The invention discloses a
tuna pond multi-fish culture evaluation method and
system based on ecological analysis, and the method comprises the steps: synchronously collecting the
ecological data and video
stream data of a culture
water area through the deployment of a sensor and a camera device; dividing a multi-fish culture cycle into a plurality of time points, calculating ecological differences based on
ecological data, extracting video association frames, adopting a YOLOV5 model to identify fish school behavior characteristics, and constructing an ecological difference sequence containing
time sequence information and a behavior difference sequence; vectorization
processing is carried out on the difference sequence,
time sequence information is reconstructed through a K-medoids clustering model, and an optimal
time sequence is generated; and importing the optimal time sequence into an LSTM prediction model, respectively establishing an ecological anomaly prediction sequence and a behavior anomaly prediction sequence, and realizing dynamic evaluation and early warning of the breeding state. According to the method, through multi-
modal data fusion and time sequence information reconstruction technologies and in combination with a
deep learning model, the prediction precision of ecological
abnormality and culture risks in a multi-fish
polyculture environment is effectively improved, and a
technical support is provided for intelligent
mixed culture management.