CIG Ottawa Branch

Mosaicking Mexico - The Big Picture of Big Data

New earth observation satellites provide petabytes of high resolution imagery on a daily basis, and offer high temporal revisit rates that make it possible to monitor changes on a global scale. Despite these new sources of data, processing and converting the imagery available into seamless and cloud free mosaics continues to present challenges. While low and medium resolution composites already have been produced for large areas of the world, a closer review of these datasets through online platforms, such as Google Earth, reveal issues with the quality of the composites when viewed at high resolutions. For example, visible seamlines and clouds exist in the imagery.

The presentation will include a description of how to create a completely seamless and cloud-free mosaic of Mexico at a resolution of 5m, using Rapideye ortho-imagery. Mexico is the world’s 13th largest country with regards to land area. As a result, producing a mosaic to cover the entire landmass at 5m resolution requires approximately 4,500 Rapideye images. To complete this project in a timely manner and with limited operators, a number of processing architectures were required to handle the volume. Furthermore, Mexico contains a variety of diverse ecosystems with different land cover and seasonal effects. The diversity of the land cover results in images having different radiometric profiles, ranging from very bright desert regions to dark dense vegetation with atmospheric contaminants (cloud and haze). As such, an algorithm capable of balancing the images while maintaining the natural fidelity of the colors and excluding cloud cover was required.

In view of these statistical benchmarks, a number of operations were required to overcome the challenges of generating a cloud-free seamless mosaic for Mexico at 5m resolution. This paper will discuss the different operations required to complete this project, which include, preprocessing, mosaic generation and post mosaic editing. Prior to mosaic generation, it was necessary to filter the 50,000 Rapideye images captured over Mexico between 2011 and 2014 to identify the top candidate images, based on season and cloud cover. Upon selecting the top candidate images, PCI Geomatics’ GXL system was used to reproject, color balance and generate seamlines for the output 1TB+ mosaic. This paper will also discuss innovative techniques used by the GXL for color balancing large volumes of imagery with substantial radiometric differences. Furthermore, post-mosaicking steps, such as, exposure correction, cloud and cloud shadow elimination, and LUT editing will be presented.

Kevin Jones, PMP, EMBA (Candidate)

Director – Marketing and Communications

PCI Geomatics

Kevin Jones has worked in the field of Geomatics since 1996, and currently holds the position of Director, Marketing and Communications at PCI Geomatics. Mr. Jones has been involved in the acquisition, processing, and dissemination of information products derived from Earth Observation imagery throughout his career. Mr. Jones also has extensive experience in bringing technology to market including product design, development, testing, evaluation, and commercialization. He holds an Undergraduate degree in Geography from Carleton University, is a Project Management Professional, and is currently working towards the completion of an Executive Masters in Business Administration.