On the eve of another GXP user conference, with its focus on major new releases, SOCET GXP® v3.1 and SOCET SET® v5.5, it’s timely to pause a moment and ponder about our directions for the future. As SOCET GXP gains increasing traction and is well received in the image analysis world, commercial customers are eager for v4.0 in 2010, with the remaining capabilities that they need to achieve what they can do with SOCET SET, yet with the updated style and ease of use of the younger product.
SOCET GXP v3.1 has many new features, but one that’s worth a moment’s thought is our first group of capabilities for analysis of hyperspectral and multispectral imagery (HSI and MSI). All of us are familiar with imagery from satellites such as GeoEye®-1, with one panchromatic and four multispectral bands, and the availability of pan-sharpening and other MSI capabilities is welcome. We await with anticipation the launch of WorldView-2, with its high-resolution panchromatic and eight multispectral bands. And the appearance of studies such as land-cover classifications bears out the manufacturers’ claims that high-performance commercial airborne digital sensors not only replace aerial film cameras but also offer accurate radiometry and thus open up a whole new range of applications. Hyperspectral sensors, with tens or hundreds of bands, often stretching beyond the visible as far as the thermal infrared, are more challenging to understand and demand powerful software to render the huge image cubes into useful information.
Our goal for SOCET GXP is not to offer specialist HSI and MSI software but to provide basic capabilities that are fully integrated with all of the other functionality and equally easy to use. The initial offering in SOCET GXP v3.1 includes destriping, principal components analysis, supervised and unsupervised classifications, change detection, anomaly detection and spectral unmixing. A little beyond v3.1 will come atmospheric correction and more algorithms for pan-sharpening, including the innovative Ehlers fusion. But we realize that customers’ needs in HSI and MSI processing vary, so, starting with a workshop during the GXP User Conference and proceeding via a series of customer visits by product managers, we shall work with customers to see what else to add so that the product is of maximum value. This is no easy task. Every HSI and MSI analyst has favorite algorithms and there are thousands to choose from. I’ve just read a review paper, “Change detection techniques,” by Dengsheng Lu and various co-authors and on just change detection it assesses 31 algorithms and cites no fewer than 273 references! So we have to be selective, however we want to cater to customers’ requirements as best we can in this important new area of functionality.
More challenging than HSI and MSI is radar. SOCET SET and SOCET GXP include sensor models and ingest capabilities for numerous radar sensors. Customers can share in the excitement being generated by the superb imagery and metadata streaming from TerraSAR-X and COSMO-SkyMed. But we treat these as images rather than take full advantage of this particularly informative data type. We have special feature matching in SOCET SET to facilitate point matching for triangulation between electro-optical and radar imagery, but it’s still all imagery. Now we are looking at going further and digging into the complex radar data to extract more information. For example, radar images collected from similar platform positions at different points in time can be used in a process called coherent change detection, which can yield information about height changes that are much smaller than a pixel. Applications of improvised explosive devices detection and analysis of subsidence caused by mining spring to mind. We are putting together a roadmap of radar capabilities that we can add in the years to come.
For acronym lovers, image analysts and photogrammetrists alike, radar leads naturally to LiDAR. It’s another active sensor, returning XYZ coordinates and intensity values for many millions of points per hour of flying. As laser pulse rates increase with technological progress and Multiple Pulse in the Air electronics break down another barrier, LiDAR point clouds can now be as dense as several points per square meter. How do we process all this data? SOCET SET can ingest LiDAR, for example in the LAS 1.2 standard format, display points, create intensity orthophotos and stereomates, and allow the user to apply the powerful ITE tools to edit the point cloud. BAE Systems is not a provider of specialist LiDAR software, but it’s remarkable how many photogrammetric capabilities are easily applicable to LiDAR data. Remember too that more than 50% of LiDAR systems are sold with imaging sensors, typically medium-format digital frame cameras: the strength of photogrammetric software is the easy fusion and manipulation of the LiDAR and image datasets. And we can do more by building on our strengths. What about a LiDAR sensor model, or how could we use NGATE’s simultaneous generation of DSM and DEM elevation data to process LiDAR point clouds? We’re looking hard at these issues.
With HSI and MSI, radar, and LiDAR, we have exciting work ahead to meet customers’ requirements for processing data from these popular sensors. That will keep us well occupied, but where could we go next? Someone is always excited by a novel computer platform. What about SOCET GXP on an iPhone? Not very practical, perhaps, but an operational system on a ruggedized handheld with a built-in GPS receiver is intriguing. Or could we offer image analysts, perhaps, better links between the imagery and one of the modern geospatial text searching services, so the analyst can find out more about the places on the image and their culture? How can we distill your aspirations and our ideas into a practical direction for SOCET GXP? Tell us what you would like to have – or all this will just be RUMINT!
Dr. A. Stewart Walker
Director of Product Initiatives
BAE Systems GXP