OSSIM began out of frustration. The core developers worked for ImageLinks Inc. - a venture capital spin off from Harris Corporation. ImageLinks specialized in advanced commercial satellite image processing with a Harris Government Off The Shelf (GOTS) toolkit - the Multi-image Exploitation Tool (MET). MET was developed over period of almost 15 years with an average team of 40 scientists, developers and engineers. It was developed for classified government programs with autonomous image registration, mosaicking, and rigorous sensor modeling as just a few of it many capabilities. To this day it remains one of the leading software tools sets for photogrammetry, remote sensing and image processing.
ImageLinks was granted commercial rights to the MET package and subcontracted a team of Harris software developers to modify it for limited commercial use. Harris and ImageLinks agreed to share developed technology for the next two years. In April of 1996, ImageLinks Inc was formed, Mark Lucas became one of the founders and the Chief Technology Officer.
ImageLinks was started on the premise that scheduled launches of the new commercial imaging satellites coupled with advanced MET technology would provide a successful business model. MET required expensive SGI and Sun workstations and software licenses. The cost per seat was in the range of $25,000. Delays and failures in the commercial imaging business strained the financial resources of the company. The small dedicated development team found it necessary to be in a constant state of crisis development to accomplish difficult projects in a small market. Cash flow was very tight and the company found it difficult to expand due to the expensive hardware and licensing costs. In retrospect, this was a blessing in disguise as it forced the development team to work side by side with the production team and encouraged management to consider alternatives to the standard enterprise solutions.
In 1997 several members of the development team were experimenting with Linux. One day at lunch we wondered if it would be possible to compile the MET software on the Linux platform. It was no believed that it would match the performance of the commercial workstations, but the team thought it might be able to accomplish some functions. A detour to the local computer parts store with a company credit card acquired what was needed to assemble a back room Linux PC. This was a skunk works project - the development team would steal time to build up the system and try to compile the software. Jeff Largent built the hardware and got Linux up and running. Ken Melero and Dave Burken took turns working on the software. Linux and the tools were still evolving and it took a couple of months and Linux software releases before the first version was up and running.
A substantial amount of code, libraries, and tools was involved in building the MET baseline. On an SGI workstation it took almost 12 hours of compiling and linking to assemble the technology. The team was amazed to see the Linux workstation complete the build in just under two hours. Much of this performance increase was due to the rapid advance of PC technology. We were comparing two year old workstations to rapidly evolving commodity PCs with more memory and faster disks. The software performed all of the functions and ran faster on hardware that was one tenth the cost. In a very short time span we converted the ImageLinks production to Linux boxes.
The development team quickly became aware of the advantages of using open source technologies. Expensive development tools were replaced with open source equivalents eliminating licensing costs and headaches. When problems were encountered with open source projects the response was immediate - there were several cases where problems were identified, posted on the project mailing list, and then fixed before the next working day began. Cash flow in the company improved, our cost for production dropped dramatically, and the business eventually became profitable.
After a couple of years the business environment and relationship with Harris Corporation had changed. ImageLinks was limited in its pursuit of government development contracts with the MET software product. The technology exchange between the two companies was not working to either companies satisfaction. ImageLinks could not use the Harris MET package in its government pursuits. ImageLinks decided to replace MET functionality with open source software as a strategic objective.
ImageLinks did not have the financial resources to develop a replacement for the MET capabilities. The team reasoned that they could slowly work towards that goal with an open source approach. Hopefully, others would help, progress would be made - eventually they would get there. When we established the website we were hoping that we wouldn’t embarrass ourselves. The team continued to support open source geospatial technologies and subsequently established remotesensing.org as a portal for those technologies.
The website immediately began to attract a community of interest. Frank Warmerdam, Norman Vine and others brought their software and network of contacts into play. The site quickly hosted Frank’s Geospatial Data Access Library (GDAL), the tiff libraries and geotiff extensions, and a number of other related visualization and mapping projects. Oscar Kramer, Garrett Potts, Dave Burken, and Scott Bortman performed much of the initial software development on the OSSIM baseline. Oscar created much of the sensor modeling and photogrammetry framework, Garrett is the overall system architect, Dave has maintained and improved the overall baseline, and Scottie has focused on Java / Web Mapping services.
ImageLinks continued to make reasonable progress with open technologies as many new capabilities were added to the production system. The remotesensing.org site began to attract attention.
Wired magazine published an online story about remotesensing.org - in typical dotcom fashion, the website and the attention exploded overnight. This triggered calls and interviews from numerous publications. Interested parties in various government agencies contacted us to find out more including the white house science advisor, congressional staffers, and interested parties in the defense and intelligence community. Overnight the remotesensing.org community grew and the number of hosted projects increased. This exposure, and the subsequent technology demonstrations, led to government support of the technology and the continuation of a virtual development team that has survived despite changes in programs and company involvement.
The following is a summary of some of the projects that OSSIM has been involved with:
1999 NIMA/NTA Time Critical Mapping
The National Imagery Mapping Agency (NIMA) through the National Technical Alliance (NTA) provided funding for continued development and support of the remotesensing.org website. The NTA hosts projects to explore advanced geo-spatial technologies. During this period NIMA released the certified source code for conversions between map projections and datums. This code was modernized and added to the open source baseline at remotesensing.org. The director of the NTA was Ms Sue Payton, she became interested in the open source approach.
1999 National Reconnaisance Office (NRO) and NIMA Multi-Modal Image Fusion Conference
The Multi-Modal Image Fusion Conference was established to reach beyond the traditional government contractors and discover what innovative technologies and approaches might be available for processing, analyzing and fusing various types of satellite, aerial and LIDAR data. ImageLinks was selected to participate in this conference hosted by Kodak in Rochester, New York.
With open source tools the team was able to demonstrate various types of cross sensor fusion, change detection and parallel processing with PC clusters on Linux. Immediately after the presentation the NRO and NIMA scheduled follow up briefing and projects to investigate the technology and the open source model. Ed Mahen and Steve Jayjock from the NRO and Bill Allder from NIMA quickly became very interested in the open source model and the technologies being developed. This led to a continuing series of studies, projects and development efforts that have enhanced the OSSIM baseline over the last ten years.
1999 NRO and NIMA OSSIM Phase 1
NRO and NIMA hosted an experiment to evaluate the open source development model. They provided funding for a couple of developers for a year to develop and integrate remote sensing and image processing capabilities with open source practices. Progress was measured against traditional methods, security, configuration management, documentation, and performance. Steve Jayjock and Ed Mahen were our sponsors and they named the project and software OSSIM (Open Source Software Image Map) - a contrived acronym pronounced “Awesome”.
The development team was both excited and apprehensive at the start of the project. The goals were ambitious, and the team knew that open source success is dependent on building and maintaining a contributing community of interest. To attract contributors a project typically has to reach critical mass by solving problems and attracting new users. Therefore, the assumption was that intense development would be required before we could demonstrate the benefits.
Garrett Potts joined the team at the beginning of the project, he began to define all of the software that would have to be written and made sure that even internal discussion used the newly formed OSSIM mailing list. Although many people signed up on the list, the project didn’t see much in the way of contribution. As the software and plans were discussed, an interesting thing occurred - the listeners on the mailing list began to point to other open source projects that had already implemented those capabilities. Often, in the case of mathematical transforms, file management, etc., there would be several implementations to choose from. Rapid progress was made as software development shifted more towards integration with other projects. The GDAL, GeoTrans?, and advanced mathematics libraries were rapidly integrated. At the end of the project the software was able to generate blended image maps and provide simple annotation with a RecceVue? tool developed by Scott Bortman.
One of the challenges was coordinating the changes and additions to the baseline with our government sponsors. The development team was empowered to make technical and functional decisions - from meeting to meeting, the implementation path would change based on the discovery of a new project or capability. At the end of the project the metrics indicated a 15:1 increase in development performance and many of the questions on configuration management, security, governance and discrepancy tracking had been addressed. But what about security and classified development? A follow on experiment was arranged to address those questions.
2000 NRO and NIMA OSSIM Classified Phase 2 - Classified Accuracy Assessment
During the first phase the question of how to address classified capability arose. One of the keys to a successful open source project is attracting and maintaining a large, contributing community of interest. When the question of government use comes up a natural reaction is to propose taking a snapshot of the software inside or restricting development to a gated community. Both of these approaches severely limit the available resource. Taking a snapshot of a code base and going off in a different direction is known as ‘forking’ in the open source community. It is considered a sign of failure as it divides the critical resource - the community of interest. The system needed to be designed to support classified requirements without limiting the unclassified community.
A closer analysis typically shows that classified or sensitive capability can easily be segregated into the data that is used by a program or isolated in a separate shared library or plugin. The focus of OSSIM phase 2 was to build the infrastructure for segregated plugins, develop classified sensor plugins, and test the resulting accuracy in a classified environment. It was during this time that Garrett Potts assumed the role of system architect of the OSSIM baseline. Dave Burken performed most of the development on the classified plugin and maintains it to this day. Observera Inc. was brought in to perform the classified analysis. Accuracy studies were performed in high and low relief areas with the OSSIM baseline and certified government tools. The results were equivalent within mathematical round off error. The plugin architecture that was developed has since been used to add image registration capabilities provided by Dr Frederic Claudel of the South African Council for Scientific and Industrial Research (CSIR) and support for the proprietary Pixia NUI format. It continues to be a key part of the OSSIM infrastructure.
2001 NIMA/NTA Open Source Prototype Research
The previous studies attracted the interest of a key decision maker at NIMA - Mr Bill Allder. Over the years NIMA had funded GOTS packages to help end users with NIMA products and maps. MET and Matrix were examples of government funded development activities that were primarily supported by government funding. Unfortunately, this good will put NIMA in the software development business. The push to move these tools from GOTS to COTS was a well intentioned attempt to continue user support while shifting the development and maintenance costs to a larger customer base. Unfortunately, these tools were very expensive - often in the tens of thousands of dollars per seat, and had a limited market. The end result was that the government transferred control of the source code and still ended up footing the vast majority of the support costs for the tools. Bill wanted to understand the pros and cons of the open source model and determine whether it could address the software situation. Development continued on the OSSIM baseline and additional investigation was performed on the process, governance, and tools that were being used. It is important to remember that open source software development was considered a radical idea during this time period. It is a credit to the government decision makers that sponsored OSSIM that they could see the potential and took the initiative and risks involved to pursue it. The tools and processes were also rapidly evolving in the open source world as well. The reports and studies were updated to show the latest processes involved. At the end of the research, NIMA management wanted to know how they would apply these practices. The team was asked to lay out a transition plan and recommendations for NIMA.
2002 NIMA/NTA Open Source eXtraordinary Program (OSXP)
The OSXP was not another study, it was a transition plan. The previous studies and experiments had demonstrated the value of this approach. Technically, open source development practices significantly out performed traditional approaches. The very nature of the geographically distributed environment evolved tools and practices that maintained configuration management and effective communication between the team members. The OSXP outlined a transition plan and was briefed up the management chain at NIMA.
Ms Sue Payton, the director of the NTA, became a strong supporter of open source development. During this period she was selected as the Deputy Undersecretary of Defense for Advanced Science and Concepts. She started the process of introducing open source there.
2003 CELTIC PATHFINDER evaluation
In the continuing process to reduce government software development costs, NIMA sponsored a formal evaluation to search for and provide commercial replacements for MET/MATRIX GOTS solutions. We decided to sponsor OSSIM through the formal evaluation. OSSIM was one of the tools selected and approved through the CELTIC Pathfinder process. This evaluation serves as the technical approval for NGA and military users of OSSIM for geo-spatial applications. The primary focus areas were image mosaicking, electronic light table functions, and the ability to handle commercial and NIMA formats. As part of the evaluation process a complete set of NIMA data sets was provided to each of the participants. During the outbrief it was stated that OSSIM was the only tool that could load all of the data sets and the evaluation team was impressed with the speed and performance of the tools.
Living beyond a single company.
In April of 2003 Harris Corporation decided to re-acquire ImageLinks Inc. The production team was successfully integrated and continues as a world class geo-spatial production capability to this day. The management at Harris made it clear that they had no intention of supporting OSSIM or open source business models - but offered the development team good positions on existing Harris contracts. The core OSSIM development team decided to leave and went to work for Intelligence Data Systems (IDS) where they were encouraged to continue OSSIM support. This demonstrated another aspect of open source development - the software baseline is not owned by a single company or agency. Contributors are free to continue to work on the baseline if they change companies. Collaboration occurs by default as any improvements to the baseline automatically benefit other agencies using the baseline. It avoids the issues encountered when one agency or company ‘controls and limits’ use of the software as a means of acquiring additional funding or lock in on a program. As the years have gone by the core team has moved between companies and programs, but have continued to collaborate on the OSSIM baseline and leverage the results back into their projects.
2003 GIS Conflation Research Hub
Additional investigation of open source software technologies into specific NIMA problem areas - intelligently merging geo-spatial vector data into geo-spatial databases. The OSSIM team helped coordinate the efforts of external open source projects towards these goals. Open source standards were beginning to rapidly evolve with the OGC WMS interfaces. These interfaces were added to the OSSIM baseline and provided a standardized way to discover and share geo-spatial information over the web.
2004 NGA (formerly NIMA) Geo-spatial One Stop
As part of federally mandated government information sharing, NGA used OSSIM and open source tools to establish a portal to the internet. Unclassified NGA geo-spatial data was shared through the portal as part of the larger eGov initiative.
2004 SCOPES - ossimPlanet
IDS was contacted by Marc Herklotz from Space Command about updating the SCOPES program. SCOPES was primarily developed by Marc to track satellites and their visibility to ground stations located around the world. It contained two and three dimensional displays of the globe, real time and simulated tracks and visibility cones to the globe. It needed to be upgraded to handle more data sets and replace the simple spherical globe with an accurate one that handled full zooming. OSSIM was combined with another major open source project, OpenSceneGraph?. The result was ossimPlanet - an accurate 3D global visualization tool. Garrett integrated the initial version of ossimPlanet into SCOPES. SCOPES is deployed to hundreds of military installations around the world.
2005 - Present NRO/NGA ILabs Classified Applications
The NRO and NGA co-host a number of classified laboratories that provide advanced technology support to the intelligence community. The director Dr. Williams, the deputy director Vic Gonzalez, and the system architect Al Kelly have been strong supporters of the OSSIM baseline. Many of the uses of OSSIM within the labs are classified. OSSIM is being used on a daily basis to provide capabilities for highly classified activities requiring high accuracy and performance.
2005 - OSGeo Foundation
The Open Source Geospatial Foundation was formed to bring together the top open source geospatial projects and take the integration of these technologies to the next level. OSSIM is one of the founding projects, Mark Lucas is a charter member and was elected to the board of directors.
2005 Fighting fires in South Africa
The South African Council for Scientific and Industrial Research (CSIR) has built an open source fire-mapping system that enables organizations to monitor fire outbreaks across South Africa. OSSIM was used to provide the web based processing for the system.
2005 - Emergency Assistance during Hurricane Katrina
Immediately after Hurricane Katrina OSSIM was used to rapidly process aftermath imagery. the processed mosaics were placed on a number of web sites for rapid damage assessment.
2005 - Open Technology Development (OTD)
Ms Payton DUSD AS&C initiated the Open Technology Development effort. The authors of this document were introduced and collaborated on the OTD Roadmap. Cross agency collaboration on OTD was expanded and continues to accelerate.
2006 - Present - Large Data JCTD
OSSIM and ossimPlanet are an integral part of the LD JCTD. The Large Data JCTD employs ossimPlanet for distributed collaboration on very large geo-spatial data sets. Being open source it can quickly be modified and enhanced to establish new interfaces and take advantage of the new infrastructure and capabilities.