ODIS (Obstacle Detection Inside the Sea)

• GIF

• 

• 

• 

• 

Abstract

• One of the problems facing by underwater vehicles and ships is in detecting any incoming obstacle to avoid a collision which might lead to hazards. Thus, this project focused on the design of an underwater obstacle detection system using a sonar sensor.
• This project uses the sonar sensor as distance detection to determine the distance between sensor and obstacle and camera to predict the type of obstacle is it using Deep Learning.

Introduction

• Every day, many boats and ships run on the water, and most of the ships have well defined navigating systems with radar technologies but most boats don’t have! Also, the navigating systems are present for the upward detection not underwater.

  • There are many cases we find most of the time where the boats or ships sink due to the underwater obstacles like huge ice blocks, rocks, or any other underwater hindrances. But we found there were no certain devices or technologies which help in predicting underwater obstacles to detect the objects accurately and warn the pilot for the same. Also, if some devices are there, those are much expensive which are usually not affordable by the boatmen! Considering the said problem, we have come up with a unique and low-cost solution.
  • Introducing, ODIS (Object Detection inside Sea); a smart device that will detect the underwater obstacles with the help of underwater radar and image processing. The device will detect the underwater obstacles with the help of radar technology and with the help of image processing we can get to know about what type of object is it and image processing with deep learning will increase the accuracy of the system.

Components

1. Arduino Uno (for now later in the actual model we can use Raspberry Pi for the whole system)
2. Underwater radar senor/Ultrasonic Sensor
3. Camera with thermal image processing
4. Laptop/PC/Mac (a system where we can run the model for now).

Radar Sensing

-The radar sensing will be based on simple ultrasonic technology. The ultrasonic waves can travel inside the water. It works in the following principle:

-Using the help of this technology, we can detect underwater obstacles. The sensors will be fitted around the four corners of the ship/boat for better accuracy.

Block Diagram

Workflow

Image Processing

-Image processing with Deep learning has great scope in detecting the objects accurately. It can also detect what type of object is it then it can respond accordingly.

-We’ll be using the Keras Model for the prediction and the images are trained using Google’s Teachable Machine. It’ll basically detect the object under the water and give a result of whether there are obstacles or not.

In details

• The radar sensor will create a virtual underwater map that will be displayed on the dashboard. The camera will provide the live feedback of the underwater scenario and the algorithm will scan the obstacles and give a perfect prediction before warning. If there will be an obstacle in front of the boat/ship then it'll give pre-warnings to the pilot. The warning will be via buzzer, lights! -The combined system, radar sensing, and objection will make the entire system more accurate. -In this way, ODIS can control the hazards due to the obstacles which will save many lives and cost also!

Testing Link: Link

Challenges we ran into

Due to the pandemic, COVID19, we can't get the chance to go out and test our work. We faced major problems in the image processing part where we have to work with the mobile phone for photo testing.

Accomplishments that we're proud of

We are proud that we created the system which will help many navigators and boats during their journey!

What we learned

During the work, we found many exciting things about the ocean and ocean lives. Also, we learned Tensorflow and Keras during our project!

What's next for ODIS (Obstacle Detection Inside the Sea)

• The next plan for ODIS is to implement and test. Realtime data will help us to get more accurate value and results!

Built With

• arduino
• javascript
• machine-learning
• python
• tensorflow