Koddie: The Coral Guardian

The Problem

Coral reefs provide vital importance to marine organisms by providing habitats, shelter, and spawning grounds. It also provides economic importance in terms of fishing and tourism for many communities. Around the world, the current state of these coral reefs are under treat, from dynamite fishing to coral bleaching. Marine researcher wills usually spend some hours diving in a small area to conduct photo-plot or transect and do assessment . These are time consuming and sometimes dangerous tasks. We need to have a proactive and wider approach in terms of monitoring and assessment for conservation of these coral reefs.

Our Proposal

Koddie is a swarm of AUV or drones that will perform a combination of photo-plot and transect on the corals. Each AUV will be fitted with three cameras. One high resolution camera at the bottom and two medium resolution cameras at the sides. The bottom camera will capture photographs of the corals. The image processing will be used to identify/classify corals and color matching for for coral health/bleaching indicator (similar to Coral Watch) . The two cameras at the sides will be used to estimate the distance between each Koddie (using triangle similarity algorithm). Each AUV will have an identifying marker on both sides and at the top. The top marker will be used by the monitoring drone to identify and estimate its distance from each other. The monitoring drone will came in two modules: the above water and below water modules. Below water module will be fitted with camera and will estimate and triangulate each position of the Koddies The captured locations of each Koddie will be sent to the above module. Above module will have a GPS to track it s current location and triangulate each position of the Koddies. A processing unit will be able to create a virtual photo-plot or transect while the swarm move above the corals. Images, videos and locations from each Koddies will be downloaded to a computer that will combined all of data to create mapped area and demographics which can then be used to monitor and make assessment of the state and health of the corals.

We Assume that...

7. ROV, AUV and machine learning enthusiast/hobbyist can provide important advise help on the design of the drone. The system will likely create an IoT under water network and drone swarms.

8. There will be a lot of technical challenges in implementing machine vision under water and the cost to build each drone ("Koddies")

6. Power requirements for each modules should be considered. Solar panel for the monitoring drones and longer battery requirements for each Koddies.

5. Minimize the over all cost per unit using available off-the-shelves components.

4. Other researchers will contribute in the image processing algorithms.

3. System can also be used for Benthic and Fish population assessment.

2. Marine researchers/biologists would provide key insights on how they conduct photo-plot and transect assessment.

1. Marine researchers/biologists would like to cover greater area for their survey and assessment.

Constraints to Overcome

The two most important barrier are the image processing implementation and cost. There a lot of image processing algorithms and models out there finding the best fit one will take time and lots of testing. High resolution cameras, single board computers and GPS components will incur junk of the cost. Off the shelves components and open source AUV designs are available for used and will lower the cost for the prototype.

Current Work

Tasks 1: a. Conduct interview with marine biologist/researcher/divers and get information on how they conduct photo-plot and transect b. Conduct interview/email/forum with marine biologist/researcher on how they assess coral bleaching using color chart (as being done by CoralWatch using their CoralHealthChart) Tasks 2: a. Conduct test implementation of machine learning (Object identification and classification) using OpenCV and YOLO. Investigate a new algorithm ("Sea-Thru") developed by Derya Akkaynak. b. Conduct test implementation of machine vision for distance measurement ( triangle similarity or similar algorithms) Tasks 3: a. Refine the over system architecture based from inputs from other researchers and technical experts b. Create technical design and BOM for the monitoring drone and Koddie c. Build a Koddie prototype as cheaply as possible using off-the-shelves components d. Build the monitoring module as cheaply as possible using off-the-shelves components

Current Needs

a. Technical experts/researchers in marine surveys and assessment. Email endorsement to make the initial contact b. ROV and AUV enthusiast. Email endorsement to make the initial contact c. Machine vision and learning experts. Email endorsement to make the initial contact d. Access if possible to Google Coral Dev Boards, GPS devices, Camera (like GoPro or similar) e. Funding for the prototypes and field testing.