As promised, here is the second post from the paper “Benefits of UltraMap for Keystone Aerial Surveys” from David Day, Director of IT for Keystone Aerial Surveys Inc. Because of the length of this section of the article, I am breaking it down into two parts, the second to follow in a next post. If you did not read the first post on this topic, you can find it here.
– Jerry Skaw, Sales and Marketing, Microsoft UltraCam Team
Monolithic Stitching and Accuracy The concept of using lower resolution imagery to enhance the tie point matching of higher resolution imagery is not new to aerial triangulation and is used by all automated AT packages. The idea of using it to refine the point matching used in creating the virtual image in a multi-cone camera seems to be a unique idea for camera manufacturers. Monolithic stitching is also expected to increase the stability of the frame and, as the manufacturer notes, leads to ‘one sensor geometry’. In an effort to determine the extent of the increased stability, Keystone conducted extensive testing using a repeatedly flown block located near its facility. The block consists of 4 lines of 7 exposures each flown north/south in an alternating pattern at approximately 1250m for a GSD of 8.8cm. The block is covered with nearly 100 control points and is part of our boresight.
Applanix PosPac software was used to process the IMU/GPS and INPHO’s Match-AT was used for aerial triangulation. Each block flown with the old COS (version 3.2) was processed using the Office Processing Center (OPC version 3.5.1) and each block flown with the new COS (version 3.6) was processed using UltraMap (version 2.1.2). Several of the blocks flown with the old COS were also processed from LVL00 to LVL03 using UltraMap. The blocks were acquired over an 18 month period with 15 unique flight dates and 3 dual OPC and UltraMap processing blocks.
In order to try to maximize the triangulation’s dependency on the geometry of the imagery and the quality of the tie points, four control points were used at each corner of the block and no automatic camera calibrations were used. The remaining points were used as check points and were never used in the automatic tie point extraction or in the bundle block adjustment. The results of these runs showed that the blocks run using four control points and IMU data proved to be the most accurate (having the lowest RMS at the check points). While the RMS values are not as low as similar studies, they are acceptable and not the main objective of the exercise. Instead, the goal of these runs was to determine any improvement when using UltraMap as compared to OPC. The table below presents the results.
Of the three blocks that were processed with both OPC and UltraMap, UltraMap processing showed higher, lower and nearly identical RMS values respectively. The results above are inconclusive, however the average RMS shown below confirm that there is some improvement with using UltraMap processing.
This is especially true when UltraMap is used with the new COS and does seem to show that monolithic stitching provides a level of increased vertical accuracy even in areas that are heterogeneous in nature. Factors such as atmospheric conditions, quality of the flight lines and radiometric factors certainly could contribute to the fluctuations in the Z. But an increased number of quality tie points due to a more geometrically sound image may account for the nearly 20cm improvement.
[Watch for the next blog post for this topic: Monolithic Stitching and Accuracy Part II]