edge and near-infrared (NIR) bands to measure vegetation indices (VIs) have been recently introduced in agriculture research. Consequently, VIs originally developed for traditional airborne and spaceborne sensors have become applicable to UAV systems. In this study, we investigated the difference in tillage treatments for cotton and sorghum using various RGB and NIR VIs. Minimized tillage has been known to increase farm sustainability and potentially optimize productivity over time; however, repeated tillage is the most commonly-adopted management practice in agriculture. To this day, quantitative comparisons of plant growth patterns between conventional tillage (CT) and no tillage (NT) fields are often inconsistent. In this study, high-resolution and multi-temporal UAV data were used for the analysis of tillage effects on plant health and the performance of various vegetation indices investigated. Time series data over ten dates were acquired on a weekly basis by RGB and multispectral (MS) UAV platforms: a DJI Phantom 4 Pro and a DJI Matrice 100 with the SlantRange 3p sensor. Ground reflectance panels and an ambient illumination sensor were used for the radiometric calibration of RGB and MS orthomosaic images, respectively. Various RGB and NIR-based vegetation indices were then calculated for the comparison between CT and NT treatments. In addition, a one-tailed Z-test was conducted to check the significance of VIs’ difference between CT and NT treatments. The results showed distinct differences in VIs between tillage treatments during the whole growing season. NIR-based VIs showed better discrimination performance than RGB-based VIs. Out of 13 VIs, the modified soil adjusted vegetation index (MSAVI) and optimized soil adjusted vegetation index (OSAVI) showed better performance in terms of quantitative difference measurements and the Z-test between tillage treatments. The modified green red vegetation index (MGRVI) and excess green (ExG) showed reliable separability and can be an alternative for economic RGB UAV application.

Estimating Biomass of Black Oat Using UAV-Based RGB Imaging

The spatial and temporal variability of crop parameters are fundamental in precision agriculture. Remote sensing of crop canopy can provide important indications on the growth variability and help understand the complex factors influencing crop yield. Plant biomass is considered an important parameter for crop management and yield estimation, especially for grassland and cover crops. A recent approach introduced to model crop biomass consists in the use of RGB (red, green, blue) stereo images acquired from unmanned aerial vehicles (UAV) coupled with photogrammetric softwares to predict biomass through plant height (PHT) information. In this study, we generated prediction models for fresh (FBM) and dry biomass (DBM) of black oat crop based on multi-temporal UAV RGB imaging. Flight missions were carried during the growing season to obtain crop surface models (CSMs), with an additional flight before sowing to generate a digital terrain model (DTM). During each mission, 30 plots with a size of 0.25 m² were distributed across the field to carry ground measurements of PHT and biomass. Furthermore, estimation models were established based on PHT derived from CSMs and field measurements, which were later used to build prediction maps of FBM and DBM. The study demonstrates that UAV RGB imaging can precisely estimate canopy height (R2 = 0.68–0.92, RMSE = 0.019–0.037 m) during the growing period. FBM and DBM models using PHT derived from UAV imaging yielded R2 values between 0.69 and 0.94 when analyzing each mission individually, with best results during the flowering stage (R2 = 0.92–0.94). Robust models using datasets from different growth stages were built and tested using cross-validation, resulting in R2 values of 0.52 for FBM and 0.84 for DBM. Prediction maps of FBM and DBM yield were obtained using calibrated models applied to CSMs, resulting in a feasible way to illustrate the spatial and temporal variability of biomass. Altogether the results of the study demonstrate that UAV RGB imaging can be a useful tool to predict and explore the spatial and temporal variability of black oat biomass, with potential use in precision farming.

UAV/SATELLITE MULTISCALE DATA FUSION FOR CROP MONITORING AND EARLY STRESS DETECTION

Abstract. Early stress detection is critical for proactive field management and terminal yield prediction, and can aid policy making for improved food security in the context of climate change and population growth. Field surveys for crop monitoring are destructive, labor-intensive, time-consuming and not ideal for large-scale spatial and temporal monitoring. Recent technological advances in Unmanned Aerial Vehicle (UAV) and high-resolution satellite imaging with frequent revisit time have proliferated the applications of this emerging new technology in precision agriculture to address food security challenges from regional to global scales. In this paper, we present a concept of UAV and satellite virtual constellation to demonstrate the power of multi-scale imaging for crop monitoring. Low-cost sensors integrated on a UAV were used to collect RGB, multispectral, and thermal images during the growing season in a test site established near Columbia, Missouri, USA. WorldView-3 multispectral data were pan-sharpened, atmospherically corrected to reflectance and combined with UAV data for temporal monitoring of early stress. UAV thermal and multispectral data were calibrated to canopy temperature and reflectance following a rigorous georeferencing and ortho-correction. The results show that early stress can be effectively detected using multi-temporal and multi-scale UAV and satellite observation; the limitations of satellite remote sensing data in field-level crop monitoring can be overcome by using low altitude UAV observations addressing not just mixed pixel issues but also filling the temporal gap in satellite data availability enabling capture of early stress. The concept developed in this paper also provides a framework for accurate and robust estimation of plant traits and grain yield and delivers valuable insight for high spatial precision in high-throughput phenotyping and farm field management.

INTRASEASONAL VARIABILITY OF GUANO STAINS IN A REMOTELY SENSED PENGUIN COLONY USING UAV AND SATELLITE

Abstract. Remote sensing of penguins gives a unique opportunity to observe ecosystem changes in the Antarctic and the Southern Ocean at a continent-wide scale. The extent of guano is the best proxy to the size of penguin populations but frequent cloud cover limits the number of available images. This study focuses on the correlation between guano coverage visible in aerial and satellite images and breeding pair numbers in the course of the breeding seasons 2016/17 and 2017/18 in a colony of Pygocelid penguins on Ardley Island (South Shetland Islands, Antarctica). Multitemporal UAV (Unmanned Aerial Vehicle) orthomosaics and high-resolution satellite images were collected of Ardley Island as well as data on breeding phenology, weather conditions and snow coverage. “Fresh” guano stains were classified using different methods of Geographical Object-based Image Analysis (GEOBIA) and differentiated from weathered guano stains. Analysis of this data shows that guano stains in an Antarctic Pygoscelid penguin colony undergo significant intraseasonal changes in extent, texture and spectral signature. Hence, the timing of image acquisition and the advance of snow melt during Antarctic spring matter when determining penguin populations and should be considered during the analysis. Our results show changes of up to 25 % of the total guano covered surface due to individual weather events and changes up to 80 % in the time between the peak of egg laying and the occurrence of the first crèche.

Tidal and Meteorological Influences on the Growth of Invasive Spartina alterniflora: Evidence from UAV Remote Sensing

Rapid invasion of Spartina alterniflora into Chinese coastal wetlands has attracted much attention. Many field and remote sensing studies have examined the spatio-temporal dynamics of S. alterniflora invasion; however, spatially explicit quantitative analyses of S. alterniflora invasion and its underlying mechanisms at both patch and landscape scales are seldom reported. To fill this knowledge gap, we integrated multi-temporal unmanned aerial vehicle (UAV) imagery, light detection and ranging (LiDAR)-derived elevation data, and tidal and meteorological time series to explore the growth potential (lateral expansion rates and canopy greenness) of S. alterniflora over the intertidal zone in a subtropical coastal wetland (Zhangjiang estuarine wetland, Fujian, China). Our analyses of patch expansion indicated that isolated S. alterniflora patches in this wetland experienced high lateral expansion over the past several years (averaged at 4.28 m/year in patch diameter during 2014–2017), and lateral expansion rates ( y , m/year) showed a statistically significant declining trend with increasing inundation ( x , h/day; 3 ≤ x ≤ 18 ): y = − 0.17 x + 5.91 , R 2 = 0.78 . Our analyses of canopy greenness showed that the seasonality of the growth potential of S. alterniflora was driven by temperature (Pearson correlation coefficient r = 0.76 ) and precipitation ( r = 0.68 ), with the growth potential peaking in early/middle summer with high temperature and adequate precipitation. Together, we concluded that the growth potential of S. alterniflora was co-regulated by tidal and meteorological regimes, in which spatial heterogeneity is controlled by tidal inundation while temporal variation is controlled by both temperature and precipitation. To the best of our knowledge, this is the first spatially explicit quantitative study to examine the influences of tidal and meteorological regimes on both spatial heterogeneity (over the intertidal zone) and temporal variation (intra- and inter-annual) of S. alterniflora at both patch and landscape scales. These findings could serve critical empirical evidence to help answer how coastal salt marshes respond to climate change and assess the vulnerability and resilience of coastal salt marshes to rising sea level. Our UAV-based methodology could be applied to many types of plant community distributions.

Insights from the failure and dynamic characteristics of two sequential landslides at Baige village along the Jinsha River, China

The focus of this paper is on two large sequential landslides which formed a dam and the resulting lake that occurred along the Jinsha River on October 11 and November 3, 2018. About 24 and 9 × 106 m3 of material failed and rushed into the river, respectively. Both landslides totally blocked the river and induced hazard chains. Four multi-temporal unmanned aerial vehicle (UAV)-based high-resolution images and digital elevation models (DEMs) before and after the landslides were obtained and merged. The initial deformation of the two landslides can be clearly captured by optical remote-sensing images and interferometric synthetic aperture radar (InSAR) analysis. The depth-integrated continuum method was adopted to analyze the dynamic process of both landslides, using the same basal friction angle and cohesion. The computational results matched well with field measurements and observations. It is shown that the evaluation of potentially landslide-prone areas by the depth-integrated continuum method is feasible and is able to provide significant information before a possible event.

Impact of Texture Information on Crop Classification with Machine Learning and UAV Images

Unmanned aerial vehicle (UAV) images that can provide thematic information at much higher spatial and temporal resolutions than satellite images have great potential in crop classification. Due to the ultra-high spatial resolution of UAV images, spatial contextual information such as texture is often used for crop classification. From a data availability viewpoint, it is not always possible to acquire time-series UAV images due to limited accessibility to the study area. Thus, it is necessary to improve classification performance for situations when a single or minimum number of UAV images are available for crop classification. In this study, we investigate the potential of gray-level co-occurrence matrix (GLCM)-based texture information for crop classification with time-series UAV images and machine learning classifiers including random forest and support vector machine. In particular, the impact of combining texture and spectral information on the classification performance is evaluated for cases that use only one UAV image or multi-temporal images as input. A case study of crop classification in Anbandegi of Korea was conducted for the above comparisons. The best classification accuracy was achieved when multi-temporal UAV images which can fully account for the growth cycles of crops were combined with GLCM-based texture features. However, the impact of the utilization of texture information was not significant. In contrast, when one August UAV image was used for crop classification, the utilization of texture information significantly affected the classification performance. Classification using texture features extracted from GLCM with larger kernel size significantly improved classification accuracy, an improvement of 7.72%p in overall accuracy for the support vector machine classifier, compared with classification based solely on spectral information. These results indicate the usefulness of texture information for classification of ultra-high-spatial-resolution UAV images, particularly when acquisition of time-series UAV images is difficult and only one UAV image is used for crop classification.

Multitemporal Classification of River Floodplain Vegetation Using Time Series of UAV Images

The functions of river floodplains often conflict spatially, for example, water conveyance during peak discharge and diverse riparian ecology. Such functions are often associated with floodplain vegetation. Frequent monitoring of floodplain land cover is necessary to capture the dynamics of this vegetation. However, low classification accuracies are found with existing methods, especially for relatively similar vegetation types, such as grassland and herbaceous vegetation. Unmanned aerial vehicle (UAV) imagery has great potential to improve the classification of these vegetation types owing to its high spatial resolution and flexibility in image acquisition timing. This study aimed to evaluate the increase in classification accuracy obtained using multitemporal UAV images versus single time step data on floodplain land cover classification and to assess the effect of varying the number and timing of imagery acquisition moments. We obtained a dataset of multitemporal UAV imagery and field reference observations and applied object-based Random Forest classification (RF) to data of different time step combinations. High overall accuracies (OA) exceeding 90% were found for the RF of floodplain land cover, with six vegetation classes and four non-vegetation classes. Using two or more time steps compared with a single time step increased the OA from 96.9% to 99.3%. The user’s accuracies of the classes with large similarity, such as natural grassland and herbaceous vegetation, also exceeded 90%. The combination of imagery from June and September resulted in the highest OA (98%) for two time steps. Our method is a practical and highly accurate solution for monitoring areas of a few square kilometres. For large-scale monitoring of floodplains, the same method can be used, but with data from airborne platforms covering larger extents.

Multi-Temporal Remote Sensing Image Registration Using Deep Convolutional Features

Registration of multi-temporal remote sensing images has been widely applied in military and civilian fields, such as ground target identification, urban development assessment, and geographic change assessment. Ground surface change challenges feature point detection in amount and quality, which is a common dilemma faced by feature-based registration algorithms. Under severe appearance variation, detected feature points may contain a large proportion of outliers, whereas inliers may be inadequate and unevenly distributed. This paper presents a convolutional neural network (CNN) feature-based multi-temporal remote sensing image registration method with two key contributions: (i) we use a CNN to generate robust multi-scale feature descriptors and (ii) we design a gradually increasing selection of inliers to improve the robustness of feature point registration. Extensive experiments on feature matching and image registration are performed over a multi-temporal satellite image data set and a multi-temporal unmanned aerial vehicle image dataset. Our method outperforms four state-of-the-art methods in most scenarios.

Optimal spatial resolution of Unmanned Aerial Vehicle (UAV)-acquired imagery for species classification in a heterogeneous grassland ecosystem

Species composition is an essential biophysical attribute of vegetative ecosystems. Unmanned aerial vehicle (UAV)-acquired imagery with ultrahigh spatial resolution is a valuable data source for investigating species composition at a fine scale, which is extremely important for species-mixed ecosystems (e.g., grasslands and wetlands). However, the ultrahigh spatial resolution of UAV imagery also poses challenges in species classification since the imagery captures very detailed information of ground features (e.g., gaps, shadow) which would add substantial noise to image classification. In this study, we obtained multi-temporal UAV imagery with 5 cm resolution and resampled them to acquire imagery with 10, 15, and 20 cm resolution. The images were then utilized for species classification using Geographic Object-Based Image Analysis (GEOBIA) aiming to assess the influence of different imagery spatial resolution on the classification accuracy. Results show that the overall classification accuracy of imagery with 5, 10, and 15 cm resolution are close, while the classification accuracy on 20-cm imagery is much lower. These results are expected because the object features (e.g., vegetation index values and standard deviation) of same species vary slightly between 5 and 15 cm resolution, but not at the 20-cm resolution. We also found that the same species show different producer’s and user’s accuracy when using imagery with different spatial resolutions. These results suggest that it is essential to select the optimal spatial resolution of imagery for investigating a vegetative ecosystem of interest.

A New Approach to Performing Bundle Adjustment for Time Series UAV Images 3D Building Change Detection

Successful change detection in multi-temporal images relies on high spatial co-registration accuracy. However, co-registration accuracy alone cannot meet the needs of change detection when using several ground control points to separately geo-reference multi-temporal images from unmanned aerial vehicles (UAVs). This letter reports on a new approach to perform bundle adjustment—named united bundle adjustment (UBA)—to solve this co-registration problem for change detection in multi-temporal UAV images. In UBA, multi-temporal UAV images are matched with each other to construct a unified tie point net. One single bundle adjustment process is performed on the unified tie point net, placing every image into the same coordinate system and thus automatically accomplishing spatial co-registration. We then perform change detection using both orthophotos and three-dimensional height information derived from dense image matching techniques. Experimental results show that UBA co-registration accuracy is higher than the accuracy of commonly-used approaches for multi-temporal UAV images. Our proposed preprocessing method extends the capacities of consumer-level UAVs so they can eventually meet the growing need for automatic building change detection and dynamic monitoring using only RGB band images.