基于三维模型的棕点石斑鱼多维表型解析方法研究OA

Fish phenotypic data serve as the foundational basis for intelligent breeding in aquaculture.Modern breeding technologies—such as genomic selection and genome-wide association study—require phenotypic data of greater dimensionality and precision.Conventional manual measurements and two-dimensional image analysis are limited to obtaining a small set of one-dimensional size parameters,which cannot adequately represent the rich morphological information of fish as three-dimensional entities.To address this,a method for three-dimensional phenotypic analysis of fish was developed based on multi-view 3D reconstruction.Multi-view images of the fish body were captured using a rotating three-camera system.Dense point clouds were reconstructed via a structure-from-motion algorithm,followed by the conversion of these point clouds into structured 3D data through the following steps:PCA-based orientation standardization,surface meshing,midline extraction,and cross-sectional slicing.Multi-dimensional phenotypic parameters were then extracted at three levels:(1)3D geometric traits,(2)basic dimensional traits and(3)high-dimensional morphological traits.Validation experiments conducted on 30 individuals of Epinephelus fuscoguttatus demonstrated a 100%success rate in 3D reconstruction.Measurements of total length and body height exhibited high agreement with manual measurements(TL:R2=0.9939;BH:R2=0.9648).The prediction accuracy of body mass using point cloud-derived volume(R2=0.94)was significantly higher than that achieved with conventional total length alone(R2=0.81).Additionally,cross-sectional morphological profiles along the body midline revealed shape pattern variations among individuals with different condition factors,which could not be detected using traditional size-based measurements.In conclusion,the proposed method extends the phenotypic characterization of fish from a handful of conventional one-dimensional size parameters to high-dimensional morphological information—encompassing 3D geometry and along-axis shape variation.This offers a norval phenotypic data foundation for advancing intelligent breeding practices in aquaculture.