Welcome to the repository for GLCM based texture analysis of side scan sonar echograms collected with a recreational-grade system.

Contents

This repository contains all of the data and scripts used to prepare the figures in Alluvial mapping by automated texture segmentation of recreational-grade side scan sonar imagery submitted to Environmental Modelling and Sofware. This repository contains a collection of files and scripts required to perform the analysis.

Organization

This repository is organized as follows:

• /ss_rasters/ contains georeferenced side scan sonar echograms
• /shapefiles/ contains shapefiles of the vidisually identified sediment patches
• /sedclass_rasters/ contains georeferenced sediment classification maps
• /scripts/ contains all of the scripts requried to create sediment classication maps
• /glcm_stats/ contains csv files of GLCM distribuions and GLCM summary statistics
• /glcm_rasters/ contains georeferenced GLCM texture feature rasters

Dependencies

All of the scripts were developed using python 2.7.11 in a windows 10 enviroment. I used the Anaconda distribution 4.0.0 (64 bit) with the MSC v.1500 64 bit (AMD64) compiler. The following dependencies are required:

• gdal 1.11.4 python bindings
• gdal 1.11.4
• rasterstats
• numpy
• pandas
• sklearn
• scikit-image
• joblib
• pyproj 1.9.4

Workflow

All of the continuous side scan sonar recordings were processed using PyHum. In the interest of space, I have not included any of the binary side scan sonar files, intermediate PyHum files, or georeferenced point clouds. If any of those files are of interest, please contact me and I will provide them outside of this repository.

To start, I recommend cloning this repository to c:\workspace. If you want to clone the repository to a different directory, there is a variable clone_root at the beginning of each python script where you can indicate where the appropriate directory.

Beginning with the side scan sonar echogram rasters in /ss_rasters/, you will first need to calculate GLCM texture features using /scripts/GLCM_calc.py. This script will produce georeferenced GLCM texture features in the directory /output/glcm_rasters/.

python glcm_calc.py

Next, you will have to use the shapefiles provided in /shapefiles/ to calculate sediment type statistics. Aggregated distributions and summary statistic CSVs will be saved to /glcm_stats/.

python glcm_stats.py

The statistics saved in /glcm_stats/ are used to calibrate all of the automated texture segmentation algorithms. All of the sediment classification rasters will be output to /sedclass_rasters/. There are individual python scripts for each of the texture segmentation methods in scripts.

python LSQ.py
python gmm2.py
python gmm4.py