Genie Pro can be used for remote sensing, in this photo it is used to detect aircraft on a runway.
Organizing Data—€”and Separating Wheat from Chaff
December 2007 / January 2008
Volume 5 Number 6
Information—€”it's everywhere. We depend upon it. In some arenas we cannot function without it. Our ability to solve complex problems, to answer complex questions and to make informed and sound decisions depends on our ability to effectively gather, store, manipulate and analyze information, or data. We are collecting information in greater amounts and at ever increasing rates—€”in some cases, faster than anyone or anything can analyze it. Some is stored for possible future use. Some may never be analyzed. Yet the necessity to collect and store data continues to grow.
One of the challenges with data today is how to distinguish the relevant pieces from the irrelevant. The sophisticated methods for collecting and analyzing data today are high-tech and high-dollar, meaning they cost a lot to make and run. Two Los Alamos National Laboratory teams have developed methods, through software or hardware innovations, to harness the data monster—€”and, as an added bonus, they've made them user friendly.
LANL's technology transfer division is working with the ISIS Team (Intelligent Searching of Images and Signals) and the PixelVizion Team on three ideas for solving the data organization and storage problem, each based on very different approaches.
"RaveGrid, Genie Pro and PixelVizion are three extraordinary examples of advanced imagery related technologies originating from Los Alamos that offer strong commercial promise across a broad spectrum of applications," said Jerome Garcia at the technology transfer division, "from enabling superior picture-based search engines to assisting with accurately diagnosing health issues to providing real-time visualization of massive 3-D data sets."
PixelVizion
PixelVizion is the result of many years of work with advanced computer hardware, both traditional and non-traditional, and investigation of optimal methods to map or fit algorithms into hardware to achieve improved results. The PixelVizion team, Carolyn Connor, Laura Monroe and brothers Andrew and David DuBois, has been working with computer hardware technologies for decades and specialize in hardware acceleration. PixelVizion is a unique image compositing algorithm that uses hardware to process streaming, complex and massive data in real-time and from multiple sources. The software then composites those data streams into a single scene. This compositing algorithm enables computer visualization acceleration and on-the-fly data manipulation.
For example, a scientist might do an entire experiment, put data into a visualization software and then try to manipulate that data virtually. With current software, manipulating the data can be tricky. Say a scientist's data are visualized as a 3D sphere and the scientist wants to look at another side of that sphere. Using the mouse, he or she can click on the sphere and drag it to the right (or left) just a bit. With current visualization software, it's likely that getting a smooth movement and actually moving the sphere to the exact position the scientist would like to look at could be very tough because the software and hardware are not fast enough. With PixelVizion, because of the accelerated visualization capabilities, the sphere would move as if it were a real sphere that the scientist actually turns with his hand.
The underlying expertise in PixelVizion—€”the team's ability to map complex algorithms to specialized hardware to achieve improved capability—€”can be applied to many algorithms and applications to accelerate or improve data analysis. In terms of streaming data, this acceleration can be very important because it allows data to be analyzed quickly and in real-time, allowing the significant data to be found and minimizing time to solution. In many cases, real-time analysis can profoundly reduce the volume of data that must be stored.
Because PixelVizion uses a unique combination of hardware and software acceleration, it can be used for numerous applications, including virtual medical training, weather patterns, animation and special effects, video game graphics and more.
The team is building partnerships with companies interested in working with the laboratory to develop PixelVizion and expand its applications. Some of these applications include the entertainment and simulation industries.
RaveGrid
RaveGrid, developed by Lakshman Prasad and Sriram Swaminarayan, is a unique raster-to-vector image analysis software solution. Unlike other image vectorization software, RaveGrid models principles of visual perception, using geometric algorithms, to extract features in a raster pixel image using only the edges in the image. These features are represented as polygons of various colors, shapes and sizes that make up a vector representation of the raster image. The polygons in the vector representation actually correspond to visual structures in the image and may be analyzed further for resemblance in shape and color to objects of interest. RaveGrid analyzes images quickly, creates smaller files and maintains the ability to zoom in and out of an image without the blocky artifacts characteristic of raster images.
RaveGrid's potential for automating image analysis tasks was demonstrated in a pilot project for detecting and counting sea scallops in ocean floor imagery captured by the Woods Hole Oceanographic Institution in Massachusetts. This study has led to a three-year project with Woods Hole funded by the National Oceanic and Atmospheric Administration for developing a mobile ocean observatory to map and study the ocean marine habitat to determine the effects of fishing on the habitat and to assess other environmental concerns. In the GIS industry, RaveGrid can be used a lot like it has been used for marine image analysis, to find features of interest and map them. It can also be used to convert large GIS imagery into scalable and economical vector representation for display on various platforms.
Mobile devices like standard cellular and BlackBerry telephones have limited processing and storage resources and small screens. Hence, these devices typically have relatively slow central processing unit speeds and limitations on the number of pixels that can be displayed and the amount of data that can be downloaded to them. RaveGrid solves both of these problems by providing vectorized images that can be displayed free of scale or number of available pixels and in small file formats that can be easily transmitted to mobile devices.
Image searching is also an exciting area for using RaveGrid. The developers suggest a tag- or label-free method of image searching. Currently, images are tagged by humans with some name, for example, a digital photo of your dog might be labeled "Sparky." When using a search engine like Google, you use text to search for data included with images. Prasad and Swaminarayan propose to develop methods to search images without the need for preassigned textual information. RaveGrid's unique image analysis capabilities will be used to characterize a query image in terms of polygons representing the image features. This vectorized image is then compared to all other images, similarly vectorized, in a particular image database using similarity of polygons with respect to shape, color, size, context, etc., to find matches to the query image. Thus, one is likely to find more meaningful search results when searching for images.
Genie Pro
There is a small community of analysts who possess the unique skills required to analyze satellite imagery and extract the required data of interest. Remote sensing (satellite and aircraft surveillance) makes use of the full light spectrum (infrared to ultraviolet) to find important information. The Genie team at Los Alamos has designed software to make the job much easier and to broaden the capabilities for any researcher or scientist who might need to analyze data.
Genie Pro isn't just software—€”it's software that learns. The original version of Genie Pro, known as Genie, was developed to analyze multispectral satellite data. Genie Pro uses evolutionary algorithms to analyze spectral and spatial data of images. These algorithms allow users to identify a small set of significant data, which it uses to teach itself. The algorithm mimics evolution by testing the algorithms for identifying areas of significance on the entire image, based on the significant data selected by the user. For example, the Forest Service is currently testing Genie Pro to identify different types of land in the United States using satellite images. One of the challenges to making maps from these images is that some of the land areas look the same except for minor characteristics; for example, grassland tends to look a lot like fields of planted agriculture because both are green on the satellite image. However, the planted agriculture always has a different texture, in other words, it's planted in rows which you can also see on the satellite image. By selecting a few of these areas of interest through an easy-to-use graphical user interface, e.g., green areas with a particular texture, Genie Pro will evolve an algorithm to best identify these areas on the rest of the image. Genie Pro can establish differences in all the areas in the image based on the color or array of colors (spectral) and the textures (spatial).
Using satellite data alone, Genie Pro can be used to analyze things such as damage caused by wildfires, snowstorms, tornados or hurricanes; floods, tsunamis, earthquakes, volcanoes or terrorist attacks; and to monitor environmental changes or crop health.
In 2000, the Cerro Grande Fire threatened the laboratory and destroyed numerous homes in and around Los Alamos. Genie Pro was used to analyze the effects of the fire. Seven years late, Genie Pro has many more users within and beyond the lab. The Genie team is working with the Environmental Protection Agency and Yale University on subjects ranging from analyzing pollution data to medical applications. In addition, Genie Pro was recently licensed to two companies, each with exclusive field of use in a particular area.
Krystal Zaragoza is a communications specialist at Los Alamos National Laboratory.
Copyright © 2012 | Innovation America