A Highly-Integrated Power-Efficient Underwater Observing Device

Project Intro
Shrimp aquaculture is highly developed in Taiwan and throughout the world, but conventional aquaculture methods for shrimp are time-consuming and laborious. Therefore, advanced technology should be used to reduce costs and increase production through the use of underwater image observation, water quality monitoring, and artificial intelligence data analysis. To overcome the low visibility issue in the shrimp pond, this project will combine the deep learning of artificial intelligence and VLSI design techniques to develop a highly integrated, low-power, low-bandwidth intelligent underwater observing device. Through the clear images obtained by the underwater observing device, we can recognize and analyze the characteristics of shrimp and environmental conditions, and then understand the status of the shrimp colony at any time and achieve intelligent, automated aquaculture decision-making and management. In the first and second years, we will develop the shrimp image detection and recognition system based on the deep neural network technique and the corresponding hardware acceleration circuit of the neural network. As a result, the observing device will have recognition and analysis capabilities, and then only the useful images are recorded and delivered to reduce the required network bandwidth, storage space, power consumption, as well as the workload of subsequent image monitoring and analysis. In the third year, we will develop the super-resolution image reconstruction technique based on deep neural network, and the high-performance and low-cost image enhancement circuit to enhance the image quality and improve the recognition rate. In the fourth year, the developed hardware/software components will be integrated with the video coding circuit into a complete embedded system to reduce the overall building and operating costs of underwater observing device to reduce the total cost. The embedded system will be displayed in the outdoor aquaculture pond to demonstrate the advantage of the intelligent underwater observing device and management system.