OBJECT BASED IMAGE ANALYSIS COMBINING HIGH SPATIAL RESOLUTION IMAGERY AND LASER POINT CLOUDS FOR URBAN LAND COVER

Abstract. With the rapid developments of the sensor technology, high spatial resolution imagery and airborne Lidar point clouds can be captured nowadays, which make classification, extraction, evaluation and analysis of a broad range of object features available. High resolution imagery, Lidar dataset and parcel map can be widely used for classification as information carriers. Therefore, refinement of objects classification is made possible for the urban land cover. The paper presents an approach to object based image analysis (OBIA) combing high spatial resolution imagery and airborne Lidar point clouds. The advanced workflow for urban land cover is designed with four components. Firstly, colour-infrared TrueOrtho photo and laser point clouds were pre-processed to derive the parcel map of water bodies and nDSM respectively. Secondly, image objects are created via multi-resolution image segmentation integrating scale parameter, the colour and shape properties with compactness criterion. Image can be subdivided into separate object regions. Thirdly, image objects classification is performed on the basis of segmentation and a rule set of knowledge decision tree. These objects imagery are classified into six classes such as water bodies, low vegetation/grass, tree, low building, high building and road. Finally, in order to assess the validity of the classification results for six classes, accuracy assessment is performed through comparing randomly distributed reference points of TrueOrtho imagery with the classification results, forming the confusion matrix and calculating overall accuracy and Kappa coefficient. The study area focuses on test site Vaihingen/Enz and a patch of test datasets comes from the benchmark of ISPRS WG III/4 test project. The classification results show higher overall accuracy for most types of urban land cover. Overall accuracy is 89.5% and Kappa coefficient equals to 0.865. The OBIA approach provides an effective and convenient way to combine high resolution imagery and Lidar ancillary data for classification of urban land cover.

Improvement in remote sensing techniques in spatial/spectral resolution strengthens their applicability for urban environmental study. Unfortunately, high spatial resolution imagery also increases internal variability in land cover units and can cause a ‘salt-and-pepper’ effect, resulting in decreased accuracy using pixel-based classification results. Region-based classification techniques, using an image object (IO) rather than a pixel as a classification unit, appear to hold promise as a method for overcoming this problem. Using IKONOS high spatial resolution imagery, we examined whether the IO technique could significantly improve classification accuracy compared to the pixel-based method when applied to urban land cover mapping in Tampa Bay, FL, USA. We further compared the performance of an artificial neural network (ANN) and a minimum distance classifier (MDC) in urban detailed land cover classification and evaluated whether the classification accuracy was affected by the number of extracted IO features. Our analysis methods included IKONOS image data calibration, data fusion with the pansharpening (PS) process, Hue–Intensity–Saturation (HIS) transferred indices and textural feature extraction, and feature selection using a stepwise discriminant analysis (SDA). The classification results were evaluated with visually interpreted data from high-resolution (0.3 m) digital aerial photographs. Our results indicate a statistically significant difference in classification accuracy between pixel- and object-based techniques; ANN outperforms MDC as an object-based classifier; and the use of more features (27 vs. 9 features) increases the IO classification accuracy, although the increase is statistically significant for the MDC but not for the ANN.

In these days, the demand on methodologies and processing schemes for urban application or analysis using high spatial resolution imagery is increasing. In fact, most methodologies for intermediate or coarse spatial resolution imagery can also be applied for high-resolution one, and some processing methods among them show practical usefulness over those imagery data. In use of high spatial resolution imagery, change detection based on accurate extraction of target features of interests in a certain region including complex urban area or inaccessible area is of importance. For this change detection approach, two processing techniques such as homomorphic filtering and morphological operations are considered in this study. After applying other basic image rectification processing, homomorphic filtering which combines non-linear mapping function and spatial frequency filtering is helpful to reveal distinct characteristics for some complicated and multi-associated features on high spatial resolution imagery, and then morphological operations are fast and efficient processing method to detect change with homomorphic filtered and multi-temporal imagery datasets. As a result, this approach with both techniques is thought to be one of the applicable schemes for the change detection problem with high spatial resolution imagery, in the aspects of increased accuracy and automatic extraction of target features.

Fusion of high spatial resolution WorldView-2 imagery and LiDAR pseudo-waveform for object-based image analysis

Mapping landcover in cities is essential for urban ecology and landuse management, yet urban landcover is often highly heterogeneous at fine spatial scales. Pixel-based approaches are shown to be less successful for effectively mapping urban landcover due to high heterogeneity, with relatively low accuracies reported despite the use of high spatial resolution optical imagery. Alternatively, geographic object-based image analysis (GEOBIA) has yielded higher accuracies across a range of urban applications. We combine three-dimensional (3-D) information from airborne laser scanning (ALS) data with RapidEye high-spatial-resolution imagery in a GEOBIA approach to classify urban landcover in a large metropolitan region in Vancouver, Canada. Results indicate that 12 urban classes could be accurately mapped at 2-m spatial resolution across 150,000 ha with an overall accuracy of 88% (kappa 0.87). Though 5-m RapidEye multispectral pixels were often mixed in heterogeneous urban areas, the additional insight provided by the 3-D ALS information enabled accurate classification of fine spatial objects such as street trees and single-family dwellings.

The Remote Sensing Technologies Center (RSTC) at Mississippi State University (MSU) is investigating the application of high-resolution imagery in the areas of precision agriculture, forest resource management, and environmental assessments of transportation systems. For precision agriculture, the focus of the research is to develop the expertise to utilize this imagery to facilitate management decisions at the farm level. Decision making at the farm level requires high spatial resolution imagery (on the order of 1 meter or less) in order to develop prescriptions for nutrient supplements, integrated pest management (insects, weeds), irrigation, and to serve as a harvesting aid. This degree of analysis of the plant and soil also requires that higher spectral resolution be available to ascertain the nuances in the target vegetation's reflectance and to be able to relate this to specific agricultural problems. In addition, decisions at the farm level must be made within a few days of onset of a problem. Therefore, revisit frequencies for precision agriculture are on the order of 10-14 days. The requirement for high spatial, spectral, and temporal resolution imagery creates special image processing and analysis challenges. These include-data management, geometric and radiometric corrections, ground control points and georeferencing, and data analysis via such approaches as data fusion. The purpose of this paper is twofold. First, to illustrate the diversity of high spatial, spectral, and temporal resolution imagery required for precision agriculture applications. Second, to offer the opportunity to other data fusion researchers to join the author of this paper in developing new data fusion techniques for high-resolution imagery.

ABSTRACTFocusing on the semi-arid and highly disturbed landscape of San Clemente Island (SCI), California, we test the effectiveness of incorporating a hierarchical object-based image analysis (OBIA) approach with high-spatial resolution imagery and canopy height surfaces derived from light detection and ranging (lidar) data for mapping vegetation communities. The hierarchical approach entailed segmentation and classification of fine-scale patches of vegetation growth forms and bare ground, with shrub species identified, and a coarser-scale segmentation and classification to generate vegetation community maps. Such maps were generated for two areas of interest on SCI, with and without vegetation canopy height data as input, primarily to determine the effectiveness of such structural data on mapping accuracy. Overall accuracy is highest for the vegetation community map derived by integrating airborne visible and near-infrared imagery having very high spatial resolution with the lidar-derived canopy height data. The results demonstrate the utility of the hierarchical OBIA approach for mapping vegetation with very high spatial resolution imagery, and emphasizes the advantage of both multi-scale analysis and digital surface data for accurately mapping vegetation communities within highly disturbed landscapes.

Detailed urban land cover classification using very high resolution (VHR) imagery is a challenging task, partly due to the spectral similarity between land cover classes and high spectral variability within individual classes. In this paper, shape features extracted from segmented objects using three types of invariant moments, namely Hu’s moments, Zernike moments, and wavelet moments, were applied to urban land cover classification using VHR imagery. The performance of invariant moments was evaluated by comparison with the performance of the object-based classification utilizing spectral information alone. The results showed that overall accuracy improved significantly when shape features defined by invariant moments were combined with spectral bands in urban land cover classification. In particular, significant improvements in accuracy for individual classes were achieved for those classes which have similar spectral features but different shape features, such as building, road, and other impervious surface. The study also found that selection of segmentation results is crucial in extraction of shape features for improved classification.

Understanding the often-heterogeneous land cover in urban areas is critical for, among other things, environmental monitoring, spatial planning, and enforcement. Recently, several earth observation satellites were developed with an enhanced spatial resolution that provides for precise and detailed representations of image objects. Morphological image analysis techniques provide useful tools for extracting spatial features from high-resolution, remotely sensed images. This study investigated the efficacy of mathematical morphological (MM) techniques in the land cover classification of a heterogeneous urban landscape using very high-resolution pan-sharpened Pleiades imagery. Specifically, the study evaluated two morphological profiles (MP) techniques (i.e., concatenation of morphological profiles (CMPs) and multi-morphological profiles (MMPs)) in the classification of a heterogeneous urban land cover. The overall accuracies for CMP were 83.14% and 83.19% over the two study areas. Similarly, the MMP overall accuracies were 84.42% and 84.08% for the two study sites. The study concluded that CMP and MMP can greatly improve the classification of heterogeneous landscapes that typify urban areas by effectively representing the structural landscape information necessary for discriminating related land cover classes. In general, similar and visually acceptable results were produced for land cover classification using either CMP or MMP image analysis techniques

Impervious surfaces are manmade surfaces which are highly resistant to infiltration of water. Previous attempts to classify impervious surfaces from high spatial resolution imagery with pixel-based techniques have proven to be unsuitable for automated classification because of its high spectral variability and complex land covers in urban areas. Accurate and rapid classification of impervious surfaces would help in emergency management after extreme events like flooding, earthquakes, fires, tsunami, and hurricanes, by providing quick estimates and updated maps for emergency response. The objectives of this study were to: (1) compare classification accuracy between pixel-based and OBIA methods, (2) examine whether the object-based image analysis (OBIA) could better detect urban impervious surfaces, and (3) develop an automated, generalized OBIA classification method for impervious surfaces. This study analyzed urban impervious surfaces using a 1-meter spatial resolution, four band Digital Orthophoto Quarter Quad (DOQQ) aerial imagery of downtown New Orleans, Louisiana taken as part of post Hurricane Katrina and Rita dataset. The study compared the traditional pixel-based classification with four variations of the rule-based OBIA approach for classification accuracy. A four-class classification scheme was used for the analysis, including impervious surfaces, vegetation, shadow, and water. The results show that OBIA accuracy ranges from 85.33% through 91.41% compared with 80.67% classification accuracy from using the pixel-based approach. OBIA rule-based method 4 utilizing a multi-resolution segmentation approach and derived spectral indices such as Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and the Spectral Shape Index (SSI) was the best method, yielding a 91.41% classification accuracy. OBIA rule-based method 4 can be automated and generalized for multiple study areas. A test of the segmentation parameters show that parameter values of scale ≤ 20, color/shape ranging from 0.1 - 0.3, and compactness/smoothness ranging from 0.4 - 0.6 yielded the highest classification accuracies. These results show that the developed OBIA method was accurate, generalizable, and capable of automation for the classification of urban impervious surfaces.

To extract GIS features from high spatial resolution imagery is an important task in remote sensing applications. However, traditional pixel-based classification methods, which were developed in the era of 10-100 m ground pixel size imagery, cannot exploit the advantages of new images provided by IKONOS and QuickBird. To successfully extract various land covers from high resolution imagery, a Target-Clustering Fusion (TCF) system is presented in this work. Compared to the conventional classification methods that typically produce more salt-and-pepper-like results, the proposed TCF system can preserve detailed spatial information on each classified target related to its neighbors.

Maintaining accurate, current, rural road network data: An extraction and updating routine using RapidEye, participatory GIS and deep learning

The advent of new high spatial resolution optical satellite imagery has greatly increased our ability to monitor land cover changes from space. Satellite observations are carried out regularly and continuously, and provide a great deal of insight into the temporal changes of land cover use. High spatial resolution imagery better resolves the details of these changes and makes it possible to overcome the "mixed-pixel" problem that is inherent with more moderate resolution satellite sensors. At the same time, high-resolution imagery presents a new challenge over other satellite systems, in that a relatively large amount of data must be analyzed and corrected for registration and classification errors to identify the land cover changes. To obtain the accuracies that are required by many applications to large areas, very extensive manual work is commonly required to remove the classification errors that are introduced by most methods. To improve on this situation, we have developed a new method for land surface change detection that greatly reduces the human effort that is needed to remove the errors that occur with many classification methods that are applied to high-resolution imagery. This change detection algorithm is based on neural networks, and it is able to exploit in parallel both the multiband and the multitemporal data to discriminate between real changes and false alarms. In general, the classification errors are reduced by a factor of 2-3 using our new method over a simple postclassification comparison based on a neural-network classification of the same images.

Effect of coregistration error on patchy target detection using high-resolution imagery

Urban land cover and land use mapping plays an important role in urban planning and management. In this paper, novel multi-scale deep learning models, namely ASPP-Unet and ResASPP-Unet are proposed for urban land cover classification based on very high resolution (VHR) satellite imagery. The proposed ASPP-Unet model consists of a contracting path which extracts the high-level features, and an expansive path, which up-samples the features to create a high-resolution output. The atrous spatial pyramid pooling (ASPP) technique is utilized in the bottom layer in order to incorporate multi-scale deep features into a discriminative feature. The ResASPP-Unet model further improves the architecture by replacing each layer with residual unit. The models were trained and tested based on WorldView-2 (WV2) and WorldView-3 (WV3) imageries over the city of Beijing. Model parameters including layer depth and the number of initial feature maps (IFMs) as well as the input image bands were evaluated in terms of their impact on the model performances. It is shown that the ResASPP-Unet model with 11 layers and 64 IFMs based on 8-band WV2 imagery produced the highest classification accuracy (87.1% for WV2 imagery and 84.0% for WV3 imagery). The ASPP-Unet model with the same parameter setting produced slightly lower accuracy, with overall accuracy of 85.2% for WV2 imagery and 83.2% for WV3 imagery. Overall, the proposed models outperformed the state-of-the-art models, e.g., U-Net, convolutional neural network (CNN) and Support Vector Machine (SVM) model over both WV2 and WV3 images, and yielded robust and efficient urban land cover classification results.