MJP-DCS: A noncontact measurement method for fish phenotypic in recirculating aquaculture systems within simplified calibration environment

Accurate extraction of fish phenotypic parameters is a prerequisite for effective management of digital recirculating aquaculture systems (RAS), which critically influences growth evaluation and feeding strategies. Conventional manual measurement induces physiological stress in fish, limiting the feasibility of high-frequency and non-destructive sampling. Existing machine vision methods are compromised by aquatic environments and varying lighting conditions, resulting in reduced imaging quality and accuracy. The major challenge lies not only in optimizing camera parameters but also in maintaining reliable calibration under limited calibration field conditions in practical RAS environments. In response to problems, a noncontact fish phenotypic measurement system applicable to RAS has been developed. It integrates a full-domain covered scan strategy with a multi-target joint principal distance calibration method, in which (1) a cross-medium multi-camera vision system is developed, integrating synchronized multi-camera triggering with full-domain scanning to achieve noncontact capture of critical feature points information from fish; (2) a joint principal distance calibration algorithm that overcomes traditional calibration field constraints is proposed, which compensates for principal distance deviations from underwater refraction, water turbulence, and camera variations by geometric consistency constraints on critical feature points to achieve high-precision coordinate mapping; (3) from joint calibration results, integrating observation information in a highly accurate forward intersection calculation to extract exterior orientation coordinates for the feature points set, computing total length and interorbital distance while mitigating cross-medium measurement errors and equipment noise. Unlike classical bundle adjustment, which primarily estimates camera parameters via global reprojection error minimization, MJP-DCS focuses on recovering calibration degradation in RAS by introducing full-domain principal distance scanning to identify the optimal geometric solution within a feasible parameter space before joint optimization. Results from laboratory and aquaculture environments demonstrate that the system achieves mean absolute errors under 2 mm for both total length and interorbital distance measurements, demonstrating millimeter agreement with manual reference measurements. The proposed method provides a practical geometric acquisition solution for digital twin aquaculture systems by improving calibration robustness under constrained field conditions.