Triton Perspective Map -- SeaClass User Guide -- By: Tony M.
Triton Imaging Inc. Engineering Office 2121 41st Avenue, Suite 211 Capitola, CA 95010 USA 831-722-7373 831-475-8446 sales@tritonimaginginc.com support@tritonimaginginc.com © 2014 TRITON This user guide is provided as a means to become familiar with TRITON’s software through an explanation of the options available for classifying imagery. The user interface presented in this guide is subject to change to accommodate software upgrades and revisions.
Table of Contents 1.0 SeaClass Module ................................................................................. 1 1.1 Supported Formats ............................................................................................................................................. 1 1.2 Classification Process .......................................................................................................................................... 1 1.3 Available Options ..................................
1.0 SeaClass Module SeaClass is an add-on to MosaicOne which allows users to classify sediment types based on a neural net training set created from bottom sample information and select points or other ground truth methods. 1.1 Supported Formats .TMAP_MOZ (Perspective Sidescan mosaic) .DDS_VIF (Triton Map Visual Information File) .TIF (GeoTIFF) The classification procedure operates on the above file formats, and works the same for any of these file types. 1.
2.0 SeaClass File Tree Options 2.1 Tree Structure The SeaClass layer contains three sub-layers as shown in the image below: Manual Training - where training data points are stored Neural Nets - for displaying the neural nets generated from the training data Classification - includes the results of the classification process 2.2 Manual Training Right-clicking on the Manual Training layer or any of its sublayers will give the following options: Color: Available at the class file tree level.
Resample: If a new sample size was set when editing the training set, selecting this will resample the training set at the new size. Remove: Removes the selected layer from the project. This is available at the training set layer and the class layer nodes. Remove All: Removes all sub-layers from the current tree node. Export: Allows users to save the SeaClass training set created. Create: Launches the Create Training Set wizard. Add: Adds an existing training set to the project and Map View.
2.4 Classification Right-clicking on the Classification layer or any of its sub-layers will give the following options: Color Settings: Opens following window for changing background color, opacity, and line drawing settings. Display All On: Makes all samples in the selected training set or the selected sample type visible in the Perspective map view. Display All Off: Makes all samples in the selected training set or the selected sample type not visible in the Perspective map view.
2.0 SeaClass Process 2.1 Prepare Data 2.1.1 Bottom Samples One requirement for the classification process using SeaClass is having bottom information to use for the training set. This information can come from diver observations, video transects, or classical sediment sampling programs. Before beginning the classification process, it is very useful to load the sample information into Perspective as Feature files.
2.2 Create Training Set The first step in the classification process is to enter ground truth information to generate a set of data points to be used in the neural net training process. This can originate from bottom sampling programs or video/diver surveys, or can be entered manually if no bottom information is available.
the classification process, the area around each sample location is compared to the texture of the imagery to develop an understanding of what each bottom type should look like in the imagery. Within each sampling area (set by the Sampling Grid Size), the number of unique values to interpret is a function of the imagery resolution! This is really important to understand!!! So for 0.
Click on the 'Add Class' button to create the class groups. Once a Class has been created, add samples by either selecting the correct feature group in the drop-down list and selecting Add, or by clicking the red Record button and manually double-clicking locations in the map view to add them to the selected Class type. The example to the right shows three classes defined with samples added to the Reef class from the selected feature group.
Feature Groups - Once the "Classes" are defined, the drop-down menu below the manual recording options allow the user to load samples directly from Feature groups in Perspective. - Select the correct feature group from the drop-down list. - Click 'Add' to transfer the samples from the selected feature group to the selected class. - Repeat until all samples have been automatically added to the SeaClass training set.
Click 'Next' to open a dialog to for naming the training set XML file as shown here: After a name is entered for the training class, click ‘Save’ to write the information to an XML file and to load the training set points into the SeaClass file tree as shown below.
3.2 Neural Net Training The second step in the classification process is to use the training set created to train the neural net.
Enter a file name and select 'Save'.
Options available include: Create - use to create a neural net for automated image classification Back - allows users to return to previous pages of wizard to change values entered Finish - select when done creating neural nets and setting parameters 3.3 Classify Bottom Types The last step in the classification process is to use the neural net to automatically classify an image or mosaic file.
Enter a file name and select Save. This will start the classification process.
Page 15
4.0 Edit Training Set As more bottom information is obtained either from additional samples or in-situ observations by a diver or ROV, these locations with known bottom types can be added to the existing training set.
the map view. When finished select the pause button. The record and pause buttons can be seen in the window below: To add points to an existing class, select the class so it is highlighted in the Training Classes and Samples list as shown in the window above. When finished adding the new bottom information to the training set, select Next. This will open the window below for saving the training set information to an XML file for use in the classification process. Enter a name and select Save.
create a new one using the Create button shown in the window below: Click finish when done.
5.0 Quick Classification The creation of the Classification image can take some time, although it is worth noting that the actual classification of the image does not take very long. There is a option called 'Quick Classification ID' that can be used to check if the Classification image will be what was expected before the actual image is created. To activate the Quick Classify button you must complete all steps up to the end of Training the Neural net.
6.0 SeaClass Exports Results from the classification process can be exported through the following methods: 1. Composite Export: This will export all layers turned on in the file tree including the SeaClass results as a composite GeoTiff file. Selecting the following menu option 'File > Export > Composite GeoTiff' will open the export dialog window. 2.
3.
