Shrimp feeding timing mastered by AI smart agriculture underwater cameras, 30% fatter

On the 23rd, the National Sun Yat-sen University research team held an "AI Smart Aquaculture Results Presentation," using underwater cameras to record shrimp activity and automatic feeding equipment, allowing shrimp to gain 3 grams per week, 30% fatter than traditional shrimp pond farming methods. Currently, 2 shrimp farms have installed underwater cameras and feeding equipment. Over the next 3 years, the research team will take on outdoor farms, hoping to improve the recognition capability of underwater cameras and make them the sharpest eyes underwater for farmers.

Hong Qingzhang, a professor in the Department of Marine Science at National Sun Yat-sen University, said the research team spent 3 years on trials and developed a central drainage system that can be activated based on dissolved oxygen parameters, as well as an AI-recognition-integrated underwater camera system and feeding equipment. Farmers have shown the highest interest in the underwater camera and feeding system.

This underwater camera system uses visible light and infrared technology to monitor shrimp activity around the clock. It can be combined with AI recognition and automatic feeding equipment. When image recognition detects that the feed density in the feeding area is below 15% or another set value, the feeding equipment can be activated by remote control or by the system. On one hand, this can reduce water pollution caused by feed; on the other hand, it can precisely master the timing of feed feeding. Hong Qingzhang said that after 6 months of trials, grass shrimp (Penaeus monodon) raised through this system in the university's seawater aquaculture pond increased their weekly weight gain from 2 grams to 3 grams, 30% fatter than shrimp raised in traditional shrimp ponds.

Huang Yingzhe, a chair professor in the Department of Computer Science and Engineering at National Sun Yat-sen University, added that water quality sensors and underwater cameras have both been developed by companies, but once a camera is installed, "what to look at is what matters." Therefore, the research team contacted farmers early in the project and only began development after understanding their needs. This system can not only integrate data from the pond to help farmers manage shrimp ponds remotely, but the cloud platform also records water quality data, feeding records, and activity status data, allowing farmers' experience to be converted into storable data so it can be replicated in different locations and achieve high-quantity production results.

Hong Qingzhang, who moved from water quality research into the shrimp farming industry, said water quality control is key to the shrimp farming industry. When the water is good, bacteria and viruses are less likely to attack the shrimp. Through AI-recognition-assisted precise feeding with underwater cameras, as well as central drainage facilities integrated with variable-frequency air compressors, the research team raised the survival rate of grass shrimp (Penaeus monodon) to 70% and also saved 30% electricity compared with traditional paddlewheel aerators.

However, the underwater camera equipment has not yet been tested in fully outdoor aquaculture ponds. Hong Qingzhang explained that outdoor aquaculture ponds have different algal ecology and sunlight changes, which test both the image-recording capability of underwater cameras and the recognition capability of AI. Huang Yingzhe, a chair professor in the Department of Computer Science and Engineering at National Sun Yat-sen University, said that in order to respond to the variable and complex environmental data of outdoor aquaculture ponds, the team is currently developing edge AI chip modules. In the future, the system will be able to process data at the pond side without needing to upload it to a cloud platform before making decisions, in order to meet users' network bandwidth and data storage needs, improve the performance of the recognition system, and reduce the disadvantage of high power consumption in underwater imaging systems.