Enhancing Land Cover Mapping through Integration of Pixel-Based and Object-Based Classifications from Remotely Sensed Imagery

Mapping Roofing with Asbestos-Containing Material by Using Remote Sensing Imagery and Machine Learning-Based Image Classification: A State-of-the-Art Review

Building roofing produced with asbestos-containing materials is a significant concern due to its detrimental health hazard implications. Efficiently locating asbestos roofing is essential to proactively mitigate and manage potential health risks from this legacy building material. Several studies utilised remote sensing imagery and machine learning-based image classification methods for mapping roofs with asbestos-containing materials. However, there has not yet been a critical review of classification methods conducted in order to provide coherent guidance on the use of different remote sensing images and classification processes. This paper critically reviews the latest works on mapping asbestos roofs to identify the challenges and discuss possible solutions for improving the mapping process. A peer review of studies addressing asbestos roof mapping published from 2012 to 2022 was conducted to synthesise and evaluate the input imagery types and classification methods. Then, the significant challenges in the mapping process were identified, and possible solutions were suggested to address the identified challenges. The results showed that hyperspectral imagery classification with traditional pixel-based classifiers caused large omission errors. Classifying very-high-resolution multispectral imagery by adopting object-based methods improved the accuracy results of ACM roof identification; however, non-optimal segmentation parameters, inadequate training data in supervised methods, and analyst subjectivity in rule-based classifications were reported as significant challenges. While only one study investigated convolutional neural networks for asbestos roof mapping, other applications of remote sensing demonstrated promising results using deep-learning-based models. This paper suggests further studies on utilising Mask R-CNN segmentation and 3D-CNN classification in the conventional approaches and developing end-to-end deep semantic classification models to map roofs with asbestos-containing materials.

Integrated usage of historical geospatial data and modern satellite images reveal long-term land use/cover changes in Bursa/Turkey, 1858-2020

Land surface of the Earth has been changing as a result of human induced activities and natural processes. Accurate representation of landscape characteristics and precise determination of spatio-temporal changes provide valuable inputs for environmental models, landscape and urban planning, and historical land cover change analysis. This study aims to determine historical land use and land cover (LULC) changes using multi-modal geospatial data, which are the cadastral maps produced in 1858, monochrome aerial photographs obtained in 1955, and multi-spectral WorldView-3 satellite images of 2020. We investigated two pilot regions, Aksu and Kestel towns in Bursa/Turkey, to analyze the long-term LULC changes quantitatively and to understand the driving forces that caused the changes. We propose methods to facilitate the preparation of historical datasets for the LULC change detection and present an object-oriented joint classification scheme for multi-source datasets to accurately map the spatio-temporal changes. Our approach minimized the amount of manual digitizing required for the boundary delineation of LULC classes from historical geospatial data. Also, our quantitative analysis of LULC maps indicates diverging developments for the selected locations in the long period of 162 years. We observed rural depopulation and gradual afforestation in Aksu; whereas, agricultural land abandonment and deforestation in Kestel.

Tree Species Mapping on Sentinel-2 Satellite Imagery with Weakly Supervised Classification and Object-Wise Sampling

Information on forest composition, specifically tree types and their distribution, aids in timber stock calculation and can help to better understand the biodiversity in a particular region. Automatic satellite imagery analysis can significantly accelerate the process of tree type classification, which is traditionally carried out by ground-based observation. Although computer vision methods have proven their efficiency in remote sensing tasks, specific challenges arise in forestry applications. The forest inventory data often contain the tree type composition but do not describe their spatial distribution within each individual stand. Therefore, some pixels can be assigned a wrong label in the semantic segmentation task if we consider each stand to be homogeneously populated by its dominant species. Another challenge is the spatial distribution of individual stands within the study area. Classes are usually imbalanced and distributed nonuniformly that makes sampling choice more critical. This study aims to enhance tree species classification based on a neural network approach providing automatic markup adjustment and improving sampling technique. For forest species markup adjustment, we propose using a weakly supervised learning approach based on the knowledge of dominant species content within each stand. We also propose substituting the commonly used CNN sampling approach with the object-wise one to reduce the effect of the spatial distribution of forest stands. We consider four species commonly found in Russian boreal forests: birch, aspen, pine, and spruce. We use imagery from the Sentinel-2 satellite, which has multiple bands (in the visible and infrared spectra) and a spatial resolution of up to 10 meters. A data set of images for Leningrad Oblast of Russia is used to assess the methods. We demonstrate how to modify the training strategy to outperform a basic CNN approach from F1-score 0.68 to 0.76. This approach is promising for future studies to obtain more specific information about stands composition even using incomplete data.

From Land Cover Map to Land Use Map: A Combined Pixel-Based and Object-Based Approach Using Multi-Temporal Landsat Data, a Random Forest Classifier, and Decision Rules

It is essential to produce land cover maps and land use maps separately for different purposes. This study was conducted to generate such maps in Binh Duong province, Vietnam, using a novel combination of pixel-based and object-based classification techniques and geographic information system (GIS) analysis on multi-temporal Landsat images. Firstly, the connection between land cover and land use was identified; thereafter, the land cover map and land use function regions were extracted with a random forest classifier. Finally, a land use map was generated by combining the land cover map and the land use function regions in a set of decision rules. The results showed that land cover and land use were linked by spectral, spatial, and temporal characteristics, and this helped effectively convert the land cover map into a land use map. The final land cover map attained an overall accuracy (OA) = 93.86%, with producer’s accuracy (PA) and user’s accuracy (UA) of its classes ranging from 73.91% to 100%. Meanwhile, the final land use map achieved OA = 93.45%, and the UA and PA ranged from 84% to 100%. The study demonstrated that it is possible to create high-accuracy maps based entirely on free multi-temporal satellite imagery that promote the reproducibility and proactivity of the research as well as cost-efficiency and time savings.

Accuracy Improvements to Pixel-Based and Object-Based LULC Classification with Auxiliary Datasets from Google Earth Engine

The monitoring and assessment of land use/land cover (LULC) change over large areas are significantly important in numerous research areas, such as natural resource protection, sustainable development, and climate change. However, accurately extracting LULC only using the spectral features of satellite images is difficult owing to landscape heterogeneities over large areas. To improve the accuracy of LULC classification, numerous studies have introduced other auxiliary features to the classification model. The Google Earth Engine (GEE) not only provides powerful computing capabilities, but also provides a large amount of remote sensing data and various auxiliary datasets. However, the different effects of various auxiliary datasets in the GEE on the improvement of the LULC classification accuracy need to be elucidated along with methods that can optimize combinations of auxiliary datasets for pixel- and object-based classification. Herein, we comprehensively analyze the performance of different auxiliary features in improving the accuracy of pixel- and object-based LULC classification models with medium resolution. We select the Yangtze River Delta in China as the study area and Landsat-8 OLI data as the main dataset. Six types of features, including spectral features, remote sensing multi-indices, topographic features, soil features, distance to the water source, and phenological features, are derived from auxiliary open-source datasets in GEE. We then examine the effect of auxiliary datasets on the improvement of the accuracy of seven pixels-based and seven object-based random forest classification models. The results show that regardless of the types of auxiliary features, the overall accuracy of the classification can be improved. The results further show that the object-based classification achieves higher overall accuracy compared to that obtained by the pixel-based classification. The best overall accuracy from the pixel-based (object-based) classification model is 94.20% (96.01%). The topographic features play the most important role in improving the overall accuracy of classification in the pixel- and object-based models comprising all features. Although a higher accuracy is achieved when the object-based method is used with only spectral data, small objects on the ground cannot be monitored. However, combined with many types of auxiliary features, the object-based method can identify small objects while also achieving greater accuracy. Thus, when applying object-based classification models to mid-resolution remote sensing images, different types of auxiliary features are required. Our research results improve the accuracy of LULC classification in the Yangtze River Delta and further provide a benchmark for other regions with large landscape heterogeneity.

Forest Road Detection Using LiDAR Data and Hybrid Classification

Knowledge about forest road networks is essential for sustainable forest management and fire management. The aim of this study was to assess the accuracy of a new hierarchical-hybrid classification tool (HyClass) for mapping paved and unpaved forest roads with LiDAR data. Bare-earth and low-lying vegetation were also identified. For this purpose, a rural landscape (area 70 ha) in northwestern Spain was selected for study, and a road network map was extracted from the cadastral maps as the ground truth data. The HyClass tool is based on a decision tree which integrates segmentation processes at local scale with decision rules. The proposed approach yielded an overall accuracy (OA) of 96.5%, with a confidence interval (CI) of 94.0–97.6%, representing an improvement over pixel-based classification (OA = 87.0%, CI = 83.7–89.8%) using Random Forest (RF). In addition, with the HyClass tool, the classification precision varied significantly after reducing the original point density from 8.7 to 1 point/m2. The proposed method can provide accurate road mapping to support forest management as an alternative to pixel-based RF classification when the LiDAR point density is higher than 1 point/m2.

Mapping Annual Land Use and Land Cover Changes in the Yangtze Estuary Region Using an Object-Based Classification Framework and Landsat Time Series Data

A system understanding of the patterns, causes, and trends of long-term land use and land cover (LULC) change at the regional scale is essential for policy makers to address the growing challenges of local sustainability and global climate change. However, it still remains a challenge for estuarine and coastal regions due to the lack of appropriate approaches to consistently generate accurate and long-term LULC maps. In this work, an object-based classification framework was designed to mapping annual LULC changes in the Yangtze River estuary region from 1985–2016 using Landsat time series data. Characteristics of the inter-annual changes of LULC was then analyzed. The results showed that the object-based classification framework could accurately produce annual time series of LULC maps with overall accuracies over 86% for all single-year classifications. Results also indicated that the annual LULC maps enabled the clear depiction of the long-term variability of LULC and could be used to monitor the gradual changes that would not be observed using bi-temporal or sparse time series maps. Specifically, the impervious area rapidly increased from 6.42% to 22.55% of the total land area from 1985 to 2016, whereas the cropland area dramatically decreased from 80.61% to 55.44%. In contrast to the area of forest and grassland, which almost tripled, the area of inland water remained consistent from 1985 to 2008 and slightly increased from 2008 to 2016. However, the area of coastal marshes and barren tidal flats varied with large fluctuations.

Large-scale probabilistic identification of boreal peatlands using Google Earth Engine, open-access satellite data, and machine learning

Freely-available satellite data streams and the ability to process these data on cloud-computing platforms such as Google Earth Engine have made frequent, large-scale landcover mapping at high resolution a real possibility. In this paper we apply these technologies, along with machine learning, to the mapping of peatlands–a landcover class that is critical for preserving biodiversity, helping to address climate change impacts, and providing ecosystem services, e.g., carbon storage–in the Boreal Forest Natural Region of Alberta, Canada. We outline a data-driven, scientific framework that: compiles large amounts of Earth observation data sets (radar, optical, and LiDAR); examines the extracted variables for suitability in peatland modelling; optimizes model parameterization; and finally, predicts peatland occurrence across a large boreal area (397, 958 km²) of Alberta at 10 m spatial resolution (equalling 3.9 billion pixels across Alberta). The resulting peatland occurrence model shows an accuracy of 87% and a kappa statistic of 0.57 when compared to our validation data set. Differentiating peatlands from mineral wetlands achieved an accuracy of 69% and kappa statistic of 0.37. This data-driven approach is applicable at large geopolitical scales (e.g., provincial, national) for wetland and landcover inventories that support long-term, responsible resource management.

Improving Remote Sensing Image Super-Resolution Mapping Based on the Spatial Attraction Model by Utilizing the Pansharpening Technique

The spatial distribution information of remote sensing images can be derived by the super-resolution mapping (SRM) technique. Super-resolution mapping, based on the spatial attraction model (SRMSAM), has been an important SRM method, due to its simplicity and explicit physical meanings. However, the resolution of the original remote sensing image is coarse, and the existing SRMSAM cannot take full advantage of the spatial–spectral information from the original image. To utilize more spatial–spectral information, improving remote sensing image super-resolution mapping based on the spatial attraction model by utilizing the pansharpening technique (SRMSAM-PAN) is proposed. In SRMSAM-PAN, a novel processing path, named the pansharpening path, is added to the existing SRMSAM. The original coarse remote sensing image is first fused with the high-resolution panchromatic image from the same area by the pansharpening technique in the novel pansharpening path, and the improved image is unmixed to obtain the novel fine-fraction images. The novel fine-fraction images from the pansharpening path and the existing fine-fraction images from the existing path are then integrated to produce finer-fraction images with more spatial–spectral information. Finally, the values predicted from the finer-fraction images are utilized to allocate class labels to all subpixels, to achieve the final mapping result. Experimental results show that the proposed SRMSAM-PAN can obtain a higher mapping accuracy than the existing SRMSAM methods.

Monitoring Land-Use/Land-Cover Changes at a Provincial Large Scale Using an Object-Oriented Technique and Medium-Resolution Remote-Sensing Images

An object-based image analysis (OBIA) technique is replacing traditional pixel-based methods and setting a new standard for monitoring land-use/land-cover changes (LUCC). To date, however, studies have focused mainly on small-scale exploratory experiments and high-resolution remote-sensing images. Therefore, this study used OBIA techniques and medium-resolution Chinese HJ-CCD images to monitor LUCC at the provincial scale. The results showed that while woodland was mainly distributed in the west, south, and east mountain areas of Hunan Province, the west had the largest area and most continuous distribution. Wetland was distributed mainly in the northern plain area, and cultivated land was distributed mainly in the central and northern plains and mountain valleys. The largest impervious surface was the Changzhutan urban agglomerate in the northeast plain area. The spatial distribution of land cover in Hunan Province was closely related to topography, government policy, and economic development. For the period 2000–2010, the areas of cultivated land transformed into woodland, grassland, and wetland were 183.87 km2, 5.57 km2, and 70.02 km2, respectively, indicating that the government-promoted ecologically engineered construction was yielding some results. The rapid economic growth and urbanization, high resource development intensity, and other natural factors offset the gains made in ecologically engineered construction and in increasing forest and wetland areas, respectively, by 229.82 km2 and 132.12 km2 from 2000 to 2010 in Hunan Province. The results also showed large spatial differences in change amplitude (LUCCA), change speed (LUCCS), and transformation processes in Hunan Province. The Changzhutan urban agglomerate and the surrounding prefectures had the largest LUCCA and LUCCS, where the dominant land cover accounted for the conversion of some 189.76 km2 of cultivated land, 129.30 km2 of woodland, and 6.12 km2 of wetland into impervious surfaces in 2000–2010. This conversion was attributed to accelerated urbanization and rapid economic growth in this region.

Learning a Multi-Branch Neural Network from Multiple Sources for Knowledge Adaptation in Remote Sensing Imagery

In this paper we propose a multi-branch neural network, called MB-Net, for solving the problem of knowledge adaptation from multiple remote sensing scene datasets acquired with different sensors over diverse locations and manually labeled with different experts. Our aim is to learn invariant feature representations from multiple source domains with labeled images and one target domain with unlabeled images. To this end, we define for MB-Net an objective function that mitigates the multiple domain shifts at both feature representation and decision levels, while retaining the ability to discriminate between different land-cover classes. The complete architecture is trainable end-to-end via the backpropagation algorithm. In the experiments, we demonstrate the effectiveness of the proposed method on a new multiple domain dataset created from four heterogonous scene datasets well known to the remote sensing community, namely, the University of California (UC-Merced) dataset, the Aerial Image dataset (AID), the PatternNet dataset, and the Northwestern Polytechnical University (NWPU) dataset. In particular, this method boosts the average accuracy over all transfer scenarios up to 89.05% compared to standard architecture based only on cross-entropy loss, which yields an average accuracy of 78.53%.

Assessing Cropland Area in West Africa for Agricultural Yield Analysis

Accurate estimates of cultivated area and crop yield are critical to our understanding of agricultural production and food security, particularly for semi-arid regions like the Sahel of West Africa, where crop production is mainly rain-fed and food security is closely correlated with the inter-annual variations in rainfall. Several global and regional land cover products, based on satellite remotely-sensed data, provide estimates of the agricultural land use intensity, but the initial comparisons indicate considerable differences among them, relating to differences in the satellite data quality, classification approaches, and spatial and temporal resolutions. Here, we quantify the accuracy of available cropland products across Sahelian West Africa using an independent, high-resolution, visually interpreted sample dataset that classifies all points across West Africa using a 2-km sample grid (~500,000 points for the study area). We estimate the “quantity” and “allocation” disagreements for the cropland class of eight land cover products in five Western Sahel countries (Burkina Faso, Mali, Mauritania, Niger, and Senegal). The results confirm that coarse spatial resolution (300 m, 500 m, and 1000 m) land cover products have higher disagreements in mapping the fragmented agricultural landscape of the Western Sahel. Earlier products (e.g., GLC2000) are less accurate than recent products (e.g., ESA CCI 2013, MODIS 2013 and GlobCover 2009). We also show that two of the finer spatial resolution maps (GFSAD30, and GlobeLand30) using advanced classification approaches (random forest, decision trees, and pixel-object combined) are currently the best available products for cropland identification. However, none of the eight land cover databases examined is consistent in reaching the targeted 75% accuracy threshold in the five Sahelian countries. The majority of currently available land cover products overestimate cultivated areas by an average of 170% relative to the cropland area in the reference data.