S. Manthira Moorthi

An automatic satellite image registration technique based on Harris corner detection and Random Sample Consensus (RANSAC) outlier rejection model

Automatic satellite image registration is a challenging task of overlaying two images for geometric conformity aligning common features by establishing a transformation model using distinguishable feature points collected simultaneously in both the images in a completely un assisted manner. Remote sensed images capture terrain features in a natural condition subjected to seasonal changes, sun illumination conditions, and cloud presence. The critical steps in image registration are collection of feature points and estimating a spatial transformation especially when outliers are present besides feature matching and resampling the slave image to the master image geometry. In this paper, the details and merit of employing automatic Harris corner detection and building a transformation model using Random Sample Consensus (RANSAC) algorithm is brought out while registering a pair of LISS-3 or AWIFS images from Indian Remote Sensing Satellite (IRS) platform. Potential available with this approach for performing large scale image registration tasks such as time series processing are highlighted.

Adaptive stochastic gradient descent optimization in multi temporal satellite image registration

Image registration of satellite imagery is challenging task as the image pair may have been captured by different sensors, view angles or at different times. Conventional satellite image registration tasks are achieved by feature based techniques which are to be identified and matched before estimating transforms and resample the moving or slave image to a fixed or master image. However, automatic image registration is very important requirement for voluminous data sets. Even if feature selection made automatic, even distribution and the density of feature points is always not ensured. A good alternative is to switch over to intensity based image registration techniques which are somewhat image content independent. We bring out details on an intensity based multi temporal satellite image registration task with metric, transform and optimization components based method. An optimization technique employed to find optimal transform parameters by maximizing the chosen similarity measure criteria, provides a robust image registration framework. Adaptive Stochastic Gradient Descent optimization (ASGD) for satellite image registration is a promising new technique, details of which are reported here.

An efficient algorithm for automatic fusion of RISAT-1 SAR data and Resourcesat-2 optical images

Satellite Image fusion generates single hybrid image from a collection of input satellite images and helps us to extract maximum information from the remotely sensed datasets to achieve optimal spatial and spectral resolution. The critical steps of image fusion framework are co-registration of Synthetic Aperture Radar(SAR) data with corresponding optical scene, enhance the images for visual clarity and then merge the multi sensor data with a standard fusion technique. The image fusion system should perform all these steps in an automatic manner for providing ease to the user. The primary attention of this work is to examine the improvement that can be obtained by fusion of low resolution multi spectral data obtained from optical Resourcesat-2 platform (LISS-4MX/LISS-III/AWIFS Sensor having 5m/24m/56m spatial resolution) with high resolution RISAT-1 (Fine Resolution STRIPMAP (FRS-1)/Medium Resolution SCANSAR(MRS) mode data having 3m/18m spatial resolution) using SAR-Optical image fusion system discussed above. This integration of optical and SAR images from Indian Remote Sensing satellites facilitates better visual and automatic image interpretation. The Maximum Likelihood algorithm is used for classification of fused image and Resourcesat-2 multispectral data. The quality improvement of the fused product can be observed by comparing the classification accuracies of merged data with original multispectral data of the same region.

Kernel based learning approach for satellite image classification using support vector machine

Machine learning is a scientific computing discipline to automatically learn to recognize complex patterns and make intelligent decisions based on the set of observed examples (training data). Support Vector Machine (SVM) is a supervised machine learning method used for classification. An SVM kernel based algorithm builds a model for transforming a low dimension feature space into high dimension feature space to find the maximum margin between the classes. In the field of geospatial data processing, there is a high degree of interest to find an optimal image classifier technique. Many image classification methods such as maximum likelihood, K-Nearest are being used for determining crop patterns, land use and mining other useful geospatial information. But SVM is now considered to be one of the powerful kernel based classifier that can be adopted for resolving classification problems. The objective of the study is to use SVM technique for classifying multi spectral satellite image dataset and compare the overall accuracy with the conventional image classification method. LISS-3 and AWIFS sensors data from Resourcesat-1, Indian Remote Sensing (IRS) platform were used for this analysis. In this study, some of the open source tools were used to find out whether SVM can be a potential classification technique for high performance satellite image classification.

An automated seamless mosaicing system of multi-charge coupled devices of panchromatic data

ABSTRACTPanchromatic data of pixel resolution 5.8 m obtained from IRS-1C and IRS-1D satellites proved to be very useful for mapping purposes. One of the popular data product is the 70 km swath mosaic which is covered by a combination of 3 CCD line sensors, each with 4096 pixels. Each CCD-line sensor with different imaging times causes geometric problems of mosaicing three strips data together. In this paper, we propose the details of the design elements of system that caters to the need for accurate and automatic multi strip image registration without any second resampling of the data. The systematic geometric correction grid mapping is improved to facilitate accurate mosaicing by automatic image registration task that makes use of the overlap data within image strips and image registration is achieved up to sub-pixel level.

Correction of Mars Colour Camera images for identification of spectral classes

Mars Colour Camera on-board the Mars Orbital Mission makes use of a Bayer pattern detector. Spectral response of RGB (red, green and blue) pixels of Bayer detector shows large overlap which reduces the spectral information content of the image. In the present paper, a simple method is suggested to correct the data for spectral overlap. It is shown that correction process significantly increases the spectral information content of the image and enhances the ability of the sensor to identify different target types like dust clouds and water ice clouds.

High dynamic range processing of Mars Color Camera images

High dynamic range processing is one of the desired step for images taken with varying exposures for the same area to capture details in dark shadows as well as bright light. Indian Mars Orbiter Mission carrying Mars Color Camera (MCC) has the capability of taking images with varying exposures. Images taken by this camera during Earth Bound Phase of the flight is processed for High Dynamic Range (HDR). Satellite images are good as well as interesting candidate for high dynamic range processing. Low dynamic range (LDR) images needs to be registered a priori, which itself is a challenging task as these images may have cloud pixels, noise, low contrast etc. Significant improvements observed after HDR processing is statistically and visually depicted in this paper as compared to single LDR frames taken by the camera.

Minnaert Topographic Correction of Mars Color Camera Images from Mars Orbiter Mission

Mars Color Camera (MCC) images obtained from Mars Orbiter Mission (MOM) are gaining scientific popularity since MOM insertion into an elliptical orbit around Mars on 24th Sep, 2014. Planetary remotely sensed images are corrected for topographic effects to normalize the radiance measures before considering the data for science analysis. It is proposed here to use non Lambertian Minnaert semi empirical approach for correcting MCC images that are used for deriving results for Mars surface science. The methodology outlined here uses terrain parameters such as slope and aspect derived from Mars Orbiter Laser Altimeter (MOLA) digital elevation model (DEM). Topographically corrected images were evaluated for the improvement in its radiometry quantitatively and it is found to reduce topographic shading and improve the image quality.