Analysis of terrain using multispectral images

2018 9th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS), 2018

We focus on the automatic 3D terrain segmentation problem using hyperspectral shortwave IR (HS-SWIR) imagery and 3D Digital Elevation Models (DEM). The datasets were independently collected, and metadata for the HS-SWIR dataset are unavailable. We explore an overall slope of the SWIR spectrum that correlates with the presence of moisture in soil to propose a band ratio test to be used as a proxy for soil moisture content to distinguish two broad classes of objects: live vegetation from impermeable manmade surface. We show that image based localization techniques combined with the Optimal Randomized RANdom Sample Consensus (RANSAC) algorithm achieve precise spatial matches between HS-SWIR data of a portion of downtown Los Angeles (LA (USA)) and the Visible image of a georegistered 3D DEM, covering a wider-area of LA. Our spectral-elevation rule based approach yields an overall accuracy of 97.7%, segmenting the object classes into buildings, houses, trees, grass, and roads/parking lots.

International Journal of Remote Sensing, 2002

This paper examines the in uence of multisensor data fusion on the automatic extraction of topographic objects from SPOT panchromatic imagery. The suitability of various grey level co-occurrence based texture measures, as well as diVerent pixel windows is also investigated. It is observed that best results are obtained with a 3×3 pixel window and the texture measure homogeneity. The synthetic texture image derived together with a Landsat Thematic Mapper (TM) imagery are then fused to the SPOT data using the additional channel concept. The object feature base is expanded to include both spectral and spatial features. A maximum likelihood classi cation approach is then applied. It is demonstrated that the segmentation of topographic objects is signi cantly improved by fusing the multispectral and texture information.

2005 IEEE International Conference on Multimedia and Expo

Typically, high-resolution remote sensing (HRRS) images contain a high level noise as well as possess different texture scales. As a result, existing image segmentation approaches are not suitable to HRRS imagery. In this paper, we have presented an unsupervised texture-based segmentation algorithm suitable for HRRS images, by extending the local binary pattern texture features and the lossless wavelet transform. Our experimental results using USGS 1ft orthoimagery show a significant improvement over the previously proposed LBP approach. With the advances in remote sensing technology high-resolution remote-sensing (HRRS) imagery has been drawing more attention from a wider audience, beyond the traditional scientific user community. The state-of-the-art remote sensing image retrieval supports image search based on metadata (i.e. sensor type, sensor name, image type, image date, and location) or on perceptual features (e.g. color, texture, and shape). However, for most non-scientific users, articulating a query using these metadata or low-level features can be non-intuitive and difficult. We believe that adding semantic annotations to HRRS images would greatly benefit in effectively formulating queries to retrieve the data that users are interested. The first step to performing automatic annotation is to segment HRRS images properly and accurately, which is quite a challenging task. To segment an image accurately means to avoid any false segmentation, whereas to segment properly implies that the segmentation algorithm should terminate at a point to eliminate over-segmenting or under-segmenting. While under-segmenting may cause loss of information, oversegmenting may bring out too many ignorable details. We believe that to accurately and properly segment a HRRS image, no single segmentation technique is suitable. This is because, in a HRRS image, different objects appear in different land use categories. For example, in a residential area, houses and roads are the main elements. To partition objects in this land use category, lines, rectangles, and directions would be helpful for further analysis. On the other hand, in a vegetation covered area, it is hard to extract any lines. Therefore, it would be more efficient to apply a multi-step segmentation procedure, where a HRRS image is segmented at the land use level first, and then different segmentation algorithms are applied based on the the category of the land use. In essence, in this paper, we argue that, to be able to segmenting HRRS images correctly and properly at land use level is an important preprocessing step to identify objects. Even though our ultimate purpose is to recognize objects in a more detailed level, in this paper, we limit our focus to the problem of efficiently and effectively segmenting remote sensing images at land-use

2011

Vision is the most important resource of information for human beings which contain several activities. Amongst these activities, object recognition and classification are widely used. Although, images are representation of vision which can be interpreted by machines, some images often do not exhibit regions of uniform intensity, but these images may contain variation of intensities which form certain repeated patterns. These repeated patterns are known as visual texture. These patterns can be the result of physical surface properties such as roughness, reflectance and color difference of surface. This special variation in pixel intensities is useful in a variety of applications. One such application is analysis of satellite images, whose purpose is to identify various regions in a given image. In this paper we try to exploit knowledge of texture i.e. repeated patterns of non uniform intensity, to recognize regions in the satellite images. There are many approaches used for texture ...

This paper introduces a new open source, knowledge-based framework for automatic interpretation of remote sensing images, called InterIMAGE. This framework owns a flexible modular architecture, in which image processing operators can be associated to both root and leaf nodes of the semantic network, which constitutes a differential strategy in comparison to other object-based image analysis platforms currently available. The architecture, main features as well as an overview on the interpretation strategy implemented in InterIMAGE is presented. The paper also reports an experiment on the classification of landforms. Different geomorphometric and textural attributes obtained from ASTER/Terra images were combined with fuzzy logic and drove the interpretation semantic network. Object-based statistical agreement indices, estimated from a comparison between the classified scene and a reference map, were used to assess the classification accuracy. The InterIMAGE interpretation strategy yielded a classification result with strong agreement and proved to be effective for the extraction of landforms.

Isprs Journal of Photogrammetry and Remote Sensing, 2017

During the last 20 years, airborne laser scanning (ALS), often combined with passive multispectral information from aerial images, has shown its high feasibility for automated mapping processes. The main benefits have been achieved in the mapping of elevated objects such as buildings and trees. Recently, the first multispectral airborne laser scanners have been launched, and active multispectral information is for the first time available for 3D ALS point clouds from a single sensor. This article discusses the potential of this new technology in map updating, especially in automated object-based land cover classification and change detection in a suburban area. For our study, Optech Titan multispectral ALS data over a suburban area in Finland were acquired. Results from an object-based random forests analysis suggest that the multispectral ALS data are very useful for land cover classification, considering both elevated classes and ground-level classes. The overall accuracy of the land cover classification results with six classes was 96% compared with validation points. The classes under study included building, tree, asphalt, gravel, rocky area and low vegetation. Compared to classification of single-channel data, the main improvements were achieved for ground-level classes. According to feature importance analyses, multispectral intensity features based on several channels were more useful than those based on one channel. Automatic change detection for buildings and roads was also demonstrated by utilising the new multispectral ALS data in combination with old map vectors. In change detection of buildings, an old digital surface model (DSM) based on single-channel ALS data was also used. Overall, our analyses suggest that the new data have high potential for further increasing the automation level in mapping. Unlike passive aerial imaging commonly used in mapping, the multispectral ALS technology is independent of external illumination conditions, and there are no shadows on intensity images produced from the data. These are significant advantages in developing automated classification and change detection procedures.

Texture features play a vital role in land cover classification of remotely sensed images. Local binary pattern (LBP) is a texture model that has been widely used in many applications. Many variants of LBP have also been proposed. Most of these texture models use only two or three discrete output levels for pattern characterization. In the case of remotely sensed images, texture models should be capable of capturing and discriminating even minute pattern differences. So a multivariate texture model is proposed with four discrete output levels for effective classification of land covers. Remotely sensed images have fuzzy land covers and boundaries. Support vector machine is highly suitable for classification of remotely sensed images due to its inherent fuzziness. It can be used for accurate classification of pixels falling on the fuzzy boundary of separation of classes. In this work, texture features are extracted using the proposed multivariate descriptor, MDLTP/MVAR that uses multivariate discrete local texture pattern (MDLTP) supplemented with multivariate variance (MVAR). The classification accuracy of the classified image obtained is found to be 93.46 %.

2019

For the ephemeral river channels in semi-arid regions of India, after every Monsoon season, prompt preparation of river sand distribution maps is often necessary for river sand auditing before resuming the sand mining operations. The process can be readily assisted by classifying the satellite-based remotely sensed imagery, although often confronted by limited accuracy levels arising due to poor distinguishing capability among the spectrally similar classcategories. This study aims to improve the classification accuracy targeting river sand deposits by systematically examining the effects of including spectral indices and textural features in the feature-space while classification. Two most common classification algorithms, viz. Maximum Likelihood Classification (MLC) and Support Vector Machine (SVM) classification were used. The results show that SVM performed even better when Normalized Difference Vegetation Index (NDVI) and correlation texture feature computed at 3x3 window size ...

Remote Sensing and Digital Image Processing, 2004

Synergy between spatial variability (e.g., texture) and spectral brightness (e.g., tone) has great potential to overcome the problem of misclassifi cation of land cover in fi ne spatial resolution 113

IEEE Transactions on Geoscience and Remote Sensing, 2016

Obtaining an up-to-date high-resolution description of land cover is a challenging task due to the high cost and laborintensive process of human annotation through field studies. This work introduces a radically novel approach for achieving this goal by exploiting the proliferation of remote sensing satellite imagery, allowing for the up-to-date generation of global-scale land cover maps. We propose the application of multilabel classification, a powerful framework in machine learning, for inferring the complex relationships between the acquired satellite images and the spectral profiles of different types of surface materials. Introducing a drastically different approach compared to unsupervised spectral unmixing, we employ contemporary ground-collected data from the European Environment Agency to generate the label set and multispectral images from the MODIS sensor to generate the spectral features, under a supervised classification framework. To validate the merits of our approach, we present results using several state-of-the-art multilabel learning classifiers and evaluate their predictive performance with respect to the number of annotated training examples, as well as their capability to exploit examples from neighboring regions or different time instances. We also demonstrate the application of our method on hyperspectral data from the Hyperion sensor for the urban land cover estimation of New York City. Experimental results suggest that the proposed framework can achieve excellent prediction accuracy, even from a limited number of diverse training examples, surpassing state-of-the-art spectral unmixing methods.