Background: The FOV/GSD Trade-Off

All aerial photography and mapping systems suffer the same trade-off - photographs with a very wide field of view (FOV) increase mapping accuracies, due to the wide stereo angles that are created between overlapping photos. Unfortunately, the same wide field of view reduces image resolution (GSD). This is a serious drawback, because the market constantly demands higher resolutions.

In most large format systems, this trade-off has resulted in a fine balance between field of view and resolution - on average, about 100mm of focal length, which in turn dictates the flight altitude that is required for any image resolution.

This focal length is good for capturing medium-sized areas in resolutions of 25-50 cm. However, when 5-15 cm resolution is mandated, or when the area is very large, many flight days are required to complete the project. This is very inefficient because it increases the cost as well as the risk, due to weather constraints and aviation logistics.

First Problem –Increasing Image Acquisition Efficiency

Since focal length dictates flight altitude for a desired GSD, higher resolution projects require lower flight altitudes. As a result, each flight line captures less area, requiring more flight lines and ultimately more costly aviation time. Because of the trade-off, adding resolution means narrowing the field of view, which means that accuracy is unacceptably compromised.

Most large format cameras attempt to overcome this by building clusters of cameras, each with its own lens and CCD. However, this growth strategy is very limited because such a cluster soon becomes too large and heavy to fly, as well as expensive to manufacture. In addition, a lens with longer focal lengths will be wider and larger, making it more difficult to cluster several lenses into a single system.

VisionMap's solution

VisionMap's team realized that a new technology was needed - one not subject to the traditional FOV/GSD trade off, and which would enable substantially better performance.

VisionMap's solution is a paradigm shift: instead of clustering many cameras into one big system, VisionMap built a single camera that could rotate and take pictures so fast that it would in fact function as one large camera.

The A3 system is comprised of two such cameras, which capture together 160 millions pixels per second, covering more than 100 degrees of field of view, with 300mm focal length. This capability enables the aerial mapping community, at last, to eliminate the FOV/GSD trade-off - substantially increasing aerial mapping efficiency.

The A3 system is comparable to clustering dozens of static cameras together, and provides significant growth potential, leveraging on huge investments in commercial off-the-shelf CCD and CMOS sensors.

In addition, VisionMap's patent-pending technology solves many associated problems, such as photogrammetric algorithms for rotating cameras, how to compress and store vast amounts of data in real-time in airborne systems, how to compensate for camera rotation accurately and more.

Second Problem – Total Automation through Computer Vision

Despite recent improvements, most mapping solutions have not been able to reach sufficient levels of automation. This translates into increased time and costs as projects grow larger.

Furthermore, even a small number of errors in automatic processes can create huge costs, because finding and correcting these errors may cost as an end-to-end manual process.

VisionMap's solved this problem by achieving complete automation - harnessing A3's extremely high pixel capacity and wide field of view to ensure that each object is photographed many times and from many different directions, at very high resolution. The imagery that is captured in each A3 flight provides very high levels of redundancy, ensuring that the automated process is robust.

To enable this, VisionMap developed technology capable of matching and solving blocks of hundreds of thousands of images, as well as millions of tie-points. The result is a totally automatic photogrammetric process that provides very high accuracies, without ground control and without IMUs.