Class cwautonav
The autonavigation tool automatically determines a navigation correction based on earth image data.
Name
cwautonav - automatically determines a navigation correction based on earth image data.
Synopsis
cwautonav [OPTIONS] locations-file variable input
Options:
-c, --correlation=FACTOR
-f, --fraction=FRACTION
-h, --help
-H, --height=PIXELS
-m, --match=PATTERN
-M, --minboxes=N
-s, --search=LEVEL
-S, --separation=DISTANCE
-t, --test
-v, --verbose
-w, --width=PIXELS
--version
Description
The autonavigation tool automatically determines a navigation correction based on earth image data. The algorithm is as follows:
- Step 1 - The user supplies a number of boxes of coastal data to use for navigation. The boxes are specified by the latitude and longitude of each box center in a text file separate from the earth data file. The box dimensions are controlled by command line options.
- Step 2 - Each box is run through an offset estimation algorithm. The algorithm first attempts to separate the pixels within a given box into two classes: land and water. If the classes are sufficiently separable, an image correlation is run by "shifting" the image data around to find the maximum correlation between land/water classes and a precomputed land mask database.
- Step 3 - All navigation boxes with successful offset estimates are used to compute the mean offset for the entire input file. All user-specified variables in the input file are then corrected with the mean offset.
Note that because of the autonavigation algorithm design, there are a number of limitations:
- Coastline features - The algorithm relies partly on the user being able to specify navigation boxes containing "wiggly" coastline features such as peninsulas and bays. A flat coastline can cause the algorithm to generate inaccurate offset estimates.
- Distinct classes - Image data in the navigation boxes must be separable into distinct land and water classes. If the image data contains cloud, or if the land and water pixels do not differ significantly (too similar in visible or thermal radiance), then the class separation step will fail for some boxes.
- Large areas - The mean offset generated from the set of successful offset estimates in Step 3 may not model the actual navigation correction for data files that cover a large physical area. If the offsets differ significantly for navigation boxes at a great distance from each other, then the user should treat a number of subsets of the data file separately.
- Rotation or scaling - An offset correction cannot model the actual navigation correction if the data requires a rotation or scale correction.
Note that satellite channel data or channel-derived variables should be corrected with navigation but GIS-derived variables such as coastline and lat/lon grid graphics should not be corrected. Applying a navigation correction simply establishes a mapping between desired and actual data coordinates — it does not change the gridded data values themselves. Once a data file has been autonavigated successfully, other CoastWatch tools in this package will take the correction into account when reading the data.
See the cwnavigate tool in this package for details on how to set a navigation correction manually, or to reset the existing navigation.
Parameters
Main parameters:
- locations-file
- The file name containing a list of navigation box centers. The file must be a text file containing center points as latitude / longitude pairs, one line per pair, with values separated by spaces or tabs. The points are specified in terms of Earth location latitude and longitude in the range [-90..90] and [-180..180].
- variable
- The variable name to use for image data.
- input
- The input data file name. The navigation corrections are applied to the input file in-situ. For CoastWatch HDF files, the corrections are applied to individual variables. No other file formats are supported.
Options:
- -c, --correlation=FACTOR
- The minimum allowed image versus land mask correlation factor in the range [0..1]. If the image data matches the precomputed land mask to within the specified correlation factor, the navigation is considered to be successful. The default correlation factor is 0.95. Caution should be used in lowering this value, as it has a significant impact on the quality of navigation results.
- -f, --fraction=FRACTION
- The minimum allowed class fraction in the range [0..1]. The class fraction is the count of land or water pixels from the class separation stage, divided by the total number of pixels in the navigation box. If the fraction of either land or water pixels is too low, the image data is rejected. The default minimum fraction is 0.05.
- -h, --help
- Prints a brief help message.
- -H, --height=PIXELS
- The navigation box height. By default, each navigation box is 100 pixels in height.
- -m, --match=PATTERN
- The variable name matching pattern. If specified, the pattern is used as a regular expression to match variable names. Only variables matching the pattern will have the navigation correction applied. By default, no pattern matching is performed and all variables are navigated.
- -M, --minboxes=N
- The minimum number of successful navigation boxes needed to apply the navigation correction. The default is 2.
- -s, --search=LEVEL
- The search level starting from 0. This option should only be used if the magnitude of the navigation correction is likely to be half or more the size of the navigation box, as it can significantly increase the algorithm running time. In these cases, the offset estimation can often fail because the image data is so far off the correct geographic features that the class separation and image correlation steps are meaningless. When this option is specified, an area of (n+1)^2 times the size of the navigation box is searched, where n is the search level. By default, only image data within the navigation box is used (n = 0).
- -S, --separation=DISTANCE
- The minimum allowed class separation distance in standard deviation units. Typical values are in the range [1..4]. The greater the distance, the more distinct the land and water classes are. The default distance is 2.5.
- -t, --test
- Turns on test mode. All operations that compute the navigation correction are performed, but no actual correction is applied to the input file. By default, test mode is off and the input file is modified if a correction can be computed.
- -v, --verbose
- Turns verbose mode on. Details on offset estimation and navigation correction are printed. The default is to run with minimal messages.
- -w, --width=PIXELS
- The navigation box width. By default, each navigation box is 100 pixels in width.
- --version
- Prints the software version.
Exit status
0 on success, > 0 on failure. Possible causes of errors:
- Invalid command line option
- Invalid input file name
- Unsupported input file format
Examples
The following example shows an automatic correction of an East Coast HDF file containing AVHHR channel 2 data. A total of 3 navigation boxes are specified in a text file, and the size of each box set to 60 by 60 pixels. The output shows that 2 of the 3 boxes were successful and a final navigation correction of (rows, cols) = (-3, 1) was applied to the file.
phollema$ cwautonav -v --width 60 --height 60 navbox.txt avhrr_ch2 2004_064_1601_n17_er.hdf [INFO] Reading input 2004_064_1601_n17_er.hdf [INFO] Testing box at 37.0503 N, 76.2111 W [INFO] Land/water class separation distance = 2.33 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 44.2783 N, 66.1377 W [INFO] Land/water class separation distance = 3.436 [INFO] Image correlation = 0.965 at offset = (-3, 1) [INFO] Box offset = (-3, 1) [INFO] Testing box at 45.1985 N, 65.9262 W [INFO] Land/water class separation distance = 3.814 [INFO] Image correlation = 0.987 at offset = (-3, 1) [INFO] Box offset = (-3, 1) [INFO] Mean offset = (-3, 1) [INFO] Applying navigation correction
The next example below shows the import and automatic correction of multiple CWF files from the Gulf of Mexico. The AVHRR channel 1, channel 2, SST, and cloud mask variables are first imported to an HDF file. The automatic correction then runs using only data from AVHRR channel 2 which provides high contrast between land and water during the day. The final correction is applied to all variables in the input file. This combination of import and autonavigation is a convenient way of correcting a set of CoastWatch HDF data files all at once, using just data from AVHRR channel 2.
phollema$ cwimport -v --match '(avhrr.*|sst|cloud)' 2004_313_1921_n16_mr*.hdf 2004_313_1921_n16_mr.hdf [INFO] Reading input 2004_313_1921_n16_mr_c1.hdf [INFO] Creating output 2004_313_1921_n16_mr.hdf [INFO] Converting file [1/4], 2004_313_1921_n16_mr_c1.hdf [INFO] Writing avhrr_ch1 [INFO] Converting file [2/4], 2004_313_1921_n16_mr_c2.hdf [INFO] Writing avhrr_ch2 [INFO] Converting file [3/4], 2004_313_1921_n16_mr_cm.hdf [INFO] Writing cloud [INFO] Converting file [4/4], 2004_313_1921_n16_mr_d7.hdf [INFO] Writing sst phollema$ cwautonav -v --width 60 --height 60 navbox2.txt avhrr_ch2 2004_313_1921_n16_mr.hdf [INFO] Reading input 2004_313_1921_n16_mr.hdf [INFO] Testing box at 26.7734 N, 82.1731 W [INFO] Land/water class separation distance = 3.239 [INFO] Image correlation = 0.945 at offset = (-2, 1) [INFO] Insufficient correlation [INFO] Box failed [INFO] Testing box at 29.1666 N, 83.0324 W [INFO] Land/water class separation distance = 3.54 [INFO] Image correlation = 0.985 at offset = (-2, 1) [INFO] Box offset = (-2, 1) [INFO] Testing box at 29.9141 N, 84.3543 W [INFO] Land/water class separation distance = 4.514 [INFO] Image correlation = 0.976 at offset = (-2, 0) [INFO] Box offset = (-2, 0) [INFO] Testing box at 30.3258 N, 88.1352 W [INFO] Land/water class separation distance = 3.006 [INFO] Image correlation = 0.954 at offset = (-3, 0) [INFO] Box offset = (-3, 0) [INFO] Testing box at 27.8423 N, 82.5433 W [INFO] Land/water class separation distance = 3.59 [INFO] Image correlation = 0.953 at offset = (-2, 1) [INFO] Box offset = (-2, 1) [INFO] Mean offset = (-2.25, 0.5) [INFO] Applying navigation correction
Another example below shows the correction of a Hawaii AVHRR HDF file using many 15 by 15 pixel navigation boxes distributed throughout the islands. AVHRR channel 2 data is used to compute the optimal offset, and the final correction is applied only to AVHRR sensor bands and derived variables.
phollema$ cwautonav -v --match '(avhrr.*|sst|cloud)' --width 15 --height 15 navbox3.txt avhrr_ch2 2005_042_0051_n16_hr.hdf [INFO] Reading input 2005_042_0051_n16_hr.hdf [INFO] Testing box at 21.7885 N, 160.2259 W [INFO] Land/water class separation distance = 1.537 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 21.9856 N, 160.0938 W [INFO] Land/water class separation distance = 1.395 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 21.6033 N, 158.2847 W [INFO] Land/water class separation distance = 1.562 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 21.7144 N, 157.9678 W [INFO] Land/water class separation distance = 1.982 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 21.0961 N, 157.3207 W [INFO] Land/water class separation distance = 1.517 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 21.2448 N, 157.2547 W [INFO] Land/water class separation distance = 2.252 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 21.2076 N, 156.9774 W [INFO] Land/water class separation distance = 3.236 [INFO] Image correlation = 0.973 at offset = (-2, 0) [INFO] Box offset = (-2, 0) [INFO] Testing box at 21.1581 N, 156.7001 W [INFO] Land/water class separation distance = 2.293 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 20.9225 N, 157.0698 W [INFO] Land/water class separation distance = 1.448 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 20.7115 N, 156.9642 W [INFO] Land/water class separation distance = 1.506 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 21.3067 N, 158.1130 W [INFO] Land/water class separation distance = 1.601 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 21.3067 N, 157.6508 W [INFO] Land/water class separation distance = 1.593 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 20.5374 N, 156.7001 W [INFO] Land/water class separation distance = 5.142 [INFO] Image correlation = 0.978 at offset = (-2, 0) [INFO] Box offset = (-2, 0) [INFO] Testing box at 20.5499 N, 156.5680 W [INFO] Land/water class separation distance = 2.834 [INFO] Image correlation = 0.947 at offset = (-2, -1) [INFO] Insufficient correlation [INFO] Box failed [INFO] Testing box at 20.5996 N, 156.4360 W [INFO] Land/water class separation distance = 2.285 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 20.8108 N, 156.5152 W [INFO] Land/water class separation distance = 2.092 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 20.9349 N, 156.4756 W [INFO] Land/water class separation distance = 3.629 [INFO] Image correlation = 0.969 at offset = (-2, -1) [INFO] Box offset = (-2, -1) [INFO] Testing box at 20.8357 N, 156.1190 W [INFO] Land/water class separation distance = 2.46 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 20.2635 N, 155.8813 W [INFO] Land/water class separation distance = 2.522 [INFO] Image correlation = 0.964 at offset = (-2, 0) [INFO] Box offset = (-2, 0) [INFO] Testing box at 19.5140 N, 154.7985 W [INFO] Land/water class separation distance = 1.64 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 19.7392 N, 155.0230 W [INFO] Land/water class separation distance = 1.833 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 19.7267 N, 155.1022 W [INFO] Land/water class separation distance = 1.688 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 18.9119 N, 155.6965 W [INFO] Land/water class separation distance = 1.378 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 19.8642 N, 155.9342 W [INFO] Land/water class separation distance = 1.583 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 19.0375 N, 155.8813 W [INFO] Land/water class separation distance = 1.638 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 22.0349 N, 159.7901 W [INFO] Land/water class separation distance = 1.919 [INFO] Insufficient separation [INFO] Box failed [INFO] Testing box at 22.1826 N, 159.3279 W [INFO] Land/water class separation distance = 1.496 [INFO] Insufficient separation [INFO] Box failed [INFO] Mean offset = (-2, -0.25) [INFO] Applying navigation correction
- Since:
- 3.1.9
- Author:
- Peter Hollemans
-
Method Summary
-
Method Details
-
main
Performs the main function.- Parameters:
argv
- the list of command line parameters.
-