Low-Cost Visual-Based Communication for Diver-Assist AUVs

The cognitive autonomous diving buddy (CADDY) is designed to serve as a companion and navigation system for human divers, leveraging an autonomous underwater vehicle (AUV). However, the implementation of this system faces challenges such as reliance on complex sensing devices and the use of wearable devices that limit flexibility. To address these issues, this study offers a cost-effective and simple solution based on computer vision (CV) technology. This approach eliminates the need for wearable devices using hand gesture recognition as a key feature. The system prototype was developed using a single camera, lightweight computing devices, and a simple algorithm. Laboratory testing demonstrated high recognition accuracy, ranging from 85% to 95%. The “Five” gesture achieved the highest accuracy at 95%, while the “Three” gesture had the lowest accuracy at 85%. This system offers not only better flexibility but also a simpler numeric gesture representation compared with conventional methods. However, the system faces challenges such as reduced accuracy for certain gestures. In real-world applications, challenges such as low visibility, varying lighting conditions, and turbulence effects can impact detection stability and gesture processing accuracy, especially if the system is further developed for multi-AUV collaboration. Nevertheless, this study makes a significant contribution by offering a cost-effective and flexible nonwearable approach as a foundation for developing reliable underwater communication systems.