Estimation of Mass Characteristics for a Rainbow Trout Based on Individual Linear Sizes in Underwater Video Surveillance System

Abstract

Video Surveillance Systems (VSS) are progressing towards the industry digitalization. In this demo, we apply the Artificial Intelligence (AI) technology to estimate the mass characteristics for a rainbow trout. To estimate the linear sizes, the key points of fish body are recognized in video recording of an individual rainbow trout. We constructed a data set for 356 individual fish with approximately 50 thousand samples. The mass depends on the fish size, so regression “length-mass” is implemented. The Histogram-based Gradient Boosting Regression tree is used as an appropriate regression model. Our early experiment shows the accuracy about 91% in the mass estimation.

Keywords

• aquaculture digitalization
• image recognition
• video surveillance system
• regression