We have developed a technique for creating large mosaics of underwater images without the need to refer to external sensors such as heading, roll, pitch, or acoustic position. The process of creating the mosaics and the people involved are described under a separate web site entitled: Underwater landscape video mosaics for coral reef mapping and monitoring.
The project was sponsored by the Strategic Environmental Research and Development Program (SERDP), for the purpose of assisting DoD (primarily Navy) efforts to map and monitor coral reefs. The mosaicing technology, however, is not inherently restricted to reef applicaitons, or even to underwater imagery.
On this page, we show the results of processing aerial imagery using the underwater mosaicing software. The raw datasets were acquired with a Nikon D200 over the the Guania Dry Forest, in southwest Puerto Rico, covering an area of about 55 km^2. This area is both a UNESCO Biosphere Reserve and monitoring site for the proposed National Ecological Observatory Network (NEON). Two study sites were processed, and results are shown below.
Figure 1: Guanica West. Mosaic created from 163 individual frames. Image is presented here at drastically reduced resolution.
Figure 2: Guanica West. Zoom at higher, but not full, spatial resolution.
Figure 3: Guanica East. Mosaic created from 413 individual frames. Image is presented here at drastically reduced resolution.
For the purposes of this experiment, no effort has been made to blend or color balance the individual frames, so some seams are visible in the final mosaics. The alignment between images is very good (see Figure 2) considering no GPS, INS, DEM, camera calibration, or manual input has been used to construct the mosaics. Processing time was about 2 hours for each mosaic using a dual 2.4 GHz quad-core workstation.
The mosaic software used here is not a substitute for traditional photogrammetric processing where the best possible accuracy is required from a final orthomosaic. Rather, the significance of this experiment is that fairly accurate results can be generated very rapidly using extremely inexpensive off the shelf hardware. Such an approach would be valuable for mosaicing imagery from light UAVs, where cost, weight, and rapid rsults are priorities.
To learn more about the underlying mosaic technology or to contact us for more information, follow the link to our project underwater landscape video mosaics for coral reef mapping and monitoring.
