From the Publisher
By Roland Mangold

Madagascar Protected Areas Mapped with GPS-logged Aerial Video and 35mm Air Photos
By Kevin P. Corbley, Lee Hannah and Dana Slaymaker

Resource mapping projects in remote areas often face a difficult choice between spatial resolution and economics. Imaging satellites offer relatively low-cost coverage for large areas, but they have insufficient spatial resolution for many applications. On the other hand, traditional aerial photography techniques acquire much higher resolution images, but are extremely expensive.
      An international conservation group, however, has teamed with a U.S. researcher to fine-tune a low-cost GPS-logged aerial photography technique for a forest monitoring project in Madagascar. The process uses a standard 35mm SLR camera, Hi8 video cameras, unmodified Cessna aircraft, and commercially available image processing software to create geo-referenced and rectified digital photo mosaics with 1.5-meter spatial resolution.

Mapping Challenges in Madagascar
In the past 50 years, rice farming on Madagascar has steadily crept up from the coastal lowlands into the dense tropical forests that lie on the escarpment of the island's central plateau. Because many of these forests are protected in reserves, the Madagascar government asked Conservation International (CI), a non-profit biodiversity conservation group in Washington, D.C., to assist in the conservation of these reserves in partnership with local communities. An important part of this program is their effort to map and monitor the encroachment of farming.
      The CI program focuses on the Ankarafantsika and Zahamena protected areas in northern Madagascar, and is part of a national Environmental Action Plan intended to improve environmental quality, as well as protect the diverse ecosystem and endemic species of the island. In recent years, Zahamena and Ankarafantsika have been threatened by the slash-and-burn practices of local rice farmers. CI is working with farmers to get them to understand and respect reserve limits, while also seeking to find development alternatives to forest destruction.
      Tree clearing is illegal in the reserves, and it requires a permit in most other forest areas. Farmers often enter reserve lands unknowingly or without realizing that they require permits for clearing elsewhere. Monitoring this situation has proved difficult for authorities because the clearings are usually quite small, often less than a hectare, and located many kilometers from the nearest road.
      CI's role is to help monitor the rate of illegal clearing, and to work with farmers in nearby villages on soil preservation techniques, rice storage methods, and sustainable agricultural practices. But right from the start, ecologists participating in this project knew that they faced the challenge of mapping a large area at high resolution, all with a limited budget.
      Project leaders at Conservation International contacted Dana Slaymaker, a researcher of renewable resources at the University of Massachusetts. He has developed a low-cost digital photo mapping procedure for application in U.S. biodiversity inventory work and international resource management projects. Mr. Slaymaker was thereupon contracted to bring his portable mapping system to Madagascar.

A Low-cost Air Photo System
Prior to the first air photo mission over the 53,000 hectare Ankarafantsika Reserve, project participants compared the cost of the 35mm system to that of an Africa-based aerial survey firm that would employ a large-format camera and modified aircraft. The cost to acquire, process, print and scan the 35mm photos was calculated at US$6,850, while the commercial firm would have charged US$41,200 to photograph the same area.
      Every element in the GPS-logged photographic mapping system is commercially available except for the camera mount, which is a custom-made brace capable of holding one 35mm and two video cameras. The mount attaches either to the window frame of a single-engine Cessna, or to the co-pilot seat rails once the seat and door have been removed.
      Other key elements include two Canon Hi8 video recorders used simultaneously, and a pair of Nikon N-90 35mm cameras with quick releases and 18mm wide-angle lenses, employed one after the other as film rolls are used and reloaded. Kodak Royal Gold 100- or 400-speed color film is preferred, depending upon lighting conditions. Two standard laptop computers and an airborne GPS also play important roles while in the air.
      Most system components are interchangeable with other similar commercial products, with the exception of the TNTmips image processing/GIS software from MicroImages of Lincoln, Neb., which was chosen for its unique mapping capabilities. Therefore, all rectification, geo-referencing and mosaicking of the 35mm images was accomplished in the field by using a Pentium desktop computer equipped with TNTmips.

Acquiring the Photos
Prior to takeoff, participants planned each mission to include 60 percent photo overlap within a flight line, and 30 percent sideways overlap between flight lines. Photos were taken about 20 seconds apart, with time enough to switch cameras as each roll of film ran out. In Madagascar, the researchers flew the still photography at an altitude of 3000 meters, dropping later to a lower altitude for the videography.
      While the aircraft was in flight, the GPS was linked to two software packages, each one on a different laptop computer. Geolink software (GeoResearch Inc., Billings, Mont.) controlled navigation by graphically displaying the aircraft's location on a map of pre-planned flight lines. The pilot simply flew up and down the lines as shown on the screen. The second software program, a homemade package called "foto," triggered the 35mm camera's shutter at the programmed coordinate locations. Foto also recorded the time and exact coordinates of each acquisition.
      Once these photos were collected, the pilot descended to 150 meters for video acquisition. Guided by the navigation software, the pilot then flew a second set of flight lines and acquired video footage that lay precisely in the middle of the sideways overlap zone. Both video cameras rolled simultaneously, with one of them set for wide-angle and the other set at 12X zoom. They were linked to a Horita GPS-3 time-code generator, a device that prints the GPS time and frame number on each video frame for accurate correlation to the GPS coordinate data, thus precisely locating each video frame.
      In U.S. biodiversity inventory work, the aerial videography was included in this technique as a ground-feature sampling method for vegetative cover mapping. The wide angle covers an approximately 200-meter swath, collecting about a four percent vegetation sampling rate of the project area at the spatial resolution of one-half meter per pixel. The 12X-zoom video records a narrow swath down the middle of the wide-angle coverage at a resolution of two centimeters per pixel, which is sharp enough to identify most individual plant species for biological inventory purposes.
      In the Madagascar application, wide-angle video was primarily used to geo-reference and rectify the 35mm frames. The video served as an excellent and inexpensive geo-referencing data source, since the geographic center of every frame could be precisely located with time and coordinate data from the GPS. This allowed the researchers to extract control points from the video to rectify the still-photo frames.

Processing the Digital Photos
The ultimate goal of the Madagascar project was to develop accurate, up-to-date maps that would assist CI in its forest conservation efforts. Fully rectified digital photo mosaics of the Ankarafantsika and Zahamena Reserves were created from the video and 35mm images. The digital mosaics were loaded into a GIS so that researchers and local officials could identify deforested areas, thereby planning development programs for each appropriate village. Hard-copy image maps were also printed and taken into the field by researchers.
      Creation of the photo mosaic was conducted on-site in Madagascar using the image processing software, TNTmips. This package was chosen specifically for its ability to display many types and combinations of raster, vector and GPS log files in its reference window during geo-rectification-a crucial capability for this procedure.
      At the end of each day's flying in Madagascar, the film rolls were taken to a one-hour photo shop that produced 4x6-inch color prints. Back at camp, these were scanned at 2700 dots per inch using a Polaroid PrintScan 35 Plus to create 2500x3800 pixel images, which were then imported into the image processing system. An outdated paper map of the reserves had also been scanned into the system, a visual reference for the technicians to use during the registration and rectification process.
      Several windows of data were open simultaneously in the image processing software: the scanned raster map with overlaid flight lines, 35mm raster photo frames, and video frames. The video frames were not imported into the image processing software itself, but instead were digitized with a video frame-grabbing package and then displayed with that software as an overlaid window.
      Using the flight-line photo centers as a guide, the technician placed the first digitized 35mm photo image from a specific flight line into its approximate location on the base map. The GPS points representing each video frame were displayed in the reference window as a series of points labeled by their time code. The technician then viewed the video until he found a frame that had an identifiable feature at its center, such as a bush or a tree. Then he found the same corresponding object in that portion of the 35mm photo image.
      By clicking on the object in both the 35mm image and the video frame, the technician created a control point, one with a known coordinate from the GPS file. With hundreds of video frames available for a single 35mm image, he then followed the video path up and down the 35mm image, repeating the feature-matching process and creating a rectangle of control points. This collection of some 15 points rectified that first 35mm image, with great accuracy, to precise GPS coordinates. The technician then added the rectified 35mm frame to the reference window so that the next frame down the line could be precisely matched within the mosaic.
      The next frame down the flight line was then brought into the geo-referencing window with its overlap area already trimmed and ready to be pieced into the mosaic. The technician proceeded by geo-referencing this, and subsequent 35mm frames, both to the video control points and to the rectified image above, thus creating a continuous image mosaic. This was yet another step where the capabilities of the image processing software proved crucial. The TNTmips offered the unique ability of allowing the technician to place the rectified frames into the reference window, so that all subsequent frames were rectified to them. This further allowed the mosaic to be built as the geo-referencing proceeded.
      The technician continued down the flight line, geo-referencing and rectifying each photo image and then adding it to the mosaic. When the end of the flight line was reached, the process began again with the next flight line, but this time the sideways overlap area of the previously rectified line was used to further rectify the frames.

Studying Results and Mapping the Future
Working at a speed of about 25 photo frames per day, the researchers ultimately built a 450-frame mosaic of Ankarafantsika. Two such mosaics of the reserve were created during two separate acquisition projects in 1996, and 1997. The project staff processed these mosaics into their GIS in order to find new clearings and to begin developing a land cover/land use database of the reserve.
      For Zahamena, individual photos and smaller mosaics were also generated for comparison to black-and-white air photos that had been acquired above the reserve in 1992. These older photos were scanned and displayed in TNTmips in juxtaposition to the 1996 mosaics. In areas where change was detected, technicians geo-referenced the old photos with the new images so that new clearings could be precisely located and mapped.
      The image processing system digitally calculated forest area that had been lost over the four-year period. It revealed a one percent to two percent annual loss in forest cover, a much lower rate than when the CI project had begun. Areas with higher-than-average levels of new clearing were targeted as priority zones for future project action.
      Conservation International has since assembled its own GPS-video logged system with Mr. Slaymaker's help. The process has been employed with several other CI projects in Mexico, Peru, Ecuador, Brazil, Botswana, and the Ivory Coast.
      In 1999, the participants expect to reunite in Madagascar to study a modified version of the technique. The University of Massachusetts researcher has added a pulse laser that will collect terrain elevation information along with the still and video frames. If successful, the researchers hope to generate a three-dimensional view of an area in southern Madagascar.

About the Author:
Kevin Corbley is a freelance writer and consultant who specializes in remote sensing, GIS and GPS. He lives in Denver, Colo. Dana Slaymaker is a researcher in the Department of Forestry and Conservation Management at the University of Massachusetts. He also works as a consultant, conducting resource evaluation projects around the world. Lee Hannah is the senior director of CI's Regional Support Office in Capetown, Republic of South Africa. His special interests include management of protected areas in partnership with local communities, and quantifying the global impact of human alteration of natural habitat.

Back