Earth Imaging Journal: Remote Sensing, Satellite Images, Satellite Imagery
Breaking News
Germans’ Purchasing Power Climbs 2.8 Percent in 2018
Bruchsal, Germany, December 12, 2017 – Germans’ 2018 purchasing...
Tersus Introduces High-end GNSS Kits with Centimeter-accurate GNSS OEM RTK Boards
Tersus GNSS Inc. (“Tersus”), a leading GNSS positioning solution...
SSL to Provide Access to Space for Small Satellite Constellation
PALO ALTO, Calif. - SSL, a business unit of...
Global Aerial Imaging Market – Expected to Reach $3.2 Billion by 2023 – Research and Markets
DUBLIN -The "Global Aerial Imaging Market Analysis (2017-2023)" report...
FLIR Announces FLIR DM166 Thermal Imaging TRMS Multimeter with IGM
WILSONVILLE, Ore. – FLIR announces the FLIR DM166 thermal...

June 26, 2013
Demystifying CLOUD COMPUTING for Remote Sensing Applications

Cloud computing is a powerful, unique value proposition that can transform massive Earth imagery datasets into useful information for users worldwide.

 By Kumar Navulur, Dan Lester, Amanda Marchetti and Greg Hammann, DigitalGlobe (www.digitalglobe.com), Longmont, Colo.

On-demand solutions for global geospatial challenges require fast access to Earth imagery and the ability to scale information extraction and insight from petabytes of data collected every year. Finer spatial resolutions, quicker revisit times, more spectral bands, higher bit-depths (radiometric resolution) and the ability to take multiangle collections have resulted in vast volumes of data. No longer is the question “How do we capture imagery?” but “How do we handle the immense volume of imagery we already have and to which we’re adding every day?”

The National Institute of Standards and Technology wrote in 2011 that “Cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources—e.g., networks, servers, storage, applications and services—that rapidly can be provisioned and released with minimal management effort or service provider interaction.” Cloud computing for remote sensing represents the notion of co-locating data with processing infrastructure rather than delivering the data to remote processing centers.

The amount of imagery in an archive and being collected each day is simply too much to deliver efficiently to each user. Furthermore, that vast volume of imagery and related data presents a daunting big data problem for conventional computing systems and techniques. A cloud-computing environment enables tremendous scalability in storage and processing power that’s way beyond the capability of all but the largest institutional users.

The cloud can be managed centrally, and new algorithms (analysis formulas) can be added quickly. On-demand processing that combines additional information from other data sources, such as geographic information system (GIS) layers, is becoming a reality for users around the world. Also, cloud computing reduces each user’s hardware and software requirement, as the hardware and software necessary for advanced production is held centrally by a service provider or large institutional user.

Maximizing Imagery Resources

The commercial remote sensing industry has seen a dramatic evolution during the last decade in terms of spatial resolution, sensor design and image access developments; Google Earth, ArcGIS Online and Bing Maps are only a few examples (see “Surveying Uncharted Territory,” page ??).

The industry has sparked increased spatial resolution of images from satellites, airborne sensors and unmanned aircraft systems. Satellite imagery’s spatial resolution has decreased to about 20 inches (0.5 meters) for commercial customers. Moreover, aerial and satellite platforms can collect stereo imagery that can be used to produce DEMs, including 3-D buildings.

Increased spatial resolution is further complemented by increased spectral resolution, especially in the satellite industry. The number of available spectral bands has increased from the traditional four visible and near-infrared (VNIR) bands to specialized bands in the VNIR regions. Landsat’s Thematic Mapper sensor has six optical and one thermal band; the newly launched Landsat 8 has eight narrow optical bands, plus a panchromatic band as well as two thermal bands similar to bands 4 and 5 on the Advanced Very High Resolution Radiometer (AVHRR) sensor carried by the U.S. National Oceanic Administration’s polar-orbiting weather satellites.

Now commercial high-resolution satellites can measure Earth in eight optical bands, and a satellite with 16 bands is under construction. More bands better characterize the physical and chemical nature of observed materials. The latest satellite trends point to more spectral bands in the short-wave infrared
region, which are longer wavelengths than the visible bands. These spectral bands open tremendous opportunities for information extraction for various vertical markets, including agriculture, precision farming, geology, mining, oil and gas, forestry, land-use planning, construction and others.

In addition, the radiometric resolution/information depth collected is increasing, with the new sensors collecting information up to 14 bits; not too long ago 8-bit imagery was all that could be stored and processed. Now satellite constellations have enabled global revisit times daily, which also contributes to the exponential growth in data collected. These factors have pushed the remote sensing industry, as well as the geospatial community in general, into the big data paradigm that’s being studied and developed for many other industries, including finance, advertising, marketing, law enforcement and more.

Today, commercial Earth observation satellites collect more than 4 million square kilometers of imagery daily, totaling petabytes every year. All of these data require processing to extract useful information. Images often need to be corrected for elevation, a process called orthorectification that’s important to produce accurate maps. The lower-resolution color bands (multispectral) can be enhanced, or pan-sharpened, with the panchromatic band.

Sometimes a satellite can’t collect an area of interest at a particular time, so different scenes must be stitched together or mosaicked. It’s important to remember that satellites view Earth from space, above the atmosphere, which is like looking through a foggy window. The data are in units of radiance, or calibrated pixel brightness, called top-of-atmosphere (TOA) radiances.

But most applications are on the land’s surface, so the TOA radiances must be corrected for atmospheric impact. In other words, it’s important to clean the window to get a good view of the surface. This process is called atmospheric correction, and the result is the surface reflectance that’s actually viewed on the ground.

These computer-intensive processes require high-volume storage and powerful computing resources that many end users don’t have the luxury of acquiring and maintaining. Surface reflectance supplies the most efficient and accurate information for extracting land features from an image, but many users must do without because they can’t afford the investment. Cloud computing offers the solution to this and other processing problems.

The spatial and spectral improvements of modern sensors enable tremendous opportunities for information extraction. Various industries, including agriculture, geology, oil and gas, mining, forestry and environmental monitoring, already benefit from traditional remote sensing methodologies in land classification and feature extraction. Improved remote sensing technologies and workflows increasingly enable similar techniques to be used within geospatial intelligence agencies for derived insight.

With petabytes of data collected each year from remote sensing vehicles, the ability to process all the data becomes a major factor that requires serious consideration as the industry dives ever deeper into big data. Furthermore, recent years have seen the adoption of machine learning and machine vision technology with spectral exploitation image analysis. Such computer-intensive techniques employ robust morphological parameters that can create hundreds of additional derived layers used for image exploitation.

Such technologies create intermediate data layers that can be several factors greater in size, compared with the original image file size. Creating a computing environment with racks of cooled servers for high-end processing power simply isn’t a realistic possibility for most end users. The problem is further exacerbated by the need to process large contiguous areas at near real-time speeds as well as to disseminate the imagery and derived products to users worldwide.

Cloud Computing and the Geospatial Industry

The geospatial industry has begun to embrace the concept of a platform for end users. Cloud computing allows for the concept of a unified remote sensing and image processing platform to be built, maintained and used throughout the globe.

Rather than a one-to-many distribution model, cloud computing’s single platform improves efficiency by supplying users with a many-to-one workflow that’s locked into a high-performance computing environment for image and information analysis and dissemination. Among the primary reasons for moving toward a platform mentality is to adopt a software as a service (SaaS) model, which makes software functionality easily accessible by reducing the information technology (IT) investment burdens on end users and ensuring fast access to data for subsequent processing and delivery.

The SaaS model has proved fruitful for location-based service companies that have embarked on developing a platform to ensure timely delivery of data to customers worldwide. Cloud computing offers an economical option for the geospatial industry to leverage public and private on-demand clouds for computing power, security, storage and delivery capability regardless of a user’s physical location.

Pixels and Pictures Aren’t Enough

Pixels, or picture elements, are an image’s minimum unit; a pixel is one data point in space and time. A single pixel or point will not create an image. Similarly, an image won’t tell the entire story.

The legacy remote sensing industry concentrated efforts on better data collection, new and better (lower signal-to-noise) sensors and new analysis methods for information extraction. The Internet bandwidth was too low to easily and economically download the data, and no central processing centers were available for the general public. Specialized hardware and software has been expensive and hard to use without training.

The commercial business was designed to collect data, perform the minimum or required processing (basically calibration) for the data to be usable by others and to deliver calibrated pixels to the end user much later. In those days, end users had to have the knowledge and the infrastructure to derive desired information based on the problem at hand. So the remote sensing industry basically has supplied raw material for scientists and companies with the ability to exploit it.

But geospatial questions and problems are common. The remote sensing industry will enjoy tremendous growth if it can provide answers to relevant questions and solve real problems for a broader group of users instead of just providing raw data (pixels) for highly trained geospatial and remote sensing professionals. Insight is the power to make decisions, but remote sensing data providers must mature to provide users the insight they seek.

One way to look at it is to follow a workflow that goes from a single data point (the pixel) to an image (many pixels) where the information shown can be extracted. The image provides new knowledge and understanding, but it still doesn’t answer the question. With further analysis and typically by including many different data sources, the question can be solved, providing insight and the ability to make an informed decision.

Cloud computing—co-locating raw material with a centralized storage center and processing power—offers the ability to create and disseminate insight. Future user interfaces will have friendly tools to help formulate questions so computers can perform on-demand analyses and provide results in a required format, whether it’s an image, a thematic map, a metric such as an area, item counts or distances, a yes or no answer, or a full report. The products provided are of ever-increasing information content, thereby increasing the value proposition for the user and in turn for the provider and investors.

Disseminating On-Demand Insight to End Users Globally

Current market trends reflect a need for the immediate delivery of imagery, information and insight to users’ desktops while decreasing their costs. Imagery is integrated deeply into geospatial workflows and typically is invoked with Open Geospatial Consortium standards for online delivery. Cloud services offer a new solution with near real-time insight delivery. When employing the best and most current cloud computing practices, the anticipated timeline for cloud calls and services is less than five seconds after a request is made anywhere on the globe.

With the increased complexity of data analysis and the need to extract true insight for broad-area analyses, cloud computing offers a superb solution for the industry as well as the end user. One platform can be empowered with the top technological techniques for feature extraction and predictive analytics to leverage big data computing methodologies that aren’t available to the average image or geospatial analyst. Once a new image arrives to the platform, all derived information layers can be shared by multiple users to produce many derived datasets in a fraction of the time of typical analytical techniques. The power to get the insight needed when it’s needed is the revolutionary power of the many-to-one enterprise cloud-computing model.

Improving Efficiency While Reducing Cost and Risk

Cloud computing offers the remote sensing industry an interesting value proposition as the need arises to store and analyze immense amounts of imagery. Spatial, spectral, temporal and angular advances in the field of remote sensing are reshaping users’ expectations, but the costs of computing environments for big data storage, analysis and information security beg for new financial strategies in the marketplace.

This challenge is a tremendous opportunity. Cloud computing fulfills a necessary role by serving as a near real-time insight platform that can rapidly disseminate big data analysis. It’s an established platform that enables access to high-end computing resources for users worldwide without large budgets for IT investment. Cloud computing offers an efficient, low-cost solution to the ever-increasing technological demand of the geospatial industry.

The on-demand creation of insight, using cutting-edge remote sensing tools and related information layers, without the burden of IT infrastructure, will be a major catalyst in the future of situational awareness and information extraction. The freedom and security provided by public and private computing clouds enable data providers and algorithm developers the ability to host their intellectual property with confidence. Unlocking the full value of imagery for data producers, algorithm developers and end users requires the community to embrace a paradigm that understands the big data landscape and resonates within the evolving remote sensing field.

Captions

Cloud computing offers an economical option for the geospatial industry to leverage public and private on-demand clouds for a wide range of applications.

Cloud computing can provide insight for a broad group of users instead of just raw data for geospatial and remote sensing professionals.

Cloud computing offers myriad image processing benefits, including the ability to create imagery from digital elevation models extracted from satellite stereo imagery (insets).

The trend toward the use of more spectral bands in satellite imagery opens tremendous opportunities for information extraction for various vertical markets.

Comments are closed.