Multiresolution Segmentation Algorithm Research Articles

Automated parameterisation for multi-scale image segmentation on multiple layers

We introduce a new automated approach to parameterising multi-scale image segmentation of multiple layers, and we implemented it as a generic tool for the eCognition® software. This approach relies on the potential of the local variance (LV) to detect scale transitions in geospatial data. The tool detects the number of layers added to a project and segments them iteratively with a multiresolution segmentation algorithm in a bottom-up approach, where the scale factor in the segmentation, namely, the scale parameter (SP), increases with a constant increment. The average LV value of the objects in all of the layers is computed and serves as a condition for stopping the iterations: when a scale level records an LV value that is equal to or lower than the previous value, the iteration ends, and the objects segmented in the previous level are retained. Three orders of magnitude of SP lags produce a corresponding number of scale levels. Tests on very high resolution imagery provided satisfactory results for generic applicability. The tool has a significant potential for enabling objectivity and automation of GEOBIA analysis.

Optimizing multi-resolution segmentation scale using empirical methods: Exploring the sensitivity of the supervised discrepancy measure Euclidean distance 2 (ED2)

Multiresolution segmentation (MRS) has proven to be one of the most successful image segmentation algorithms in the geographic object-based image analysis (GEOBIA) framework. This algorithm is relatively complex and user-dependent; scale, shape, and compactness are the main parameters available to users for controlling the algorithm. Plurality of segmentation results is common because each parameter may take a range of values within its parameter space or different combinations of values among parameters. Finding optimal parameter values through a trial-and-error process is commonly practiced at the expense of time and labor, thus, several alternative supervised and unsupervised methods for supervised automatic parameter setting have been proposed and tested. In the case of supervised empirical assessments, discrepancy measures are employed for computing measures of dissimilarity between a reference polygon and an image object candidate. Evidently the reliability of the optimal-parameter prediction heavily relies on the sensitivity of the segmentation quality metric. The idea behind pursuing optimal parameter setting is that, for instance, a given scale setting provides image object candidates different from the other scale setting; thus, by design the supervised quality metric should capture this difference. In this exploratory study, we selected the Euclidean distance 2 (ED2) metric, a recently proposed supervised metric, whose main design goal is to optimize the geometrical discrepancy (potential segmentation error (PSE)) and arithmetic discrepancy between image objects and reference polygons (number-of segmentation ratio (NSR)) in two dimensional Euclidean space, as a candidate to investigate the validity and efficacy of empirical discrepancy measures for finding the optimal scale parameter setting of the MRS algorithm. We chose test image scenes from four different space-borne sensors with varying spatial resolutions and scene contents and systematically segmented them using the MRS algorithm at a series of parameter settings. The discriminative capacity of the ED2 metric across different scales groups was tested using non-parametric statistical methods. Our results showed that the ED2 metric significantly discriminates the quality of image object candidates at smaller scale values but it loses the sensitivity at larger scale values. This questions the meaningfulness of the ED2 metric in the MRS algorithm’s parameter optimization. Our contention is that the ED2 metric provides some notion of the optimal scale parameter at the expense of time. In this respect, especially in operational-level image processing, it is worth to re-think the trade-off between execution time of the processor-intensive MRS algorithm at series of parameter settings targeting a less-sensitive quality metric and an expert-lead trial-and-error approach.

Optimization of Object-Based Image Analysis With Random Forests for Land Cover Mapping

A prerequisite for object-based image analysis is the generation of adequate segments. However, the parameters for the image segmentation algorithms are often manually defined. Therefore, the generation of an ideal segmentation level is usually costly and user-depended. In this paper a strategy for a semi-automatic optimization of object-based classification of multitemporal data is introduced by using Random Forest (RF) and a novel segmentation algorithm. The Superpixel Contour (SPc) algorithm is used to generate a set of different levels of segmentation, using various combinations of parameters in a user-defined range. Finally, the best parameter combination is selected based on the cross-validation-like out-of-bag (OOB) error that is provided by RF. Therefore, the quality of the parameters and the corresponding segmentation level can be assessed in terms of the classification accuracy, without providing additional independent test data. To evaluate the potential of the proposed concept, we focus on land cover classification of two study areas, using multitemporal RapidEye and SPOT 5 images. A classification that is based on eCognition's widely used multiresolution segmentation algorithm (MRS) is used for comparison. Experimental results underline that the two segmentation algorithms SPc and MRS perform similar in terms of accuracy and visual interpretation. The proposed strategy that uses the OOB error for the selection of the ideal segmentation level provides similar classification accuracies, when compared to the results achieved by manual-based image segmentation. Overall, the proposed strategy is operational and easy to handle and thus economizes the findings of optimal segmentation parameters for the Superpixel Contour algorithm.

Evaluation of data fusion and image segmentation in earth observation based rapid mapping workflows

This paper is an exploratory study, which aimed to discover the synergies of data fusion and image segmentation in the context of EO-based rapid mapping workflows. Our approach pillared on the geographic object-based image analysis (GEOBIA) focusing on multiscale, internally-displaced persons’ (IDP) camp information extraction from very high spatial resolution (VHSR) images. We applied twelve pansharpening algorithms to two subsets of a GeoEye-1 image scene that was taken over a former war-induced ephemeral settlement in Sri Lanka. A multidimensional assessment was employed to benchmark pansharpening algorithms with respect to their spectral and spatial fidelity. The multiresolution segmentation (MRS) algorithm of the eCognition Developer software served as the key algorithm in the segmentation process. The first study site was used for comparing segmentation results produced from the twelve fused products at a series of scale, shape, and compactness settings of the MRS algorithm. The segmentation quality and optimum parameter settings of the MRS algorithm were estimated by using empirical discrepancy measures. Non-parametric statistical tests were used to compare the quality of image object candidates, which were derived from the twelve pansharpened products. A wall-to-wall classification was performed based on a support vector machine (SVM) classifier to classify image objects candidates of the fused images. The second site simulated a more realistic crisis information extraction scenario where the domain expertise is crucial in segmentation and classification. We compared segmentation and classification results of the original images (non-fused) and twelve fused images to understand the efficacy of data fusion. We have shown that the GEOBIA has the ability to create meaningful image objects during the segmentation process by compensating the fused image’s spectral distortions with the high-frequency information content that has been injected during fusion. Our findings further questioned the necessity of the data fusion step in rapid mapping context. Bypassing time-intensive data fusion helps to actuate EO-based rapid mapping workflows. We, however, emphasize the fact that data fusion is not limited to VHSR image data but expands over many different combinations of multi-date, multi-sensor EO-data. Thus, further research is needed to understand the synergies of data fusion and image segmentation with respect to multi-date, multi-sensor fusion scenarios and extrapolate our findings to other remote sensing application domains beyond EO-based crisis information retrieval.

Landslide susceptibility assessment using object mapping units, decision tree, and support vector machine models in the Three Gorges of China

Due to the particular geographical location and complex geological conditions, the Three Gorges of China suffer from many landslide hazards that often result in tragic loss of life and economic devastation. To reduce the casualty and damages, an effective and accurate method of assessing landslide susceptibility is necessary. Object-based data mining methods were applied to a case study of landslide susceptibility assessment on the Guojiaba Town of the Three Gorges. The study area was partitioned into object mapping units derived from 30 m resolution Landsat TM images using multi-resolution segmentation algorithm based on the landslide factors of engineering rock group, homogeneity, and reservoir water level. Landslide locations were determined by interpretation of Landsat TM images and extensive field surveys. Eleven primary landslide-related factors were extracted from the topographic and geologic maps, and satellite images. Those factors were selected as independent variables using significance testing and correlation coefficient analysis, including slope, profile curvature, engineering rock group, slope structure, distance from faults, land cover, tasseled cap transformation wetness index, reservoir water level, homogeneity, and first and second principal components of the images. Decision tree and support vector machine (SVM) models with the optimal parameters were trained and then used to map landslide susceptibility, respectively. The analytical results were validated by comparing them with known landslides using the success rate and prediction rate curves and classification accuracy. The object-based SVM model has the highest correct rate of 89.36 % and a kappa coefficient of 0.8286 and outperforms the pixel-based SVM, object-based C5.0, and pixel-based SVM models.

Accuracy assessment measures for image segmentation goodness of the Land Parcel Identification System (LPIS) in Denmark

This letter evaluates the performance of eCognition’s multi-resolution segmentation algorithm on the task of delimiting agricultural parcels in Denmark using very high spatial resolution 8-band WorldView-2 (WV2) imagery. Fifty-seven different scale parameters setting, ranging from 100 to 1500, was employed in order to assess the quality of segmentation. An accuracy assessment was performed using seven metrics based on the topological or geometric similarity between segmented polygons and reference polygons, which were derived through manual delineation. The results indicate that (1) segmentation accuracy is influenced by the size of the reference polygons and (2) the presence of clear boundaries (e.g. hedgerow, ponds, ditches and road) drives the segmentation algorithm when the scale parameter exceeds a certain value.

Village forms Classification by Object-Based Image Analysis

This paper presents and discusses an application of object-based image analysis for rural land use/land cover classification based on village forms and shapes in Northeastern, Thailand. Increasing availability of VHRS data and object-based classification techniques can be extremely effective to delineate the complexity of land cover in the study area. Quick Bird pan-sharpened imagery with spatial resolution of 0.6 meters is being used in this study. First, a multi resolution segmentation algorithm was used for creating image objects from heterogeneous pixel values. Transportation network and topographic information were also incorporated as ancillary data into the segmentation procedure. It allows a creation of different levels of segments supporting a hierarchical structure, including spatial relations between objects and sub-objects. Second, in order to use the best criterion and threshold selection, the NN classifier had been used for features analysis. The class hierarchy was divided into 3levels, i.e. AOI/Non-AOI (L1), analysis level (L2) and top-most level (L3). Land use was classified into 8 classes based on the classification system, developed by Land Development Department, Thailand (e.g. urban/built-up land, agricultural area, range land, forest land and water). The classification was integrating not only spectral information, but also contextual and spatial relationships among image objects. A rule-based classifier has been used for the classification. The criterion of membership functions and the class descriptions were carried out from spectral information and spatial relation derived from image objects. Last, the classification result consists of 2 classes, non-village vs. residential. This study demonstrates that OBIA with topographic variables produces better classification results than OBIA with spectral information only. In the accuracy assessment, the result has been compared with reference ground truth points, including digital land use map, visual interpretation and TTA masks obtained from the pre-classification process. The overall accuracy achieved has reached approximately 70%, and Kappa index of agreement 0.64. OBIA technique is an effective tool for land use/land cover classification from VHRS data in rural areas aimed at land use planning, land use monitoring and ultimately for increasing the quality of life in rural societies.

A Segmentation Algorithm for Remote Sensing Imaging Based on Edge and Heterogeneity of Objects

Image segmentation is the fundamental technology used in object-oriented remote sensing image analysis. To improve on the approach of object-oriented image analysis for the extraction of ground objects of interest, a segmentation algorithm for remote sensing images based on edge and heterogeneity of objects has been proposed in this paper. Canny algorithm combined with a mathematical morphology operator is used first to detect the edge information of a remote sensing image. Then, based on edge and heterogeneity of images, homogenous regions are obtained by Fractal Net Evolution approach. Finally, the edge information is used as a constrain in region merging, the final segmentation results are acquired. In this paper, QuickBird data from a certain area of Kunming have been tested and compared with a multi-resolution segmentation algorithm in the eCognition platform. The experiment results show that the method can segment the remote sensing images effectively and precisely.

OBJECT-BASED ANALYSIS OF AERIAL PHOTOGRAMMETRIC POINT CLOUD AND SPECTRAL DATA FOR LAND COVER MAPPING

Abstract. The acquisition of 3D point data with the use of both aerial laser scanning (ALS) and matching of aerial stereo images coupled with advances in image processing algorithms in the past years provide opportunities to map land cover types with better precision than before. The present study applies Object-Based Image Analysis (OBIA) to 3D point cloud data obtained from matching of stereo aerial images together with spectral data to map land cover types of the Nord-Trøndelag county of Norway. The multi-resolution segmentation algorithm of the Definiens eCognition™ software is used to segment the scenes into homogenous objects. The objects are then classified into different land cover types using rules created based on the definitions given for each land cover type by the Norwegian Forest and Landscape Institute. The quality of the land cover map was evaluated using data collected in the field as part of the Norwegian National Forest Inventory. The results show that the classification has an overall accuracy of about 80% and a kappa index of about 0.65. OBIA is found to be a suitable method for utilizing 3D remote sensing data for land cover mapping in an effort to replace manual delineation methods.

Identifying permafrost slope disturbance using multi-temporal optical satellite images and change detection techniques

Abstract Active layer detachments (ALDs) are a common form of permafrost slope disturbance that pose a serious risk for infrastructure and can impact environmental and ecological stability in Arctic regions. Effective recognition and detection of slope disturbances are critical for future hazard analysis. Historically, this has primarily been done through manual image interpretation and field mapping, both of which are cost-intensive. Semi-automatic detection techniques have been successfully applied in more temperate regions to identify slope failures, however, little work has been done to map permafrost disturbances. In this paper we present a methodology to detect and map ALDs using multi-temporal IKONOS satellite imagery in combination with vegetation index differencing and object-based image analysis, to semiautomatically identify landscape change associated with ALDs. A normalized difference vegetation index (NDVI) was computed for each of the two dates (2004 and 2010) and then subtracted generating a NDVI difference surface. Using areas where vegetation was removed as a proxy for the presence of ALDs, a multi-resolution segmentation algorithm was used to threshold the NDVI difference map into objects to demarcate regions of similarity (i.e., potential ALDs). To discriminate between disturbed and undisturbed zones a NDVI threshold was applied removing false positives. The thresholded image was then verified with a disturbance inventory collected from the field. These methods were successfully applied to the study area achieving 43% detection accuracy when identifying all ALDs. Morphometric characteristics were used to separate ALDs into two forms, elongate and compact, with accuracies assessed for each. Elongate ALDs, with a detection accuracy of 67%, are typically more destructive, moving substantially more material downslope over longer distances and posing a greater risk for infrastructure. By contrast, compact ALDs are associated with minimal downslope sliding distances (

Distinguishing wetland vegetation and channel features with object-based image segmentation

Mapping landscape features within wetlands using remote-sensing imagery is a persistent challenge due to the fine scale of wetland pattern variation and the low spectral contrast among plant species. Object-based image analysis (OBIA) is a promising approach for distinguishing wetland features, but systematic guidance for this use of OBIA is not presently available. A sensitivity analysis was tested using OBIA to distinguish vegetation zones, vegetation patches, and surface water channels in two intertidal salt marshes in southern San Francisco Bay. Optimal imagery sources and OBIA segmentation settings were determined from 348 sensitivity tests using the eCognition multiresolution segmentation algorithm. The optimal high-resolution (≤1 m) imagery choices were colour infrared (CIR) imagery to distinguish vegetation zones, CIR or red, green, blue (RGB) imagery to distinguish vegetation patches depending on species and season, and RGB imagery to distinguish surface water channels. High-resolution (1 m) lidar data did not help distinguish small surface water channels or other features. Optimal segmentation varied according to segmentation setting choices. Small vegetation patches and narrow channels were more recognizable using small scale parameter settings and coarse vegetation zones using larger scale parameter settings. The scale parameter served as a de facto lower bound to median segmented object size. Object smoothness/compactness weight settings had little effect. Wetland features were more recognizable using high colour/low shape weight settings. However, an experiment on a synthetic non-wetland image demonstrated that, colour information notwithstanding, segmentation results are still strongly affected by the selected image resolution, OBIA settings, and shape of the analysis region. Future wetland OBIA studies may benefit from strategically making imagery and segmentation setting choices based on these results; such systemization of future wetland OBIA approaches may also enhance study comparability.

Monitoring Seasonal Hydrological Dynamics of Minerotrophic Peatlands Using Multi-Date GeoEye-1 Very High Resolution Imagery and Object-Based Classification

The La Grande River watershed, located in the James Bay region (54°N, Quebec, Canada), is a major contributor to the production of hydroelectricity in the province. Peatlands cover up to 20% of the terrestrial environment in this region. Their hydrological behavior is not well understood. The present study is part of a multidisciplinary project which is aimed at analyzing the hydrological processes in these minerotrophic peatlands (fens) in order to provide effective monitoring tools to water managers. The objective of this study was to use VHR remote sensing data to understand the seasonal dynamics of the hydrology in fens. A series of 10 multispectral pan-sharpened GeoEye-1 images (with a spatial resolution of 40 cm) were acquired during the snow-free season (May to October) in 2009 and 2010, centered on two study sites in the Laforge sector (54°06'N; 72°30'W). These are two fens instrumented for continuous hydrometeorological monitoring (water level, discharge, precipitation, air temperature). An object-based classification procedure was set up and applied. It consisted of segmenting the imagery into objects using the multiresolution segmentation algorithm (MRIS) to delineate internal structures in the peatlands (aquatic, semi-aquatic, and terrestrial). Then, the objects were labeled using a fuzzy logic based algorithm. The overall classification accuracy of the 10 images was assessed to be 82%. The time series of the peatland mapping demonstrated the existence of important intra-seasonal spatial dynamics in the aquatic and semi-aquatic compartments. It was revealed that the dynamics amplitude depended on the morphological features of the fens. The observed spatial dynamics was also closely related to the evolution of the measured water levels.

Methods in general model localization

The aim of this study was to evaluate and test methods which could improve local estimates of a general model fitted to a large area. In the first three studies, the intention was to divide the study area into sub-areas that were as homogeneous as possible according to the residuals of the general model, and in the fourth study, the localization was based on the local neighborhood. According to spatial autocorrelation (SA), points closer together in space are more likely to be similar than those that are farther apart. Local indicators of SA (LISAs) test the similarity of data clusters. A LISA was calculated for every observation in the dataset, and together with the spatial position and residual of the global model, the data were segmented using two different methods: classification and regression trees (CART) and the multiresolution segmentation algorithm (MS) of the eCognition software. The general model was then re-fitted (localized) to the formed sub-areas. In kriging, the SA is modeled with a variogram, and the spatial correlation is a function of the distance (and direction) between the observation and the point of calculation. A general trend is corrected with the residual information of the neighborhood, whose size is controlled by the number of the nearest neighbors. Nearness is measured as Euclidian distance. With all methods, the root mean square errors (RMSEs) were lower, but with the methods that segmented the study area, the deviance in single localized RMSEs was wide. Therefore, an element capable of controlling the division or localization should be included in the segmentation-localization process. Kriging, on the other hand, provided stable estimates when the number of neighbors was sufficient (over 30), thus offering the best potential for further studies. Even CART could be combined with kriging or non-parametric methods, such as most similar neighbors (MSN).

Landslide intelligent prediction using object-oriented method

Landslides are one of the most serious geological disasters in the world and happen quite frequently in the Three Gorges. Landslide prediction is a very important measure of landslide prevention and cure in the Three Gorges. Traditional methods lack in sufficiently mining the various complex information from a landslide system. They often need much manual intervention and possess poor intelligence and accuracy. An intelligent method proposed in this paper for landslide prediction based on an object-oriented method and knowledge driving is hopeful to solve the above problem. The method adopted Landsat ETM+ images, 1:50,000 geological map and 1:10,000 relief map in the Three Gorges as the data origins. It firstly produced the key factors influencing landslide development and used multi-resolution segmentation algorithm to segment the image objects based on the key landslide factors of engineering rock group, reservoir water fluctuation, slope structure and slope level. Secondly, the method chose some sample objects and adopted the decision tree algorithm C5.0 to mine the landslide forecast criteria according to the factor values of each sample object. Finally, under knowledge driving the method classified the image objects and realized landslide susceptibility analysis and intelligent prediction in the Three Gorges. The method proposed in this paper is object-oriented. Results of a real-world example show that: (1) the object-oriented method possesses much more compact knowledge representation, higher efficiency, more continuous classifying result and higher prediction accuracy compared with the pixel-oriented method; (2) it possesses the overall accuracy of 87.64% and kappa coefficient of 0.8305 and is more accurate than the other seven methods (such as the pixel-oriented methods of Parallelpiped, Minimum Distance, Maximum Likelihood, Mahalanobis Distance, K-means and Isodata and the object-oriented method of Nearest Neighbor); (3) about 46.97% landslides lie in the high susceptibility region, 24.24% landslides lie in the moderate susceptibility region, 27.27% landslides lie in the low susceptibility region and 1.52% landslides lie in the very low susceptibility region. Therefore the method can effectively realize landslide susceptibility analysis and provides a new idea for landslide intelligent and accurate prediction.

Multiresolution video object extraction fitted to scalable wavelet-based object coding

To enable content based functionalities in video processing algorithms, decomposition of scenes into semantic objects is necessary. A semi-automatic Markov random field based multiresolution algorithm is presented for video object extraction in a complex scene. In the first frame, spatial segmentation and user intervention determine objects of interest. The specified objects are subsequently tracked in successive frames and newly appeared objects/regions are also detected. The video object extraction algorithm includes discrete wavelet transform decomposition multiresolution Markov random field (MRF)-based spatial segmentation with emphasis on border smoothness at different resolutions, and an MRF-based backward region classification that determines the tracked objects in the scene. Finally, a motion constraint, embedded in the region classifier, determines the newly appeared objects/regions and completes the proposed algorithm towards an efficient video segmentation algorithm. The results are applicable for generic segmentation applications, however the proposed multiresolution video segmentation algorithm supports scalable object-based wavelet coding in particular. Moreover, compared to traditional object extraction algorithms, it produces smoother and more visually pleasing shape masks at different resolutions. The proposed effective multiresolution video object extraction method allows for larger motion, better noise tolerance and less computational complexity.

Scalable multiresolution color image segmentation

This paper presents a novel multiresolution image segmentation method based on the discrete wavelet transform and Markov Random Field (MRF) modeling. A major contribution of this work is to add spatial scalability to the segmentation algorithm producing the same segmentation pattern at different resolutions. This property makes it suitable for scalable object-based wavelet coding. To optimize segmentation at all resolutions of the wavelet pyramid, with scalability constraint, a multiresolution analysis is incorporated into the objective function of the MRF segmentation algorithm. Examining the corresponding pixels at different resolutions simultaneously enables the algorithm to directly segment the images in the YUV or similar color spaces where luminance is in full resolution and chrominance components are at half resolution. Allowing for smoothness terms in the objective function at different resolutions improves border smoothness and creates visually more pleasing objects/regions, particularly at lower resolutions where down-sampling distortions are more visible. In addition to spatial scalability, the proposed algorithm outperforms the standard single and multiresolution segmentation algorithms, in both objective and subjective tests, yielding an effective segmentation that particularly supports scalable object-based wavelet coding.

A novel multiresolution fuzzy segmentation method on MR image

Multiresolution-based magnetic resonance (MR) image segmentation has attracted attention for its ability to capture rich information across scales compared with the conventional segmentation methods. In this paper, a new scale-space-based segmentation model is presented, where both the intra-scale and inter-scale properties are considered and formulated as two fuzzy energy functions. Meanwhile, a control parameter is introduced to adjust the contribution of the similarity character across scales and the clustering character within the scale. By minimizing the combined inter/intra energy function, the multiresolution fuzzy segmentation algorithm is derived. Then the coarse to fine leading segmentation is performed automatically and iteratively on a set of multiresolution images. The validity of the proposed algorithm is demonstrated by the test image and pathological MR images. Experiments show that by this approach the segmentation results, especially in the tumor area delineation, are more precise than those of the conventional fuzzy segmentation methods.

Study on segmentation and interpretation of multi-look polarimetric SAR images

The segmentation and interpretation of multi-look polarimetric synthetic aperture radar (SAR) images is studied. We first introduce a multi-look polarimetric whitening filter (MPWF) to reduce the speckle in multi-look polarimetric SAR images. Then, by utilizing the wavelet multiresolution approach to extract the texture information in different scales and the Markov random field (MRF) model to characterize the spatial constraints between pixels in each scale level, a multiresolution segmentation algorithm (MSA) to segment the speckle-reduced SAR images is presented. The MSA first segments the image at the lowest resolution level and then proceeds to progressively higher resolutions until individual pixels are well classified. An unsupervised step to estimate both the optimal number of texture classes and their model parameters is also included in the MSA so that the segmentation can be implemented without supervision. Finally, in order to interpret the results of the unsupervised segmentation and to understand the whole polarimetric SAR image, we develop an image interpretation approach which jointly utilizes the scattering mechanism identification and target decomposition approaches. Experimental results with the real-world multi-look polarimetric SAR image demonstrate the effectiveness of the segmentation and interpretation approaches.

Object-based image coding

Object-based compression methods describe images in terms of a set of regions (a partition), and of some information for each region to be used by the receiver to reconstruct the image (its texture). Different techniques can be used to define the partition as well as for coding it and its texture. In this paper we propose a general multi-resolution segmentation algorithm which can deal with many different types of images selecting the appropriated criteria. We also review most outstanding techniques for coding in this context.

European historical statistics and economic growth

There is increasing interest in developing automatic procedures to segment landscapes into soil spatial entities that replace conventional, expensive manual procedures for delineating and classifying soils. Geographic object-based image analysis (GEOBIA) partitions remote sensing imagery or digital elevation models into homogeneous image objects based on image segmentation. We used an object-based methodology for the detailed delineation and classification of soil types using digital maps of topography and vegetation as soil covariates, based on the Random Forests (RF) classifier. We compared the object-based method's results with those of a pixel-based classification using the same classifier. We used 18 digital elevation model derivatives and 5 remote sensing indices that were related to vegetation cover and soil. Using 171 soil profiles with their associated environmental variable values, the RF method was used to identify the most important soil type predictors for use in the segmentation process. A stack of raster-geodatasets corresponding to the selected predictors was segmented using a multi-resolution segmentation algorithm, which resulted in homogeneous objects related to soil types. These objects were further classified as soil types using the same method, RF. We also conducted a pixel-based classification using the same classifier and soil profiles, and the resulting maps were assessed in terms of their accuracy using 30% of the soil profiles for validation. We found that GEOBIA was an effective method for soil type mapping, and was superior to the pixel-based approach. The optimized object-based soil map had an overall accuracy of 58%, which was 10% higher than that of the optimized pixel-based map.