This work uses a customized high-resolution anti-salt camera to shoot in the water for a long time. The film will be pre-imaged, such as adjusting the brightness and sharpening the sharpness. Finally, the image is input into the designed neural network (R-CNN) for training and identification. This study is currently able to identify shrimp images in the water and distinguish the shrimp head with its tail in order to conclude whether the identified image is a complete shrimp.

At present, in terms of image recognition aspect, artificial intelligence (AI) is currently used for shrimp image recognition after the AI training process by the usage of 365 images and the test of 200 images. As a result, 76% recognition accuracy is obtained. In the future, an additional number of shrimp images will be collected for AI training to achieve higher recognition accuracy.

This system will not only reduce the demanded labor-force in the future but also allow for more immediate observation for large quantities of shrimp. In addition, shrimp farming researchers can use this information to quickly adjust the feed to increase the growth rate of the shrimp.