Towed Camera Pipeline

A Marimba Pipeline for processing deep-sea imagery from the CSIRO Marine Resources and Industry Towed Camera (MRITC) platform. The Pipeline specializes in synchronizing high-resolution stereo imagery with environmental sensor data while preserving comprehensive spatio-temporal context.

Overview

The Towed Camera Pipeline is designed to process deep-sea imagery collected during CSIRO marine monitoring campaigns. It handles data from the MRITC platform, which captures both high-resolution stereo still images and video footage synchronized with environmental sensor measurements including Ultra-Short Baseline (USBL) positioning data.

Key capabilities include:

• Automated synchronization of imagery with USBL sensor data
• Organization of imagery and data by deployment
• Integration of comprehensive metadata including position, depth, and camera orientation
• Generation of thumbnail and overview images for visual validation
• Preservation of precise temporal and spatial context
• Generation of FAIR-compliant datasets with embedded metadata

Requirements

The Towed Camera Pipeline is built on the Marimba framework which includes all necessary dependencies. No additional packages are required beyond those installed with Marimba.

Installation

Create a new Marimba project and add the Towed Camera Pipeline:

marimba new project my-towed-camera-project
cd my-towed-camera-project
marimba new pipeline towed_camera https://github.com/csiro-fair/towed-camera-pipeline.git

Configuration

Pipeline Configuration

The Pipeline uses a configuration-free approach, with all necessary parameters derived from input data.

Collection Configuration

Collections requires no specific configuration parameters, as all necessary information is extracted from the source data structure.

Usage

Importing

Import collections by deployment:

marimba import IN2018_V06_001 "/path/to/source/IN2018_V06_001" --operation link

For survey voyages, multiple deployments can be imported sequentially:

# Import multiple deployments
marimba import IN2018_V06_001 "/path/to/IN2018_V06_001" --operation link
marimba import IN2018_V06_002 "/path/to/IN2018_V06_002" --operation link
marimba import IN2018_V06_003 "/path/to/IN2018_V06_003" --operation link

Source Data Structure

The Pipeline expects towed camera data organized by deployment:

source/
└── IN2018_V06_001/
    ├── data/                 # USBL sensor data (CSV)
    ├── stills/               # High-resolution still images
    └── video/                # Video files

Processing

marimba process

During processing, the Towed Camera Pipeline:

1. Creates a hierarchical directory structure by deployment
2. Matches imagery timestamps with USBL sensor data
3. Generates thumbnails from still images
4. Creates deployment overview images for quality control

Packaging

marimba package IN2018_V06 \
--operation link \
--version 1.0 \
--contact-name "Keiko Abe" \
--contact-email "[email protected]"

The --operation link flag creates hard links instead of copying files, optimizing storage for large datasets.

Processed Data Structure

IN2018_V06/                                 # Root dataset directory
├── data/                                   # Directory containing all processed data
│   └── MRITC/                              # MRITC platform data directory
│       └── IN2018_V06_*/                   # Deployment directories
│           ├── data/                       # Deployment data files (CSV)
│           ├── stills/                     # Full resolution images
│           ├── thumbnails/                 # Image thumbnails
│           ├── video/                      # Video files
│           └── IN2018_V06_*_OVERVIEW.JPG   # Deployment overview image
├── logs/                                   # Directory containing all processing logs
│   ├── pipelines/                          # Pipeline-specific logs
│   │   └── towed_camera.log                # Logs from Towed Camera Pipeline
│   ├── dataset.log                         # Dataset packaging logs
│   └── project.log                         # Overall project processing logs
├── pipelines/                              # Directory containing pipeline code
│   └── towed_camera/                       # Pipeline-specific directory
│       ├── repo/                           # Pipeline source code repository
│       │   ├── LICENSE                     # Pipeline license file
│       │   ├── README.md                   # Pipeline README file
│       │   └── towed_camera.pipeline.py    # Pipeline implementation
│       └── pipeline.yml                    # Pipeline configuration
├── ifdo.yml                                # Dataset-level iFDO metadata file
├── manifest.txt                            # File manifest with SHA256 hashes
├── map.png                                 # Spatial visualization of dataset
└── summary.md                              # Dataset summary and statistics

Metadata

The Towed Camera Pipeline captures comprehensive metadata including:

Survey Metadata

• USBL positioning data
• Depth measurements
• Collection timestamps
• Environmental parameters

Technical Metadata

• Image acquisition parameters
• Camera configuration
• Processing parameters
• Quality metrics
• Camera orientation (pitch, roll)

Image-Specific Data

• Precise geographic coordinates
• Temporal information
• Platform details
• Deployment context

All metadata is standardized using the iFDO schema (v2.1.0) and embedded in both image EXIF tags and dataset-level files.

Contributors

The Towed Camera Pipeline was developed by:

• Christopher Jackett (CSIRO)
• David Webb (CSIRO)
• Franziska Althaus (CSIRO)
• Candice Untiedt (CSIRO)

License

The Towed Camera Pipeline is distributed under the CSIRO BSD/MIT license.

Contact

For inquiries related to this repository, please contact:

• Christopher Jackett
  Software Engineer, CSIRO
  Email: [email protected]