Indian Air Force
E-mail: [email protected]
With the launch of each remote sensing (RS) satellite, the data is increasing exponentially both in content and usage. This paper also looks at some recent work on Geo-Spatial Data Production System (GSDPS), remote sensing, on-board information management system, photogrammetry and database system capable of managing a huge amount of RS images
Today, multiple terrabytes of Remote Sensing data in the form of images and metadata are being collected by majority of nations across the globe, from diverse and complex space borne platform. It raises question how to process, manage, archive and make best use of the RS information, and share such a vast amount of RS images and value added data, which can benefit to the public at large, developmental projects besides stimulating the concepts and research developments. The need of the hour is mainly to provide an effective and efficient “connection” between the image providers, archivers and the image users, besides ‘value adding ‘ (knowledge-base) to data downloaded and link them directly with the “demand and supply cycle” for optimum utilisation.
During the past decade, the satellites have acquired large volume of data using sensors operating in different bands of spectrum, namely, Optical, microwave (synthetic aperture radar (SAR)) and other sensors, that have been systematically collected, processed and stored. The best available state of art systems provides queries mainly in the geospatial coordinates, time of acquisition and the scene types. This information is generally less relevant when compared to the context of the scene, such as pattern, tone, objects, structure, geometric, dielectric properties and scattering parameters etc.. Thus the comprehensive query of an image database is not carried out and lot of valuable contextual data is lost. It is understood after research and experience over the years that the data handling of imageries is different to any other file data handling due to the context-content knowledge-based associated with images.
Today, there is a need to emphasise on the development of on-board geo-data base management system for the future EOS enabling end user to directly downlink satellite imagery for their specific area of interest in their work (Campbell et. al 2000, Prescoll et. al 1999, Zhou 2001, Zetocha 2000). One of the key challenges in this area is how to undertake on-board data distribution with autonomous distribution of data, and automatically retrieve other knowledge based information from the data sets of database and how these data sets will be updated and maintained automatically. The answer to these key issues lie in the philosophy of artificial intelligence, data mining algorithms and the concept of neural network along with the advantages of telecommunications. KEY TECHNOLOGIES IN BUILDING RS IMAGE INFORMATION DATABASE
The RS database is different from traditional image database of telecommunications, human resource, energy, police, agriculture, forestry, land development and to some extent geo-information. Though these databases contribute towards the knowledge base data extraction, it supports and strengthens the achievements of many social and environmental objectives, to be achieved by the RS image database also. The concept of spatial data infrastructure (SDI) is an integration of components technologies, policies, standardization and human work-process (resources) necessary to acquire, process, store, distribute and improve the utilization of geo spatial data to the widest possible group of users. With rapid advancements in the field of RS technology, a huge amount of RS database and knowledge-base can be made available seamlessly and near-instantaneously to varied kinds of users simultaneously. The concepts are changing and gaining momentum towards Ortho-RS image spatial framework. Hence, the development of RS images information management system must be able to handle vast volumes of data. The key technology for achieving this lies with:
Fig 1 Integration of image data fusion in 3D environment for knowledge information base
Virtual seamless mosaic
The virtual mosaic is the computer-defined mosaic of the images to make ‘seamless mosaic’. It maintains the integrity of RS image database management system, that includes spatial, spectral and radiometric though it is independent of geometric corrections. The data used in the virtual mosaic of the RS data needs to be geometrically corrected to compensate for the distortions by earth curvature, panoramic distortions, relief displacement and atmospheric refraction so as to be re-projected to a common datum as spheroid and projections and finally drape them with the DTM. The database thus made after these functions is mainly continuous and have elevation information data though the data content is largely devoid of value added information. However, the data structure holds good in feature based co-relation mechanics and are widely used for the knowledge content based data mining and management.
Fig 2 Images hierarchical structure in database
Fig 3 Geo-spatial query model
Fig 4 On-board data management model
Data fusion of Multi source images
The RS images of the earth atmosphere, hydrosphere and geosphere are continuously created from various platforms and sensors. Hence, the fusion of multi-platform, multi-resolution, multi-temporality, multi-sensor, multi-angle, and multi spectral image information is an obvious strategy. In order to enrich our future classification in area of interest (AOI), and analytical image based knowledge base, such solution strategy will increase the reliability and accuracy of RS images. There are three methodologies of RS data fusion and it’s handling (a) Pixel based, (b) Feature based and (c) Determination based.
Multi-resolution is a very important and central aspect of RS image management systems, because different users need different levels of RS image details (Hierarchical Organisation). A pyramidal structure is the most common approach for multi resolution or hierarchical management system. In this each of the derived layers stack on the top of previous layer. Various concepts of stacking have developed over the time for the coding of entire earth surface. Google Earth is a classical example of hierarchical management of images. Though the system is able to satisfy different requirements of resolution and minimizes the search time by balancing the load throughout the processor and network, the cost of the storage system is exorbitant. The benefits of each system have to be incorporated to build a comprehensive spatial data management systems.
Spatial indexing and querying
After the multi resolution management of RS imagery, cell indexing and searching is the most important issue for building RS image data base systems, because it directly impacts on the performance of the system. The work process and the human resource management is directly linked to the capabilities of indexing and querying of the system. Spatial index is another important index approach for spatial data structure. One of the most simple spatial query systems is explained. The system has been found to be very effective for moderately large scene based image database. Various knowledge base programme such as algorithms, which can find the probabilities of most visited images and scenes for workspace management, designer query geo-algorithms, metadata text search engine etc, can be incorporated along with the query models. Besides these there are ocean of technology in the data mining and geo-algorithms, which can undertake refined indexing and querying. FUTURE OF INTEGRATED INFORMATION MANAGEMENT OF RS IMAGES
The new design concept for on-board geo-data management in future intelligent earth observing satellite is as under.
On-board Geo-database management in future RS satellites
The integration of satellite raster image data, with the already existing geo data, is one of the important challenges of image processing, image geo-coding, image data management. The traditional GIS holds the key for integration of satellite imagery, on space-borne and ground controlled and mirrored GIS platform. There are three basic models for space borne integration (a) Separated but Parallel Integration ; which means that the image processing system and GIS system are separate but are used in integrated fashion to transfer knowledge based geo-spatial information between the two data handling systems where GIS compliments the imagery database with knowledge base, (b) Seamless Integration; this means the GIS and image analysis system are stored together and the functions of both the system are interfaced and accessed through a common interface and (c) Total Integration; this means the RS and geo data support each other for analysis and processing and make full use of benefits of image analysis functionality and GIS simultaneously. The future earth observation satellite are likely to have on-board data integration functionality and will be able to manage minimum two data sets namely from geo data and DEM (vector and raster) and will also seamlessly link them together. To achieve this following data type and its modeling will be required.
Data types and data modeling
DTM data: The DTM data will provide the elevation information to the end user and will also traditionally provide data for carrying out ortho image rectification so as to link with ortho – RS database system existing on earth user. The data will be stored in the form of typical regular raster data, triangular irregular network (TIN) and hybrid data. The DTM database will also be provided as attribute associated with the imagery. This data set will also be used for programming the satellite downloads.
Satellite image data: As a practice the satellite image from the on-board sensor are co-registered with the ground co-ordinate system using on-board image processor with specific algorithms. However, in the advent of on board data management, only the change in coordinate data will be transmitted resulting in transmission of the changed area image data. To achieve high performance image filters and modulators with refined contrast matching techniques, correlation function and other contemporary technologies will be utilized.
Spatial data: The geo spatial data is an abstract entity in the real world. The spatial data has two obvious features, viz., geometric and physical characteristics identifiable by point object, line object, area object and complex objects. This database will hold the feature characteristics of the objects updated time to time from the earth station or with the privileged clients. This database will have all characteristics of typical GIS.
Client knowledge base engine: This database engine provides the client knowledge base interface and keep a track of all the data queried and in process so that the speedy scheduling of the data is achieved, by switching off and on the filter circuits and modulators so as to achieve maximum of the processing need. This engine will also have standard work process scheduler for most of the digital image processing.
Implementation of on-board data management systems
Work flow modelling: The work force modeling will be the key process controller as it functions as interface to execute query modeling in the technical functions to the human resource management in the ground based system.
DTM data management system: The purpose of a global DEM database on-board is to directly provide elevation information to user without any processing on ground. The DEM management will be carried out in block indexing fashion with an unique DEM identifier. The DEM processor will automatically generate DEM pyramid data for data management and will under take ortho-corrections.
Geo- spatial management system: This database system will hold the attribute data connected to spatial database connected by a unique identification number (geodetic co-ordinates). This will be achieved by the help of organizing attribute and spatial data in the same record of database or separating them and specifying the link. This scheme will directly employ the spatial data to a unique identification number (geodetic co-ordinates) so as to connect the attribute and spatial data.
Fig 5 Work flow of integrated management system Processing and client knowledge data mining management system: Three types of data as above are stored in three separate databases. An integrated management system is required to manage and process three sub systems and query database. A unique dynamic identification number or a pointer will connect all the databases and generate the dynamic query. A integrated management system will also be responsible for imposing scale compression process, co-ordinate transformation and transmission to on-board data coordinator and distributor.
Key technologies for the on-board integrated information management: The management system uses the image as a reference layer in which the geo-data and the DEM data is imposed for analyzing and querying in the real world environment. This type of product contains more information than the traditional real world information database generated from satellite. User can better orient themselves with geo-image maps spatially when they are not aware of image processing work process needed for undertaking the job. The realization of this function will change the imagery storage concept and requirements for the similar images will reduce considerably. There are some key thrust areas that need to be developed to make this technology successful.
- Retrieving the raster images, DEM and attributes as well as returning them to user bandwidth technology is the key technology faced for more advanced querying and that too particularly when it is on the fly.
- As the complexity increases more data especially high-resolution satellite data will be demanded for task. It will be a time consuming to query them at and simultaneously to take decision and determine which data holds priority. Hence the key technology development area will be in parallel fast processing on-board satellite.
- As mentioned the images will be orthorectified and will use geographic reference layer. The image will be in all three phases, i.e., differentially rectify, partly rectify and un rectify because of on-board processing speed. When the attribute data will be imposed on them the error are likely to be generated. The key technology area will be to on fly calculate the errors before transmitting the data to the client.
- The universal problem of inappropriate linkage in DEM images and geo-data will produce confusion and inaccurate result in downloaded imagery. The key technology area is to develop an artificial intelligence to rectify the same.
- The inconsistent work processes for the on-board processing will lead to inconsistent results. The key technology area will be to neural networks the artificial intelligence to provide solution from the heterogeneous queries.
The combination of RS and modern database technology is making possible the management of huge amount of RS image data. In addition, the RS image management not only significantly increases the system efficiencies, scalability, security and integrity it also extends the application of space of the RS image. Moreover, a RS data base system in future will provide a platform for on-board integration of satellite data with corresponding geo-data (spatial and attributable) and DEM data in the satellite, reducing the work planning, processing and acquisition delays.
Acknowledgement & References
The works carried out by Mihai Datcu , Herbert Daschiel et al. Yufei Wang , Chris Rizos , Linlin Ge & Craig Roberts from school of surveying Sydney. Paul Kaufrmann and Guoqing Zhou from old Dominium University. Works on knowledge based interpretation by Stefan Groves et al, various proceedings of IEEE , Data from various conferences, Journals and references from open source have contributed towards the development of this article. I thank Brigadier RC Padhi for his guidance and support provided for this work.