US 12,352,845 B1
PIT-based fish school trajectory monitoring method and system for marine ranch
Yunrong Yan, Zhanjiang (CN); Zhao Li, Zhanjiang (CN); Feng Wang, Zhanjiang (CN); Qing Feng, Zhanjiang (CN); Jianan Zheng, Zhanjiang (CN); Zhenyu Jiang, Zhanjiang (CN); and Qi Lin, Zhanjiang (CN)
Assigned to Guangdong Ocean University, Zhanjiang (CN)
Filed by Guangdong Ocean University, Zhanjiang (CN)
Filed on Mar. 18, 2025, as Appl. No. 19/082,845.
Claims priority of application No. 202411449241.4 (CN), filed on Oct. 17, 2024.
Int. Cl. G01S 13/75 (2006.01); G06N 20/00 (2019.01)
CPC G01S 13/75 (2013.01) [G06N 20/00 (2019.01)] 8 Claims
OG exemplary drawing
 
1. A passive integrated transponder (PIT)-based fish school trajectory monitoring method for a marine ranch, comprising:
using a PIT reader to obtain PIT tag information of each fish of a target fish school during a monitoring period to thereby obtain PIT tag information of the target fish school during the monitoring period, wherein each fish of the target fish school is implanted with a PIT before monitoring, the PIT reader is configured to: in response to the fish implanted with the PIT passing through the PIT reader, identify the PIT and generate the PIT tag information of the fish implanted with the PIT, and the PIT tag information of each fish of the target fish school comprises: a PIT tag, a time stamp, a first longitude coordinate, a first latitude coordinate and a first vertical depth;
generating, according to the PIT tag information of the target fish school during the monitoring period, a three-dimensional (3D) fish school trajectory map of the target fish school during the monitoring period, wherein the 3D fish school trajectory map consists of 3D motion trajectories of all fish in the target fish school, a 3D motion trajectory of each fish of the target fish school comprises multiple 3D spatial points, and a 3D coordinate of each of the multiple 3D spatial points comprises: a second longitude coordinate, a second latitude coordinate, and a second vertical depth;
generating, according to the 3D motion trajectory of each fish of the target fish school, a motion sequence of each fish of the target fish school, inputting the motion sequence of each fish of the target fish school into a motion trajectory correction model, and outputting, by the motion trajectory correction model, a corrected motion sequence of each fish of the target fish school to thereby obtain corrected motion sequences of all fish of the target fish school;
generating, according to the corrected motion sequences of all fish of the target fish school, a first corrected 3D fish school trajectory map of the target fish school during the monitoring period, and generating a second corrected 3D fish school trajectory map according to the first corrected 3D fish school trajectory map and the 3D fish school trajectory map; and
generating motion information of the target fish school according to the second corrected 3D fish school trajectory map, and feeding back the second corrected 3D fish school trajectory map and the motion information of the target fish school to a monitoring personnel, wherein the motion information of the target fish school comprises: a migration direction of the target fish school, and a movement speed of the target fish school during the monitoring period;
wherein a construction process of the motion trajectory correction model comprises:
obtaining historical motion sequence samples of multiple fish species, wherein each sample of the historical motion sequence samples comprises multiple historical 3D coordinate samples with continuous time stamps;
constructing an initial motion trajectory correction model, wherein the initial motion trajectory correction model comprises a classification module and a trajectory correction module, the trajectory correction module comprises trajectory correction sub-models, each of the trajectory correction sub-models corresponds to one fish species of the multiple fish species;
training the classification module of the initial motion trajectory correction model with the historical motion sequence samples of the multiple fish species until the classification module reaches a first preset convergence condition, obtaining network parameters of the classification module in the first preset convergence condition, and freezing the network parameters of the classification module to obtain frozen network parameters of the classification module;
training each of the trajectory correction sub-models with the historical motion sequence samples of the one fish species of the multiple fish species corresponding to each of the trajectory correction sub-models until the trajectory correction sub-model reaches a second preset convergence condition, obtaining network parameters of each of the trajectory correction sub-models in the second preset convergence condition, and freezing the network parameters of each of the trajectory correction sub-models to obtain frozen network parameters of each of the trajectory correction sub-models; and
adjusting network parameters of the initial motion trajectory correction model according to the frozen network parameters of the classification module and the frozen network parameters of each of the trajectory correction sub-models, and generating the motion trajectory correction model; and
wherein the generating, according to the PIT tag information of the target fish school during the monitoring period, a 3D fish school trajectory map of the target fish school during the monitoring period comprises:
classifying the PIT tag information of the target fish school according to the PIT tag of the PIT tag information of each fish of the target fish school, to obtain PIT tag information groups of all fish of the target fish school;
generating a 3D motion trajectory corresponding to each of the PIT tag information groups according the time stamp, the first longitude coordinate, the first latitude coordinate, and the first vertical depth, to thereby obtain 3D motion trajectories corresponding to the PIT tag information groups;
dividing, according to the PIT tag of the PIT tag information of each fish of the target fish school, the target fish school into fish species, and obtaining 3D motion trajectories of the fish species of the target fish school;
fitting the 3D motion trajectories of the fish species of the target fish school to generate fitted 3D motion trajectories of the fish species of the target fish school; and
generating, according to the fitted 3D motion trajectories of the fish species of the target fish school and the 3D motion trajectories corresponding to the PIT tag information groups, the 3D fish school trajectory map of the target fish school during the monitoring period.