Set Classification Thresholds and Prioritize
Use the Set Classification Thresholds and Prioritize option (Geosoft.uxo.gxnet.dll(Geosoft.GX.UXO.SetThresholdAndPrioritize;Run)*) from the menu AGC Expert User > Source Classification and Review to classify the sources as either targets of interest or not and to create a prioritized list in the source database.
Set Classification Thresholds and Prioritize dialog options
Source Database | |||||||
Name | Select the source database file. Script Parameter: UXANALYZE.SOURCEGDB | ||||||
Group | Select a group from the list of groups in the source database. Script Parameter: UXANALYZE.SOURCEGROUP | ||||||
ID channel | Select the ID channel, typically UXA_Target_ID. Script Parameter: UXANALYZE.TARGETID | ||||||
Mask channel | Select the mask channel in the source database to identify for which sources to classify. Script Parameter: UXANALYZE.TARGETMASK2 | ||||||
Gates | Select the Low, High, and Decay gates, which are used to calculate the size and decay thresholds.
Script Parameters:
| ||||||
Time of gates (ms) | Read-only parameter that displays the time of gates from the survey database. | ||||||
Library Database | |||||||
Name | Select the signature library. This library is a Geosoft database containing polarizability curves for a variety of known targets of interest (munitions) and optionally known clutter items. Script Parameter: UXANALYZE.LIBRARYGDB | ||||||
Group - TOI | Select the library group for TOI (target of interest) only. Script Parameter: UXANALYZE.LIBRARYGROUP_TOI | ||||||
Group - non TOI (Clutter) | Select the library group for non -TOI or clutter items. Script Parameter: UXANALYZE.LIBRARYGROUP_CLUTTER | ||||||
Classification Thresholds | |||||||
Minimum signal amplitude (mV) | Specify the minimum signal amplitude. If the amplitude is less than the minimum value, it will be defined as non -TOI. This value can depend on the sensor and the type of signal amplitude (Monostatic Z vs Coaxial Z) selected when inverting the data. Default value is 2. Script Parameter: UXANALYZE.MIN_SIGNAL_AMP | ||||||
Maximum signal amplitude (mV) | Specify the maximum signal amplitude. If the amplitude is greater than the maximum value, the fit results are treated with more confidence. Default value is 20. Script Parameter: UXANALYZE.MAX_SIGNAL_AMP | ||||||
Minimum fit coherence | Specify the minimum acceptable fit coherence. Default value is 0.8. Script Parameter: UXANALYZE.MIN_FIT_COHERENCE | ||||||
Maximum inverted depth BGS (m) | Specify the maximum inverted depth in meters. Default value is 2. Script Parameter: MAX_FIT_DEPTH | ||||||
Maximum distance - array position & inverted location (m) | Specify the maximum distance in meters between sensor array position and inverted location. Default value is 0.4 Script Parameter: UXANALYZE.MAX_DIFF_ARRAY_INV_LOCATION | ||||||
Maximum distance - array position & flag location (m) | Specify the maximum distance in meters between sensor array position and flag (anomaly) location. Default value is 0.75 Script Parameter: UXANALYZE.MAX_DIFF_ARRAY_FLAG_LOCATION | ||||||
Maximum distance - flag location & inverted location (m) | Specify the maximum distance in meters between flag (anomaly) location and inverted location. Default value is 0.6 Script Parameter: UXANALYZE.MAX_DIFF_FLAG_INV_LOCATION | ||||||
| Decision Statistics |
| ||||||
Minimum source distance for multiple TOI at a flag (m): | Specify the minimum distance apart for which two sources will be deemed as two separate items. Default value is 0.2. Script Parameter: UXANALYZE.MIN_MULTISOURCE_DIST | ||||||
[More] | |||||||
Classification Options | |||||||
Classify sources with noisy polarizabilities as Cannot Analyze (Cat 0) | Datasets with poor or noisy polarizabilities can result in lower-than-expected decision statistic. A conservative approach is to classify these items as "Cannot Analyze". Check this field to enable the parameters below:
Script Parameter: UXANALYZE.CLASSIFY_SOURCES_POLARIZ [0 - unchecked; 1- checked] | ||||||
Classify sources using only the primary polarizability match | For some datasets with poor or noisy secondary and tertiary polarizabilities, which can lead to a lower-than-expected decision statistic. A conservative approach is to classify them as TOI based on a high primary polarizability match only. If checked, the field below will be enabled:
Script Parameter: UXANALYZE.CLASSIFY_POLARIZABILITY_MATCH_SELECTED [0 - unchecked; 1- checked] | ||||||
Use custom classification expressions | Select the option to use custom classification expression files (see Application Notes below). If checked, the fields below will be enabled:
Script Parameter: UXANALYZE.CUSTOM_CLASSIFICATION_SELECTED [0 - unchecked; 1- checked] | ||||||
Ranking Options | |||||||
Rank using manual TOI selection | Check this option to manually select source items to be classified as TOI. If checked, the field below will be enabled:
Script Parameter: UXANALYZE.RANK_USING_MANUAL_TOI_SELECTION [0 - unchecked; 1- checked] | ||||||
Rank training data at the top of the list | Check this option to place items that were used for training data (i.e., help selecting thresholds) at the top of the list. If checked, the field below will be enabled:
Script Parameter: UXANALYZE.RANK_TRAINING_DATA_SELETED [0 - unchecked; 1- checked] | ||||||
Application Notes
*The GX tool will search in the "gx" folder. The GX.Net tools, however, are embedded in the Geosoft.uxo.gxnet.dll located in the bin folder. If running this GX interactively, bypassing the menu, first change the folder to point to the bin folder, then supply the GX.Net tool in the specified format.
How it Works
This tool will classify a list of sources as TOI (target of interest) or non -TOI and will create a prioritized list. TOI are usually discarded munitions and unexploded ordnance (UXO), while non -TOI are using clutter or other metallic debris.
The classification uses several metrics, but the key metric is the decision statistic - the average of the available library match metrics. Typically, four match metrics are calculated (with criteria weights of 111, 110, 101 and 100). The classifications are:
Cannot Analyze – Category 0; there is insufficient and/or poor-quality data. We are unable to reliably classify the source, and additional data should be collected at these locations or otherwise further investigated.
TOI – Category 1 and 2; category 1 being high confidence TOI.
Non -TOI – Category 3; the source does not match TOI, and it is likely metallic debris or clutter.
Prior to running this tool, you should run, either as part of the Classify and Rank or individually, the following:
Invert for Sources (Dynamic) or Invert for Sources (Static) – to create the source database with the polarizabilities, fit coherence, fit depth, and signal amplitude channels
Calculate Size and Decay Channels – to calculate the appropriate size and decay channels
Calculate Axial Symmetry and Polarizability Noise – to calculate the polarizability noise (optional)
Inversion QC – to determine if the source passes the inversion quality control
Perform Library Match – to calculate the library matches and metrics. You should use the standard nomenclature (i.e., 111TOI) for the suffix, and run four times using different criteria weights (i.e., 111, 110, 011 and 100).
The general process to classify the sources is:
Calculate the various distance metrics for each source.
Calculate the decision statistic, and based on the various thresholds perform the initial classification.
Calculate initial ranking and prioritization.
Temporarily update the ranking to prioritize the source that can be classified.
Sort the source list to identify sources which are essentially “duplicates”, that is within the minimum distance for multiple TOI of each other, selecting the one with the best library match.
Update the classification for sources that either:
- exceed the distance metric and are not high confidence TOI or
- are manually selected to be a TOI.
Set the “UXO Type” from the best library match.
Update the ranking and prioritization.
Finalize the classification to prioritize sources that cannot be analyzed (or optionally training data), and for those where a decision statistic could not be calculated, set it to a value of 0.5.
Finalize the prioritization ordering the list of sources.
Create the final group, which only contains the classified sources, while removing the “duplicates” and ensuring that each flag location is represented by at least one source.
Output Channels
This tool creates the following channels in the source database, with the important ones highlighted in bold:
- UXA_mask_classify
- UXA_mask_toi
- UXA_rank
- UXA_rank_index
- UXA_subcategory
- UXA_UXOTYPE
- UXA_Category
- UXA_decision_statistic
- UXA_diff_array_fit
- UXA_diff_flag_array
- UXA_diff_flag_fit
- UXA_QC_array_fit
- UXA_QC_flag_array
- UXA_QC_flag_fit
- UXA_QC_xy
- UXA_flag_id
- UXA_combined_LmStat
- UXA_cat_temp
Custom Classification Expressions
Expert users can create custom classification expressions by copying the standard expression files from the "C:\Program Files\Geosoft\Desktop Applications \etc" folder and editing them. The files are:
Rank_toi.exp or Rank_toi_clutter.exp
Rank_distance.exp
These expression files are used in steps 2 and 6 as noted in the general process above. When editing, you can place your custom file in your project folder or in a location that you prefer. Care should be taken to preserve the structure of the file and variables.
Got a question? Visit the Seequent forums or Seequent support
© 2024 Seequent, The Bentley Subsurface Company
Privacy | Terms of Use