Integration of object-based and pixel-based classification for mapping mangroves with IKONOS imagery

IKONOS 1-m panchromatic and 4-m multispectral images were used to map mangroves in a study site located at Punta Galeta on the Caribbean coast of Panama. We hypothesized that spectral separability among mangrove species would be enhanced by taking the object as the basic spatial unit as opposed to the pixel. Three different classification methods were investigated: maximum likelihood classification (MLC) at the pixel level, nearest neighbour (NN) classification at the object level, and a hybrid classification that integrates the pixel and object-based methods (MLCNN). Specifically for object segmentation, which is the key step in object-based classification, we developed a new method to choose the optimal scale parameter with the aid of Bhattacharya Distance (BD), a well-known index of class separability in traditional pixel-based classification. A comparison of BD values at the pixel level and a series of larger scales not only supported our initial hypothesis, but also helped us to determine an optimal scale at which the segmented objects have the potential to achieve the best classification accuracy. Among the three classification methods, MLCNN achieved the best average accuracy of 91.4%. The merits and restrictions of pixel-based and object-based classification methods are discussed.     

Earthquake-induced landslide mapping in the western Himalayas using medium resolution ASTER imagery

On 8 October 2005, a devastating earthquake struck northern Pakistan and several parts of Pakistani- and Indian-controlled Kashmir. The severely hit areas lie in close proximity to the most tectonically active region of the western Himalayas. The earthquake destroyed close to 400 000 houses and over 75 000 people lost their lives. The intensity of the earthquake was such that it triggered widespread landslides, which caused considerable destruction of the area's forests, and blocked the mountain roads and rivers. Satellite imagery-based analysis can be effectively used to provide critical geologic information for determining the causes of earthquakes, mapping of faults and lineaments, as well as for hazards and damage assessment mapping. In this study, Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) imagery acquired on 27 October 2005 was analysed to assess earthquake-induced land cover changes in severely hit areas. Image analysis techniques including intensity hue and saturation (IHS), principal component analysis (PCA), Normalized Difference Vegetation Index (NDVI) and Iterative Self-Organizing Data Analysis Technique (ISODATA) were applied on VNIR–SWIR bands of ASTER for the purpose of identification and mapping of landslides triggered by the earthquake in areas of northern Pakistan and Pakistani-controlled Kashmir. The techniques proved effective in identifying freshly exposed surfaces on mountain slopes, landslide scars and debris deposits. Accurate identification and mapping of landslides and slope failures can play an important role in post-earthquake damage assessment and hazards mapping in earthquake-prone mountainous areas.     

Feature selection by genetic algorithms in object-based classification of IKONOS imagery for forest mapping in Flanders, Belgium

Obtaining detailed information about the amount of forest cover is an important issue for governmental policy and forest management. This paper presents a new approach to update the Flemish Forest Map using IKONOS imagery. The proposed method is a three-step object-oriented classification routine that involves the integration of 1) image segmentation, 2) feature selection by Genetic Algorithms (GAs) and 3) joint Neural Network (NN) based object-classification. The added value of feature selection and neural network combination is investigated. Results show that, with GA-feature selection, the mean classification accuracy (in terms of Kappa Index of Agreement) is significantly higher (p < 0.01) than without feature selection. On average, the summed output of 50 networks provided a significantly higher (p < 0.01) classification accuracy than the mean output of 50 individual networks. Finally, the proposed classification routine yields a significantly higher (p < 0.01) classification accuracy as compared with a strategy without feature selection and joint network output. In addition, the proposed method showed its potential when few training data were available.     

Object-based classification using SPOT-5 imagery for Moso bamboo forest mapping

This study proposed a multi-scale, object-based classification analysis of SPOT-5 imagery to map Moso bamboo forest. A three-level hierarchical network of image objects was developed through multi-scale segmentation. By combining spectral and textural properties, both the classification tree and nearest neighbour classifiers were used to classify the image objects at Level 2 in the three-level object hierarchy. The feature selection results showed that most of the object features were related to the spectral properties for both the classification tree and nearest neighbour classifiers. Contextual information characterized by the composition of classified image objects using the class-related features assisted the detection of shadow areas at Levels 1 and 3. Better classification results were achieved using the nearest neighbour algorithm, with both the producer’s and user’s accuracy higher than 90% for Moso bamboo and an overall accuracy of over 85%. The object-based approach toward incorporating textural and contextual information in classification sequence at various scales shows promise in the analysis of forest ecosystems of a complex nature.     

Advances in classification for land cover mapping using SPOT HRV imagery

Abstract

High-resolution data from the HRV (High Resolution Visible) sensors onboard the SPOT-1 satellite have been utilized for mapping semi-natural and agricultural land cover using automated digital image classification algorithms. Two methods for improving classification performance are discussed. The first technique involves the use of digital terrain information to reduce the effects of topography on spectral information while the second technique involves the classification of land-cover types using training data derived from spectral feature space. Test areas in Snowdonia and the Somerset Levels were used to evaluate the methodology and promising results were achieved. However, the low classification accuracies obtained suggest that spectral classification alone is not a suitable tool to use in the mapping of semi-natural cover types.     

GeoEye-1 and WorldView-2 pan-sharpened imagery for object-based classification in urban environments

The latest breed of very high resolution (VHR) commercial satellites opens new possibilities for cartographic and remote-sensing applications. In fact, one of the most common applications of remote-sensing images is the extraction of land-cover information for digital image base maps by means of classification techniques. The aim of the study was to compare the potential classification accuracy provided by pan-sharpened orthoimages from both GeoEye-1 and WorldView-2 (WV2) VHR satellites over urban environments. The influence on the supervised classification accuracy was evaluated by means of an object-based statistical analysis regarding three main factors: (i) sensor used; (ii) sets of image object (IO) features used for classification considering spectral, geometry, texture, and elevation features; and (iii) size of training samples to feed the classifier (nearest neighbour (NN)). The new spectral bands of WV2 (Coastal, Yellow, Red Edge, and Near Infrared-2) did not improve the benchmark established from GeoEye-1. The best overall accuracy for GeoEye-1 (close to 89%) was attained by using together spectral and elevation features, whereas the highest overall accuracy for WV2 (83%) was achieved by adding textural features to the previous ones. In the case of buildings classification, the normalized digital surface model computed from light detection and ranging data was the most valuable feature, achieving producer's and user's accuracies close to 95% and 91% for GeoEye-1 and VW2, respectively. Last but not least and regarding the size of the training samples, the rule of ‘the larger the better' was true but, based on statistical analysis, the ideal choice would be variable depending on both each satellite and target class. In short, 20 training IOs per class would be enough if the NN classifier was applied on pan-sharpened orthoimages from both GeoEye-1 and WV2.     

Per-pixel vs. object-based classification of urban land cover extraction using high spatial resolution imagery

In using traditional digital classification algorithms, a researcher typically encounters serious issues in identifying urban land cover classes employing high resolution data. A normal approach is to use spectral information alone and ignore spatial information and a group of pixels that need to be considered together as an object. We used QuickBird image data over a central region in the city of Phoenix, Arizona to examine if an object-based classifier can accurately identify urban classes. To demonstrate if spectral information alone is practical in urban classification, we used spectra of the selected classes from randomly selected points to examine if they can be effectively discriminated. The overall accuracy based on spectral information alone reached only about 63.33%. We employed five different classification procedures with the object-based paradigm that separates spatially and spectrally similar pixels at different scales. The classifiers to assign land covers to segmented objects used in the study include membership functions and the nearest neighbor classifier. The object-based classifier achieved a high overall accuracy (90.40%), whereas the most commonly used decision rule, namely maximum likelihood classifier, produced a lower overall accuracy (67.60%). This study demonstrates that the object-based classifier is a significantly better approach than the classical per-pixel classifiers. Further, this study reviews application of different parameters for segmentation and classification, combined use of composite and original bands, selection of different scale levels, and choice of classifiers. Strengths and weaknesses of the object-based prototype are presented and we provide suggestions to avoid or minimize uncertainties and limitations associated with the approach.     

Synergistic use of QuickBird multispectral imagery and LIDAR data for object-based forest species classification

This study evaluated the synergistic use of high spatial resolution multispectral imagery (i.e., QuickBird, 2.4 m) and low-posting-density LIDAR data (3 m) for forest species classification using an object-based approach. The integration of QuickBird multispectral imagery and LIDAR data was considered during image segmentation and the subsequent object-based classification. Three segmentation schemes were examined: (1) segmentation based solely on the spectral image layers; (2) segmentation based solely on LIDAR-derived layers; and (3) segmentation based on both the spectral and LIDAR-derived layers. For each segmentation scheme, objects were generated at twelve different scales in order to determine optimal scale parameters. Six categories of classification metrics were generated for each object based on spectral data alone, LIDAR data alone and the combination of both data sources. Machine learning decision trees were used to build classification rule sets. Quantitative segmentation quality assessment and classification accuracy results showed the integration of spectral and LIDAR data, in both image segmentation and object-based classification, improved the forest classification compared to using either data source independently. Better segmentation quality led to higher classification accuracy. The highest classification accuracy (Kappa = 91.6%) was acquired when using both spectral- and LIDAR-derived metrics based on objects segmented from both spectral and LIDAR layers at scale parameter 250, where best segmentation quality was achieved. Optimal scales were analyzed for each segmentation-classification scheme. Statistical analysis of classification accuracies at different scales revealed that there was a range of optimal scales that provided statistically similar accuracy.     

Efficiency and accuracy of per-field classification for operational crop mapping

A crop map of The Netherlands was created using a methodology that integrates multi-temporal and multi-sensor satellite imagery, statistical data on crop area and parcel boundaries from a 1?:?10?000 digital topographic map. In the first phase a crop field database was created by extracting static parcel boundaries from the digital topographic map and by adding dynamic crop boundaries using on-screen digitizing. In the next phase the crop type was determined from the spectral and phenological properties of each field. The resulting crop map has an accuracy larger than 80% for most individual crops and an overall accuracy of 90%. By comparing cost and man-hours it was demonstrated that per-field classification is more efficient than per-pixel classification and decreased the effort for classification from 1500 to 500 man-hours, but the effort for creating the crop field database was estimated at 2300 man-hours. The use of image segmentation techniques for deriving the crop field database was discussed. It was concluded that image segmentation cannot replace the use of a large-scale topographic map but, in the future, image segmentation may be used to map the dynamic crop boundaries within the topographic parcels.     

The application of discriminant analysis for mapping cereals and pasture using object-based features

High mapping accuracies occur where crops differ spectrally (e.g.>90.0%; canola, corn, soybeans) and vice versa (e.g. <75.0%; cereals and pasture). Developing improved mapping methods has been an ongoing priority of Agriculture and Agri-Food Canada (AAFC) remote-sensing science. To this end, this study tests a data-driven object-based classification method using Discriminant Analysis (DA) method for mapping cereals and pasture from satellite data. In this approach, variables (number >400) derived from the image segmentation and object-based feature extraction of multi-date and multi-band optical (RapidEye) and microwave (RADARSAT-2) imagery were applied in a data-driven approach. We use in situ and satellite information collected over two study sites with different levels of heterogeneity (Winnipeg, Brandon) situated in the Canadian Prairies during the 2013 growing season to assess: (a) the type of DA model that most accurately classifies the cereals and pasture cover classes; and (b) how the classification accuracies obtained by the application of this DA model compare to those obtained from more traditional Maximum Likelihood (ML), Decision Tree (DT), and Random Forest (RF) classifications. We found that our DA-based approach was able to map cereals and pastures at our two study sites with the highest accuracies, but these accuracies did not improve significantly with the use of more complex DA model (including priori classification probabilities, more input principle components (PCs), the use of weights proportional to field area). Our results are encouraging for the wider application of the data-driven pre-processing of the inputs to the image classification by DA.     

Comparison of pixel- and object-based classification in land cover change mapping

Land use/land cover (LULC) change occurs when humans alter the landscape, and this leads to increasing loss, fragmentation and spatial simplification of habitat. Many fields of study require monitoring of LULC change at a variety of scales. LULC change assessment is dependent upon high-quality input data, most often from remote sensing-derived products such as thematic maps. This research compares pixel- and object-based classifications of Landsat Thematic Mapper (TM) data for mapping and analysis of LULC change in the mixed land use region of eastern Ontario for the period 1995–2005. For single date thematic maps of 10 LULC classes, quantitative and visual analyses showed no significant accuracy difference between the two methods. The object-based method produced thematic maps with more uniform and meaningful LULC objects, but it suffered from absorption of small rare classes into larger objects and the incapability of spatial parameters (e.g. object shape) to contribute to class discrimination. Despite the similar map accuracies produced by the two methods, temporal change maps produced using post-classification comparison (PCC) and analysed using intensive visual analysis of errors of omission and commission revealed that the object-based maps depicted change more accurately than maximum likelihood classification (MLC)-derived change maps.     

Land-cover classification of an intra-urban environment using high-resolution images and object-based image analysis

Detailed, up-to-date information on intra-urban land cover is important for urban planning and management. Differentiation between permeable and impermeable land, for instance, provides data for surface run-off estimates and flood prevention, whereas identification of vegetated areas enables studies of urban micro-climates. In place of maps, high-resolution images, such as those from the satellites IKONOS II, Quickbird, Orbview and WorldView II, can be used after processing. Object-based image analysis (OBIA) is a well-established method for classifying high-resolution images of urban areas. Despite the large number of previous studies of OBIA in the context of intra-urban analysis, there are many issues in this area that are still open to discussion and resolution. Intra-urban analysis using OBIA can be lengthy and complex because of the processing difficulties related to image segmentation, the large number of object attributes to be resolved and the many different methods needed to classify various image objects. To overcome these issues, we performed an experiment consisting of land-cover mapping based on an OBIA approach using an IKONOS II image of a southern sector of São José dos Campos city (covering an area of 12 km² with 50 neighbourhoods), which is located in São Paulo State in south-eastern Brazil. This area contains various occupation and land-use patterns, and it therefore contains a wide range of intra-urban targets. To generate the land-cover map, we proposed an OBIA-based processing framework that combines multi-resolution segmentation, data mining and hierarchical network techniques. The intra-urban land-cover map was then evaluated through an object-based error matrix, and classification accuracy indices were obtained. The final classification, with 11 classes, achieved a global accuracy of 71.91%.     

Integrating LiDAR and high-resolution imagery for object-based mapping of forest habitats in a heterogeneous temperate forest landscape

ABSTRACT

Land-cover mapping (LCM) at a fine scale would be useful for forest management across heterogeneous natural landscapes. However, the heterogeneity of land covers at such scales results in complex spectral and textural properties that hinder the applicability of LCM. Besides, the method suffers from, e.g. inconsistent representation of different land-cover types, lack of sufficient and balanced training samples, and instability of classifiers trained by a high number of predictor variables. Even well-known object-based classification approaches are challenged with an objective evaluation of segmentation outputs. Here we classified partially ambiguous land-cover types across heterogeneous forest landscapes in the Bavarian Forest National Park (Germany) by combining metrics from airborne light detection and ranging (LiDAR) and colour infrared (CIR) imagery data and a random forest classifier implemented in an object-based paradigm. We evaluated the segmentation results by creating a global quality score based on inter- and intra-measurements of variance and the number of segments. Selected segmentation outputs were combined with balanced training samples to run the classification algorithm based on representative blocks within the national park. The entire processing chain was implemented in an open-source domain. The final segmentation consisted of LiDAR-based height, image-based Normalized Difference Vegetation Index (NDVI) and red band, with 20 cluster seeds and a minimum segment size of 40 pixels. In the classification, the most important variables included the height of the top layer, NDVI, Enhanced Vegetation Index (EVI) and Green–Red Vegetation Index (GRVI). The average values of 500 random forest runs indicated an overall accuracy of 86.6% and an estimated Cohen’s kappa coefficient of 85.2%, with different probabilities of correct classification for land-cover classes. Mature deciduous, standing deadwood, fallen deadwood, meadow, and bare soil classes were classified most accurately, whereas classification of young coniferous, intermediate-age coniferous, mature coniferous, young deciduous, and intermediate-age deciduous were associated with the highest uncertainties. Our methodology is sufficiently robust to be applied to other similarly structured sites across temperate forested landscapes. The versatility of the method is partially guaranteed by the proposed segmentation quality score, which satisfactorily corrects under- and over-segmentation.     

Extraction of landslide-related factors from ASTER imagery and its application to landslide susceptibility mapping

The aim of this study is to extract landslide-related factors from remote-sensing data, such as Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) satellite imagery, and to examine their applicability to landslide susceptibility near Boun, Korea, using a geographic information system (GIS). Landslide was mapped from interpretation of aerial photographs and field surveying. Factors that influence landslide occurrence were extracted from ASTER imagery. The slope, aspect and curvature were calculated from the digital elevation model (DEM) with 25.77 m root mean square error (RMSE), which was derived from ASTER imagery. Lineaments, land-cover and normalized difference vegetation index (NDVI) layers were also estimated from ASTER imagery. Landslide-susceptible areas were analysed and mapped using the occurrence factors by a frequency ratio and logistic regression model. Validation results were 84.78% in frequency ratio and 84.20% in logistic regression prediction accuracy for the susceptibility map with respect to ground-truth data.     

Active learning for object-based image classification using predefined training objects

Object-based image analysis (OBIA) is a new remote-sensing-based image processing technology that has become popular in recent years. In spite of its remarkable advantages, the segmentation results that it generates feature a large number of mixed objects owing to the limitations of OBIA segmentation technology. The mixed objects directly influence the acquisition of training samples and the labelling of objects and thus affect the stability of classification performance. In light of this issue, this article evaluates the influence of classification uncertainty on classification performance and proposes a sampling strategy based on active learning. This sampling strategy is novel in two ways: (1) information entropy is used to evaluate the classification uncertainty of segmented objects; all segmented objects are classified as having zero or non-zero entropies, and the latter are arranged in terms of decreasing entropy. (2) Based on an evaluation of the influence of classification uncertainty on classification performance, an active learning technology is developed. A certain proportion of zero-entropy objects is acquired via random sampling used as seed training samples for active learning, non-zero-entropy objects are used as a candidate set for active learning, and the entropy query-by-bagging (EQB) algorithm is used to conduct active learning to acquire optimal training samples. In this study, three groups of high-resolution images were tested. The test results show that zero-entropy and non-zero-entropy objects are indispensable to the classifier, where the optimal range of the ratio of combination of the two is between 0.2 and 0.6. Moreover, the proposed sampling strategy can effectively improve the stability and accuracy of classification.     

Per-field classification: an example using SPOT HRV imagery

Abstract

The classification of land cover on remotely sensed imagery is usually undertaken in a per-pixel format within an image file or in a per-field format within a non-image file. The latter is more accurate but does not produce an image output and is not readily input to a vector-based geographical information system. We propose setting the pixels in each field to a representative statistic for that field and then using a per-pixel classifier to perform a per-field classification in an image file. This procedure was evaluated using SPOT high resolution visible (HRV) imagery. The highest classification accuracy of 62.1 per cent (12 class) was achieved using measures of prior probabilities and image texture within the proposed per-field format.     

Object-based image classification for burned area mapping of Creus Cape, Spain, using NOAA-AVHRR imagery

Due to the ability of the NOAA-AVHRR sensor to cover a wide area and its high temporal frequency, it is possible to quickly obtain a general overview of the prevailing situation over a large area of terrain and, more specifically, quickly assess the damage caused by a recent large forest fire by mapping the extent of the burned area. The aim of this work was to map a large forest fire that recently took place on the Spanish Mediterranean coast using innovative image classification techniques and low spatial resolution imagery. The methodology involved developing an object-based classification model using spectral as well as contextual object information. The burned area map resulting from the image classification was compared with the fire perimeter provided by the Catalan Environmental Department in terms of spatial overlap and size in order to determine to what extent they were compatible. Results of the comparison indicated a high degree (≈90%) of spatial agreement. The total burned area of the classified image was found to be 6900 ha, compared to a fire perimeter of 6000 ha produced by the Catalan Environmental Department. It was concluded that, although the object-oriented classification approach was capable of affording very promising results when mapping a recent burn on the Spanish Mediterranean coast, the method in question required further assessment to ascertain its ability to map other burned areas in the Mediterranean.     

Automatic system for operational traffic monitoring using very-high-resolution satellite imagery

Vehicle detection from very-high-resolution satellite imagery has received increasing interest during the last few years. In this article, we propose an automatic system for operational traffic monitoring using very-high-resolution optical satellite imagery (0.5–0.6 m resolution) of small highways with low traffic density and a range of different illumination conditions, including cloud-shadowed, hazy, and partially cloudy conditions. The proposed system includes cloud and cloud shadow detection, road detection, and vehicle detection, classification, and counting. The main part of the system is vehicle detection, which is constructed using an elliptical blob detection strategy followed by region growing and feature extraction steps. Vehicular objects are separated from non-vehicular objects using a K-nearest-neighbour classifier, with various classical features used for pattern recognition, as well as some proposed application-specific features, and are also classified according to vehicle size. The fully automatic processing chain has been validated on a selection of satellite scenes from different parts of Norway, including imagery with large amounts of cloud, fog, cloud shadows, and similar conditions that complicate image interpretation. The overall vehicle detection rate was 85.4% and the false detection rate was 9.2%. Overall, this demonstrates the potential of operational traffic monitoring using very-high-resolution satellites.     

Observation of submarine sand waves using ASTER stereo sun glitter imagery

Signatures of sun glitter images strongly depend on the viewing angle. Stereo observation can provide more valuable information than observation from one angle. In this work, the sun glitter patterns caused by submarine sand waves were studied using stereo images from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), which operates a high spatial resolution (15 m) optical and near-infrared sensor on board the Terra satellite. The previous imaging geometrical model has been improved, and the difference in the viewing angle among pixels is considered. We found that brightness reversal occurs in some stereo imagery of submarine sand waves. Based on the imaging geometry model, an interaction model of current topography, and a sun glitter radiance transfer model, a simulation model was developed for sun glitter from submarine sand waves at multiple viewing angles. The cases of nadir-looking and backward-looking were simulated with the model. The results show the following differences between the two viewing angles, which have also been observed in the ASTER images. Both tendency and extent of the simulated radiance are in good agreement with those in actual images. In the nadir-looking view, obvious differences have been observed between the normalized significant radiance from smooth and rough facets. This difference increases with increasing viewing angle, to a peak, and then decreases. However, in most cases, the difference in the back view is smaller and shows an opposite tendency regarding the viewing angle from the nadir-looking view. Thus, the sand-wave characteristics in the nadir-looking images seem to be more enhanced than that in the backward-looking images.