A Deep Learning-Based White Shrimp Body Length and Bait Detection Methodology for Turbid Underwater Aquaculture Environment

White shrimp accounts for one-half of all farmed shrimp species in Taiwan. It has extremely high nutritional value and is more adaptable to harsh environments than other aquatic products. The time from production to sale is relatively short. Therefore, it has a very high economic output value, and because of this high economic value, the problems that need to be considered during the breeding period will also increase. During the breeding of white shrimp, much manpower is required to measure water quality, feed, and estimate the body length of white shrimp, and finally, it can be sold when it reaches a certain body length. There are many matters in the monitoring of white shrimp farming. In this regard, this paper focuses on body length estimation and bait estimation, and the detection of white shrimp to evaluate the meat exchange rate of subsequent white shrimp farming. However, in actual outdoor aquaculture farms, the underwater quality is not clear but very turbid. Therefore, this paper proposes a deep learning-based white shrimp body length and bait detection methodology for breeding monitoring to solve the problem that the turbid underwater aquaculture environment is not easy to identify and monitor.