Underwater object tracking for autonomous surface vehicle

• Main Author: Lidya, Elke
• Other Authors: Tegoeh Tjahjowidodo
• Format: Final Year Project
• Language: English
• Published: 2017
• Subjects: DRNTU::Engineering::Mechanical engineering
• Online Access: http://hdl.handle.net/10356/71990
• Institution: Nanyang Technological University

Coral reefs are important in sustaining life both in the ocean and on land because many lives are dependant on them. Therefore, it is important to develop suitable equipments to make research easier. OpenCV - an open source computer vision library - can be used to track marine life underwater. The Autonomous Surface Vehicle is a remote-controlled boat that has been modified - a Raspberry Pi 3 and a webcam was mounted onboard to control ASV movement. There are four methods tested in this project: frame differencing, background subtraction, shape and color detection, and Haar classifier. Frame differencing and background subtraction are simple algorithms, but they are not suitable for vehicles. Shape and color detection is suitable for tracking colorful objects (most coral reef species are colorful) and it is suitable for moving camera. However, there is a lot of noise and results in false positives. Haar classifier is different from the previous methods as it is a machine-learning based algorithm. Training using thousands of positive and negative images is required to create a cascade. With this method, the program was able to track the object most of the time. False positives in Haar classifiers can also be reduced by using multiple layers of cascades. In conclusion, Haar classifier is the most suitable method for tracking underwater object in shallow waters. In the future, more research should be done on improving response time and improving accuracy.