Object-based image classification for land-cover mapping purposes using remote-sensing imagery has attracted significant attention in recent years. Numerous studies conducted over the past decade have investigated a broad array of sensors, feature selection, classifiers, and other factors of interest. However, these research results have not yet been synthesized to provide coherent guidance on the effect of different supervised object-based land-cover classification processes. In this study, we first construct a database with 28 fields using qualitative and quantitative information extracted from 254 experimental cases described in 173 scientific papers. Second, the results of the meta-analysis are reported, including general characteristics of the studies (e.g., the geographic range of relevant institutes, preferred journals) and the relationships between factors of interest (e.g., spatial resolution and study area or optimal segmentation scale, accuracy and number of targeted classes), especially with respect to the classification accuracy of different sensors, segmentation scale, training set size, supervised classifiers, and land-cover types. Third, useful data on supervised object-based image classification are determined from the meta-analysis. For example, we find that supervised object-based classification is currently experiencing rapid advances, while development of the fuzzy technique is limited in the object-based framework. Furthermore, spatial resolution correlates with the optimal segmentation scale and study area, and Random Forest (RF) shows the best performance in object-based classification. The area-based accuracy assessment method can obtain stable classification performance, and indicates a strong correlation between accuracy and training set size, while the accuracy of the point-based method is likely to be unstable due to mixed objects. In addition, the overall accuracy benefits from higher spatial resolution images (e.g., unmanned aerial vehicle) or agricultural sites where it also correlates with the number of targeted classes. More than 95.6% of studies involve an area less than 300 ha, and the spatial resolution of images is predominantly between 0 and 2 m. Furthermore, we identify some methods that may advance supervised object-based image classification. For example, deep learning and type-2 fuzzy techniques may further improve classification accuracy. Lastly, scientists are strongly encouraged to report results of uncertainty studies to further explore the effects of varied factors on supervised object-based image classification.

A review of supervised object-based land-cover image classification

Despite recent advances of deep Convolutional Neural Networks (CNNs) in various computer vision tasks, their potential for classification of multispectral remote sensing images has not been thoroughly explored. In particular, the applications of deep CNNs using optical remote sensing data have focused on the classification of very high-resolution aerial and satellite data, owing to the similarity of these data to the large datasets in computer vision. Accordingly, this study presents a detailed investigation of state-of-the-art deep learning tools for classification of complex wetland classes using multispectral RapidEye optical imagery. Specifically, we examine the capacity of seven well-known deep convnets, namely DenseNet121, InceptionV3, VGG16, VGG19, Xception, ResNet50, and InceptionResNetV2, for wetland mapping in Canada. In addition, the classification results obtained from deep CNNs are compared with those based on conventional machine learning tools, including Random Forest and Support Vector Machine, to further evaluate the efficiency of the former to classify wetlands. The results illustrate that the full-training of convnets using five spectral bands outperforms the other strategies for all convnets. InceptionResNetV2, ResNet50, and Xception are distinguished as the top three convnets, providing state-of-the-art classification accuracies of 96.17%, 94.81%, and 93.57%, respectively. The classification accuracies obtained using Support Vector Machine (SVM) and Random Forest (RF) are 74.89% and 76.08%, respectively, considerably inferior relative to CNNs. Importantly, InceptionResNetV2 is consistently found to be superior compared to all other convnets, suggesting the integration of Inception and ResNet modules is an efficient architecture for classifying complex remote sensing scenes such as wetlands.

https://www.mdpi.com/2072-4292/10/7/1119/pdf

Very Deep Convolutional Neural Networks for Complex Land Cover Mapping Using Multispectral Remote Sensing Imagery

Wetlands are some of the most important ecosystems on Earth. They play a key role in alleviating floods and filtering polluted water and also provide habitats for many plants and animals. Wetlands also interact with climate change. Over the past 50 years, wetlands have been polluted and declined dramatically as land cover has changed in some regions. Remote sensing has been the most useful tool to acquire spatial and temporal information about wetlands. In this paper, seven types of sensors were reviewed: aerial photos coarse-resolution, medium-resolution, high-resolution, hyperspectral imagery, radar, and Light Detection and Ranging (LiDAR) data. This study also discusses the advantage of each sensor for wetland research. Wetland research themes reviewed in this paper include wetland classification, habitat or biodiversity, biomass estimation, plant leaf chemistry, water quality, mangrove forest, and sea level rise. This study also gives an overview of the methods used in wetland research such as supervised and unsupervised classification and decision tree and object-based classification. Finally, this paper provides some advice on future wetland remote sensing. To our knowledge, this paper is the most comprehensive and detailed review of wetland remote sensing and it will be a good reference for wetland researchers.

A Review of Wetland Remote Sensing

Wetlands are important ecosystems around the world, although they are degraded due both to anthropogenic and natural process. Newfoundland is among the richest Canadian province in terms of different wetland classes. Herbaceous wetlands cover extensive areas of the Avalon Peninsula, which are the habitat of a number of animal and plant species. In this study, a novel hierarchical object-based Random Forest (RF) classification approach is proposed for discriminating between different wetland classes in a sub-region located in the north eastern portion of the Avalon Peninsula. Particularly, multi-polarization and multi-frequency SAR data, including X-band TerraSAR-X single polarized (HH), L-band ALOS-2 dual polarized (HH/HV), and C-band RADARSAT-2 fully polarized images, were applied in different classification levels. First, a SAR backscatter analysis of different land cover types was performed by training data and used in Level-I classification to separate water from non-water classes. This was followed by Level-II classification, wherein the water class was further divided into shallow- and deep-water classes, and the non-water class was partitioned into herbaceous and non-herbaceous classes. In Level-III classification, the herbaceous class was further divided into bog, fen, and marsh classes, while the non-herbaceous class was subsequently partitioned into urban, upland, and swamp classes. In Level-II and -III classifications, different polarimetric decomposition approaches, including Cloude-Pottier, Freeman-Durden, Yamaguchi decompositions, and Kennaugh matrix elements were extracted to aid the RF classifier. The overall accuracy and kappa coefficient were determined in each classification level for evaluating the classification results. The importance of input features was also determined using the variable importance obtained by RF. It was found that the Kennaugh matrix elements, Yamaguchi, and Freeman-Durden decompositions were the most important parameters for wetland classification in this study. Using this new hierarchical RF classification approach, an overall accuracy of up to 94% was obtained for classifying different land cover types in the study area.

Random forest wetland classification using ALOS-2 L-band, RADARSAT-2 C-band, and TerraSAR-X imagery

The applications of object-based image analysis (OBIA) in remote sensing studies of wetlands have been growing over recent decades, addressing tasks from detection and delineation of wetland bodies to comprehensive analyses of within-wetland cover types and their change. Compared to pixel-based approaches, OBIA offers several important benefits to wetland analyses related to smoothing of the local noise, incorporating meaningful non-spectral features for class separation and accounting for landscape hierarchy of wetland ecosystem organization and structure. However, there has been little discussion on whether unique challenges of wetland environments can be uniformly addressed by OBIA across different types of data, spatial scales and research objectives, and to what extent technical and conceptual aspects of this framework may themselves present challenges in a complex wetland setting. This review presents a synthesis of 73 studies that applied OBIA to different types of remote sensing data, spatial scale and research objectives. It summarizes the progress and scope of OBIA uses in wetlands, key benefits of this approach, factors related to accuracy and uncertainty in its applications and the main research needs and directions to expand the OBIA capacity in the future wetland studies. Growing demands for higher-accuracy wetland characterization at both regional and local scales together with advances in very high resolution remote sensing and novel tasks in wetland restoration monitoring will likely continue active exploration of the OBIA potential in these diverse and complex environments.

Object-Based Image Analysis in Wetland Research: A Review

The Brazilian Pantanal is a large continuous tropical wetland with large biodiversity and many threatened habitats. The interplay between the distribution of vegetation, the hydrology, the climate and the geomorphology nourishes and sustains the large diversity of flora and fauna in this wetland, but it is poorly understood at the scale of the entire Pantanal. This study uses multi-temporal L-band ALOS/PALSAR and C-band RADARSAT-2 data to map ecosystems and create spatial-temporal maps of flood dynamics in the Brazilian Pantanal. First, an understanding of the backscattering characteristics of floodable and non-floodable habitats was developed. Second, a Level 1 object-based image analysis (OBIA) classification defining Forest, Savanna, Grasslands/Agriculture, Aquatic Vegetation and Open Water cover types was achieved with accuracy results of 81%. A Level 2 classification separating Flooded from Non-Flooded regions for five temporal periods over one year was also accomplished, showing the interannual variability among sub-regions in the Pantanal. Cross-sensor, multi-temporal SAR data was found to be useful in mapping both land cover and flood patterns in wetland areas. The generated maps will be a valuable asset for defining habitats required to conserve the Pantanal biodiversity and to mitigate the impacts of human development in the region.

Using ALOS/PALSAR and RADARSAT-2 to Map Land Cover and Seasonal Inundation in the Brazilian Pantanal

Landcover classification of the Lower Nhecolândia subregion of the Brazilian Pantanal Wetlands using ALOS/PALSAR, RADARSAT-2 and ENVISAT/ASAR imagery

Large-scale habitat mapping of the Brazilian Pantanal wetland: a synthetic aperture radar approach.

The Brazilian Pantanal is an ecosystem of high biodiversity This biodiversity is largely supported by the dynamic hydrology and landscape of the region As part of the mosaic of this unique landscape, there are a variety of geochemically diverse lakes in the lower Nhecolândia (LN) sub-region The goal of this study was to provide an inventory, spatial distribution, and geochemical analysis of the lakes, and the characteristics of the associated surrounding landscape of the LN region in the Pantanal, and to use that information to further add to the hypotheses on lake formation To accomplish this, we made use of: (i) the spatial distribution of lakes and surrounding landscape units from a classification product generated from a combination of fine pixel resolution L-band ALOS/PALSAR, C-band Radarsat-2, and Envisat/ASAR; and (ii) in situ measured lake water geochemical propertiesWith an accuracy of 98 %, we found that the LN sub-region has 637 salinas and 8,214 baias The greatest degrees of salinas clustering are located in the southeast portion of the study area, and are surrounded by forested savanna, which are in higher grounds less prone to flooding Baias, however, are more broadly distributed in the region, and the highest occurrence is associated with the open mixed cover, comprised mostly of grasslands, which are in flood-prone areas Geochemically, salinas present a pH > 90 and total dissolved solids (TDS) generally >1,000 mg/L, while baias have a circum-neutral pH not much higher than 70, and TDS 90, comparable to the salinas, but relatively lower TDS and conductivity, and significant lower chloride The combination of the regional scale spatial data and associated geochemistry of lakes suggests that the geomorphological characteristics of the region, and consequent undulating terrains, create groundwater divides and groundwater-sheds that can develop local and regional flow systems and lead to lakes with different geochemistry The mechanisms responsible for the formation of the diversity of lakes in the region is likely due to combination of isolation from ground water/surface water recharge, evaporation, episodic flushing, and biogeochemical processes in the lake’s watersheds

The lakes of the Pantanal: inventory, distribution, geochemistry, and surrounding landscape

The Pantanal is a large cross-border wetland system in South America that supports a diversity of ecosystems. This internationally important wetland is under threat from water diversion but also agricultural development and there is a need for satellite-based observations to characterize, map, and monitor the changing extents and states of the wetlands. To achieve this, a number of connected studies have been undertaken using combination of Synthetic Aperture Radar (SAR) operating at C- and L-band. Different lake types were also classified using an object-oriented approach, with these linked to different levels of salinity, primarily through association with vegetation types.

Teresa L. Evans

Papers

Using ALOS/PALSAR and RADARSAT-2 to Map Land Cover and Seasonal Inundation in the Brazilian Pantanal

Landcover classification of the Lower Nhecolândia subregion of the Brazilian Pantanal Wetlands using ALOS/PALSAR, RADARSAT-2 and ENVISAT/ASAR imagery

Large-scale habitat mapping of the Brazilian Pantanal wetland: a synthetic aperture radar approach.

The lakes of the Pantanal: inventory, distribution, geochemistry, and surrounding landscape

Remote Sensing of Wetland Types: Subtropical Wetlands of Southern Hemisphere