Jungho Im

Object-based change detection using correlation image analysis and image segmentation

This study introduces change detection based on object/neighbourhood correlation image analysis and image segmentation techniques. The correlation image analysis is based on the fact that pairs of brightness values from the same geographic area (e.g. an object) between bi‐temporal image datasets tend to be highly correlated when little change occurres, and uncorrelated when change occurs. Five different change detection methods were investigated to determine how new contextual features could improve change classification results, and if an object‐based approach could improve change classification when compared with per‐pixel analysis. The five methods examined include (1) object‐based change classification incorporating object correlation images (OCIs), (2) object‐based change classification incorporating neighbourhood correlation images (NCIs), (3) object‐based change classification without contextual features, (4) per‐pixel change classification incorporating NCIs, and (5) traditional per‐pixel change classification using only bi‐temporal image data. Two different classification algorithms (i.e. a machine‐learning decision tree and nearest‐neighbour) were also investigated. Comparison between the OCI and the NCI variables was evaluated. Object‐based change classifications incorporating the OCIs or the NCIs produced more accurate change detection classes (Kappa approximated 90%) than other change detection results (Kappa ranged from 80 to 85%).

An automatic region-based image segmentation algorithm for remote sensing applications

Object-based image analysis has proven its potentials for remote sensing applications, especially when using high-spatial resolution data. One of the first steps of object-based image analysis is to generate homogeneous regions from a pixel-based image, which is typically called the image segmentation process. This paper introduces a new automatic Region-based Image Segmentation Algorithm based on k-means clustering (RISA), specifically designed for remote sensing applications. The algorithm includes five steps: k-means clustering, segment initialization, seed generation, region growing, and region merging. RISA was evaluated using a case study focusing on land-cover classification for two sites: an agricultural area in the Republic of South Africa and a residential area in Fresno, CA. High spatial resolution SPOT 5 and QuickBird satellite imagery were used in the case study. RISA generated highly homogeneous regions based on visual inspection. The land-cover classification using the RISA-derived image segments resulted in higher accuracy than the classifications using the image segments derived from the Definiens software (eCognition) and original image pixels in combination with a minimum-distance classifier. Quantitative segmentation quality assessment using two object metrics showed RISA-derived segments successfully represented the reference objects.

Machine learning approaches for forest classification and change analysis using multi-temporal Landsat TM images over Huntington Wildlife Forest

This research investigated three machine learning approaches – decision trees, random forest, and support vector machines – to classify local forest communities at the Huntington Wildlife Forest (HWF), located in the central Adirondack Mountains of New York State, and to identify forest type change over a 20-year period using multi-temporal Landsat satellite Thematic Mapper (TM) data. Because some forest species are sensitive to topographic characteristics, three terrain correction methods – C correction, statistical–empirical (SE) correction, and Variable Empirical Coefficient Algorithm (VECA) – were utilized to account for the topographic effects. Results show that the topographic correction slightly improved the classification accuracy although the improvement was not significant based on the McNemar test. Random forest and support vector machines produced higher classification accuracies than decision trees. Besides, random forest- and support vector machine-based multi-temporal classifications better reflected the forest type change seen in the reference data. In addition, topographic features such as elevation and aspect played important roles in characterizing the forest type changes.

A review of remote sensing of forest biomass and biofuel: options for small-area applications.

Forests have served as a primary reservoir of terrestrial carbon and have long been investigated in the global climate change context. In addition, increased exposure in the public domain of climate change issues has caused greater interest in the role of forests in the global energy balance. Researchers have been investigating the use of forests as carbon sequestration systems, as well as using forest products for conversion into biofuels. Remote sensing has been widely utilized as a cost-effective tool to provide forest baseline data (e.g., biomass) for effective and efficient forest management. Forest biomass is one of the forest parameters that is widely investigated using remote sensing because biomass is directly related to the productivity of forests and provides valuable information that is necessary for understanding ecosystem functions and carbon cycling. In this paper, we review remote sensing of forest biomass, focusing on recent advances and applications (published after 2000). We also explore the challenges of using forest biomass as biofuel, a topic that is often neglected in remote sensing papers.

Object-Based Land Cover Classification Using High-Posting-Density LiDAR Data

This study introduces a method for object-based land cover classification based solely on the analysis of LiDAR-derived information—i.e., without the use of conventional optical imagery such as aerial photography or multispectral imagery. The method focuses on the relative information content from height, intensity, and shape of features found in the scene. Eight object-based metrics were used to classify the terrain into land cover information: mean height, standard deviation (STDEV) of height, height homogeneity, height contrast, height entropy, height correlation, mean intensity, and compactness. Using machine-learning decision trees, these metrics yielded land cover classification accuracies > 90%. A sensitivity analysis found that mean intensity was the key metric for differentiating between the grass and road/parking lot classes. Mean height was also a contributing discriminator for distinguishing features with different height information, such as between the building and grass classes. The shape- or texture-based metrics did not significantly improve the land cover classifications. The most important three metrics (i.e., mean height, STDEV height, and mean intensity) were sufficient to achieve classification accuracies > 90%.

Machine learning approaches to coastal water quality monitoring using GOCI satellite data

Since coastal waters are one of the most vulnerable marine systems to environmental pollution, it is very important to operationally monitor coastal water quality. This study attempts to estimate two major water quality indicators, chlorophyll-a (chl-a) and suspended particulate matter (SPM) concentrations, in coastal environments on the west coast of South Korea using Geostationary Ocean Color Imager (GOCI) satellite data. Three machine learning approaches including random forest, Cubist, and support vector regression (SVR) were evaluated for coastal water quality estimation. In situ measurements (63 samples) collected during four days in 2011 and 2012 were used as reference data. Due to the limited number of samples, leave-one-out cross validation (CV) was used to assess the performance of the water quality estimation models. Results show that SVR outperformed the other two machine learning approaches, yielding calibration R2 of 0.91 and CV root-mean-squared-error (RMSE) of 1.74 mg/m3 (40.7%) for chl-a, and calibration R2 of 0.98 and CV RMSE of 11.42 g/m3 (63.1%) for SPM when using GOCI-derived radiance data. Relative importance of the predictor variables was examined. When GOCI-derived radiance data were used, the ratio of band 2 to band 4 and bands 6 and 5 were the most influential input variables in predicting chl-a and SPM concentrations, respectively. Hourly available GOCI images were useful to discuss spatiotemporal distributions of the water quality parameters with tidal phases in the west coast of Korea.

Population estimation based on multi-sensor data fusion

This research examines the utility of QuickBird imagery and Light Detection and Ranging (LiDAR) data for estimating population at the census-block level using two approaches: area-based and volume-based. Residential-building footprints are first delineated from the remote-sensing data using image segmentation and machine-learning decision-tree classification. Regression analysis is used to model the relationship between population and the area or volume of the delineated residential buildings. Both approaches result in successful performance for estimating population with high accuracy (coefficient of determination = 0.8–0.95; root-mean-square error = 10–30 people; relative root-mean-square error = 0.1–0.3). The area-based approach is slightly better than the volume-based approach because the residential areas of the study sites are generally homogeneous (i.e. single houses), and the volume-based approach is more sensitive to classification errors. The LiDAR-derived shape information such as height greatly improves population estimation compared to population estimation using only spectral data.

Prediction of the Arctic Oscillation in boreal winter by dynamical seasonal forecasting systems

This study assesses the skill of boreal winter Arctic Oscillation (AO) predictions with state-of-the-art dynamical ensemble prediction systems (EPSs): GloSea4, CFSv2, GEOS-5, CanCM3, CanCM4, and CM2.1. Long-term reforecasts with the EPSs are used to evaluate how well they represent the AO and to assess the skill of both deterministic and probabilistic forecasts of the AO. The reforecasts reproduce the observed changes in the large-scale patterns of the Northern Hemispheric surface temperature, upper level wind, and precipitation associated with the different phases of the AO. The results demonstrate that most EPSs improve upon persistence skill scores for lead times up to 2 months in boreal winter, suggesting some potential for skillful prediction of the AO and its associated climate anomalies at seasonal time scales. It is also found that the skill of AO forecasts during the recent period (1997–2010) is higher than that of the earlier period (1983–1996).

Remote Sensing Change Detection in Urban Environments

Timely and accurate change information in the urban environment is essential for successful planning and management. The change detection may range from 1) monitoring general land cover/land use found in multiple dates of imagery, to 2) anomaly (e.g., subsidence) detection on hazardous waste sites. Remote sensing approaches to change detection have been widely used due to its cost-effectiveness, extensibility, and temporal frequency. Since the advent of high-spatial resolution satellite imagery, it has become increasing popular to detect, analyze, and monitor detailed changes such as new buildings, roads, and even patios in the urban environment. Basically, there are two types of change detection methods: 1) detection of the change using various image enhancement methods, and 2) extraction of detailed types of land-cover change based on the use of classification techniques (Chan et al. 2001; Jensen 2005)

Forest Biomass and Carbon Stock Quantification Using Airborne LiDAR Data: A Case Study Over Huntington Wildlife Forest in the Adirondack Park

In response to the need for a better understanding of biosphere-atmosphere interactions as well as carbon cycles, there is a high demand for monitoring key forest parameters such as biomass and carbon stock. These monitoring tasks provide insight into relevant biogeochemical processes as well as anthropogenic impacts on the environment. Recent advances in remote sensing techniques such as Light Detection and Ranging (LiDAR) enable scientists to nondestructively identify structural and biophysical characteristics of forests. This study quantified forest biomass and carbon stock at the plot level from small-footprint full-waveform LiDAR data collected over a montane mixed forest in September 2011, using seven modeling methods: ordinary least squares, generalized additive model, Cubist, bagging, random forest, boosted regression trees, and support vector regression (SVR). Results showed that higher percentiles of canopy height and intensity made significant contributions to the predictions, while other explanatory variables related to canopy geometric volume, structure, and canopy coverage were generally not as important. Boosted regression trees provided the highest accuracy for model calibration, whereas SVR and ordinary least squares performed slightly better than the other models in model validation. In this study, the simple ordinary least squares approach performed just as well as any advanced machine learning method.