Radiometric Calibration

Sensor

Expected value: One value from the available list (required)

Select your Specim sensor from the list of supported sensors. If your sensor is not listed please contact hello@geona.io with details of the sensor you have.

Note

This field is automatically set if you upload a Sensor Lookup File

Scan lines to process

Exepcted value: Integer (optional)

If you wish to only process part of a single flight line, perhaps because of a known navigation data failure or cloud cover, you can specify the first and last lines to process. This is a zero-indexed list, meaning the first scan line is line 0. If you specify the first line or last line which is outside the maximum number of lines you will get an out-of-range error.

Flipping Options

Expected value: Check to enable (optional)

By default, the calibrated, geocorrected data cube that results from Geona Hyperspectral processing will be in the same order as the raw data which is the input. However, options are available to reverse the order, either horizontally left-to-right or by reversing the order of the bands, depending on your particular system configuration.

Flip data left to right: Also referred to as horizontal flipping, this reverses the scan line from left to right. If your instrument has been mounted backwards in reference to the direction of flight, this option should be applied.

Note

This field may be automatically set if you upload a Sensor Lookup File

Flip band order: Reorders bands such that 1…n becomes n…1.

Enforce no flipping: This will ensure that data are not flipped at all.

Masking

The masking process identifies potentially bad data and provides a mechanism to remove these data from the final product. After the calibration stage, the masking process may produce two mask files; one delivers mask information and the other describes the method used to identify masked pixels.

Mask information

The mask file flags are numbers and meanings are written into the mask .hdr files as comments. The mask file has a one-to-one representation to the radiometrically calibrated file, .e.g they have the same dimensions and pixel X in the mask applies to pixel X in the calibrated file.

Value	Meaning
0	Good data.
1	Underflows.
2	Overflows.
4	SWIR detector bad pixels.
8	Pixel affected by uncorrected smear.
16	Dropped scans.
32	Corrupt raw data.
64	Quality control failure.

Mask method files

There may be a mask bad pixel method file created also. This is a supplemenary file that contains information on the methods used in detecting the “SWIR detector bad pixels”. The information will depend on the calibration laboratory used to perform the sensor calibration and how the bad pixels have been detected and marked bad. These are methods used by the UK NCEO/NEODAAS calibration laboratories. Method G relates to those from a Specim BPR file.

Value	Method name	Methodology
0		Not flagged as a bad detector pixel.
1	methodA	ConstantLightInconstantResponse. Data for given pixel varies greater than given threshold. Thresholds=10.0, 10.0
2	methodB	DeviationFromSpectralAndSpatialNeighbours. Data for given pixel varies from spectral and spatial neighbours greater than given threshold. (spectral, spatial)= (3, 10), (3, 10). Threshold= 20.0, 10.0
4	methodC	LinearInputNonLinearResponse. Correlation of pixel integration time against data value is less than given threshold. Threshold= 0.9825, 0.9825
8	methodD	RapidSaturation. Correlation of pixel integration time against data value varies from spectral and spatial neighbours greater than given threshold. (1, 18), (1, 18). Threshold= 3.0, 3.0
16	methodE	ManualDetection. Manual detection of bad pixels from level 1 test set.
32	methodF	LowResponse. If the average spectral response of the band is less than a threshold %% of the average maximum response for the sensor. Threshold= 0.3, 0.1
64	methodG	Identified from BPR file acquired at data collection.

Mask scan lines are corrupt

(optional) Provide a space separated list of integer values for scan lines which should be removed from the raw data prior to calibration. The line number is a zero-indexed array, meaning the first scan line is line 0. For example, if you have known errors or anomalies in the flight navigation data which will produce poor results, these lines can be excluded.

Only use specific flags

(optional) The mask file created to identify poor quality or corrupt data (see Mask information) defines flags for each pixel in the data cube. By default, all flags greater than zero are applied, however, you can specify which flags should be applied as a space separated list of flags to ignore. Valid options are 1, 2, 4, 8, 16, 32, 64

List of band/sample pairs that should be masked

(optional) Select a file from the input data location. This file should be text file containing a list of space separated band sample pairs (one pair per line) of pixels to mask as quality control failures. Note, these pixels are in the raw data geometry with band/sample starting from 0.

First line of file should be sub-sensor id. This is 0 for Eagle, Hawk and Fenix VNIR, 1 for Fenix SWIR. Note the Fenix SWIR bands should start from 0 (not 348). In cases where you need to mask a sample for every band you can use ‘N’ in replace of the band number to represent all bands.

This option allows masking of pixels that are not picked up by routine processing procedures. It expects a plain text file containing a list (one per line) of space separated band sample pairs. As some sensors have more than one detector it is also important to highlight which detector the pixel belongs to. In most cases this is 0. For Fenix(1K), which has 2 detectors, 0 describes the VNIR and 1 the SWIR. To mask every band for a particular sample you can use N in place of the band number. Note that bands and samples range from 0 to N-1. An example is given below:

0
0 0
1 0
55 87
1
0 0
N 100

Explanation line-by-line:

0 = define the detector to use
0 0 = mask out band 0 and sample 0 (e.g. the first pixel of the first band on detector 0)
1 0 = mask out band 1 sample 0 (e.g the first pixel on the second band on detector 0)
55 87 = mask out sample 87 on the 56th band
1 = define the next detector to use - only relevant for Fenix sensors
0 0 = mask out band 0 sample 0 (e.g. the first pixel of the first band on detector 1)
N 100 = mask out pixel 100 for all bands in detector 1

Bad pixel file

(optional) File with a .bad file extension. This will default to using the calibration file but with a .bad extension if such a file exists.

The bad pixel file contains information about pixels that have been determined bad, usually during a laboratory calibration. It is a plain text file that lists the band and samples to be masked out including an identification of the method used to assign the pixel as bad. The file can include a header of variable length, with the size of the header given by the first line.

Example format:

headerlines=5
methodF = LowResponse.
methodG = Identified from BPR file acquired at data collection.
badpixels = 483
id band sample method(s)
0 349 581 G
1 351 32 G
2 351 77 F
3 351 140 G
4 351 490 G
....

We can see from the header that, in this example, there were 2 methods used to identify bad pixels: methods F and G, and that 483 pixels in total have been identified as bad.

Going through the example line by line, the header above tells us:

headerlines= number of lines in the header before the first bad pixel row
methodF = A description of method F used to identify bad pixels.
methodG = A description of method G used to identify bad pixels.
badpixels = the number of bad pixels in the file. This together with the headerlines should be the total number of lines in the file.
id band sample method(s)  are the Column descriptors for the data that follows, and should always be in this order

The values that follow the header are a list related to the column descriptors. The id column is an arbitrary identification number which increments each row, the band and sample describe the pixel that is bad, the method is the NERC NCEO calibration lab description of the bad pixel. Each line refers to one pixel to be masked.