基于端到端点云分割的鮡科鱼类三维形态表型提取

戴志文; 陈康飞; 常安代; 苏晓静; 谢磊; 杨庆勇; 杨瑞斌; 熊皓; 郑芳

doi:10.13300/j.cnki.hnlkxb.2026.03.013

您当前的位置：

首页 >

文章列表页 >

基于端到端点云分割的鮡科鱼类三维形态表型提取

农业人工智能专题 | 更新时间：2026-11-05

- 基于端到端点云分割的鮡科鱼类三维形态表型提取
- Three-dimensional morphological phenotypic extraction of carnichodidae based on end-to-end point cloud segmentation
- 华中农业大学学报 2026年45卷第3期页码：148-160
- 作者机构：
  
  1.华中农业大学信息学院，武汉 430070
  2.华电金沙江上游水电开发有限公司叶巴滩分公司，甘孜 62715
  3.华中农业大学水产学院，武汉430070
- 作者简介：
  
  戴志文，E-mail：zw_dai@qq.com
  熊皓，E-mail：397149876@qq.com
  郑芳，E-mail：zhengfang@mail.hzau.edu.cn
- 基金信息：
  
  华电集团金沙江上游远期放流鱼种人工繁育技术研究项目
- DOI：10.13300/j.cnki.hnlkxb.2026.03.013
  中图分类号： TP391.4
- 收稿：2025-09-29，
  
  纸质出版：2026-05-25
- 稿件说明：
移动端阅览
戴志文，陈康飞，常安代，等.基于端到端点云分割的鮡科鱼类三维形态表型提取［J］.华中农业大学学报，2026，45（3）：148‐160.

DAI Zhiwen,CHEN Kangfei,CHANG Andai,et al.Three-dimensional morphological phenotypic extraction of carnichodidae based on end-to-end point cloud segmentation[J].Journal of Huazhong Agricultural University,2026,45(03):148-160.
戴志文，陈康飞，常安代，等.基于端到端点云分割的鮡科鱼类三维形态表型提取［J］.华中农业大学学报，2026，45（3）：148‐160. DOI： 10.13300/j.cnki.hnlkxb.2026.03.013.

DAI Zhiwen,CHEN Kangfei,CHANG Andai,et al.Three-dimensional morphological phenotypic extraction of carnichodidae based on end-to-end point cloud segmentation[J].Journal of Huazhong Agricultural University,2026,45(03):148-160. DOI： 10.13300/j.cnki.hnlkxb.2026.03.013.

摘要

为实现鮡科鱼表型的高精度、自动化测量，解决二维图像处理的表型数据不够丰富立体及现有三维点云分割模型对复杂生物结构识别能力不足的问题，本研究以前臀鮡（

Pareuchiloglanis anteanalis

）为具体研究对象，提出一种适用于鮡科鱼类的端到端三维形态表型提取方法。该方法利用RGB相机采集的多视角图像序列，通过运动恢复结构（SfM）与多视图立体视觉（MVS）算法重建高保真三维点云模型。为有效分离目标与背景，设计了一种融合随机采样一致性（RANSAC）算法与HSV空间颜色特征的复合去噪方法，在保留鱼体细节的同时精确去除背景噪声。针对原始PointNet++网络特征提取不充分的问题，提出改进的PointNet++点云部件分割网络，通过引入基于密集连接的多层感知机DenseNet-MLP、点云三重注意力机制PTA和多尺度学习模块MSB，显著提升了模型对鱼体精细结构的分割性能。试验结果显示，MSE低于10e

-5

，PSNR均值为60.8，召回率平均值为93.33%。改进后的网络模型的平均交并比（mIoU）、总体精度（OA）和平均精确度（mAcc）评估指标分别达到了88.9%、95.5%、93.8%，相对原模型分别提升了3.5、1.4、1.6百分点。能够精确分割鱼头、鱼身及鱼鳍等关键部件。综上，本研究提出的方法可用于水产动物表型高通量、无接触式的精准测量。

Abstract

To achieve high-precision and automatic phenotypic measurement of carpenters，and address the limitations of insufficient phenotypic informations in two-dimensional image processing，as well as the inadequate recognition capability of existing three-dimensional point cloud segme

ntation models for complex biological structures.This study focuses on

Pareuchiloglanis anteanalis

and proposes an end-to-end point cloud segmentation-based method for extracting three-dimensional morphological phenotypes of sisorid fish.The proposed method reconstructed a high-fidelity 3D point cloud model from multi-view image sequences captured by an RGB camera using the structure from motion （SfM） and multi-view stereo （MVS） algorithms.To effectively separate the target from the background and retain detailed morphological features of the fish body，a hybrid denoising strategy is designed by integrating the random sample consensus （RANSAC） algorithm with HSV spatial color features，which enables accurate background noise removal while preserving fine fish body details.To overcome the insufficient feature extraction capability of the original PointNet++ network，an improved PointNet++ network for point cloud part segmentation is proposed.By incorporating densely connected multi-layer perceptron （DenseNet-MLP），a point triple attention mechanism（PTA），and a multi-scale learning module （MSB），the model’s segmentation performance for fine fish body structures was significantly enhanced.Experimental results demonstrated that the mean squared error （MSE） was below 10e-5，the average peak signal-to-noise ratio （PSNR） was 60.8，and the average recall rate was 93.33%.The improved network model achieved a mean intersection over union （mIoU） of 88.9%，an overall accuracy （OA） of 95.5%，and a mean accuracy （mAcc） of 93.8%，representing respective improvements of 3.5，1.4，and 1.6 percentage point compared with the original model.It can accurately segment key anatomical components such as the fish head，body，and fins.To sum up，the method proposed in this study can be used to accurately measure the phenotypic qualcomm of aquatic animals without contact.

关键词

Keywords

references

中国科学院动物研究所.前臀鮡（ Pareuchiloglanis anteanalis Fang，Xu et Cui，1984）［EB/OL］.北京：中国科学院动物研究所，2018（2018-01-17）［2025-09-29］.http：//museum.ioz.ac.cn/topic_detail.aspx？id=69081.Institute of Zoology，Chinese Academy of Sciences. Pareuchiloglanis anteanalis Fang，Xu et Cui，1984［EB/OL］.Beijing： National Zoological Museum of China，Institute of Zoo logy，Chinese Academy of Sciences，2018（2018-01-17）［2025-09-29］.http：//museum.ioz.ac.cn/topic_detail.aspx？id=69081 （in Chinese） .

冯云鹏，杨胜保，方善臣，等.野生前臀（鮡）亲鱼摄食驯化技术探究［J］.科学养鱼，2024（5）：75-77.FENG Y P，YANG S B，FANG S C，et al.Study on feeding and domestication technology of wild parent fishes［J］.Scientific fish farming，2024（5）：75-77 （in Chinese） .

姚景龙，陈毅峰，李堃，等.中华鮡与前臀鮡的形态差异和物种有效性［J］.动物分类学报，2006（1）：11-17.YAO J L，CHEN Y F，LI K，et al.Morphological variations and species validity of Pareuchiloglanis sinensis and P.anteanalis in southwestern of China［J］.Acta zootaxonomica sinica，2006（1）：11-17 （in Chinese with English abstract） .

YU H H，WANG Z M，QIN H X，et al.An automatic detection and counting method for fish lateral line scales of underwater fish based on improved YOLOv5［J］.IEEE access，2023，11：143616-143627.

房舒.基于深度学习的鱼类表型数据测量方法研究［D］.杭州：浙江大学，2021.FANG S.Research on fish phenotypic data measurement method based on deep learning ［D］.Hangzhou： Zhejiang University，2021 （in Chinese with English abstract） .

FREITAS M V，LEMOS C G，ARIEDE R B，et al.High-throughput phenotyping by deep learning to include body shape in the breeding program of pacu （ Piaractus mesopotamicus ）［J/OL］.Aquaculture，2023，562：738847［2025-09-29］.https：//doi.org/10.1016/j.aquaculture.2022.738847.

GUO Y L，WANG H Y，HU Q Y，et al.Deep learning for 3D point clouds：a survey［J］.IEEE transactions on pattern analysis and machine intelligence，2021，43（12）：4338-4364.

GUO H，LI Z B，MA Q，et al.A bilateral symmetry based pose normalization framework applied to livestock body measurement in point clouds［J］.Computers and electronics in agriculture，2019，160：59-70.

HU H，YU J C，YIN L，et al.An improved PointNet++ point cloud segmentation model applied to automatic measurement method of pig body size［J/OL］.Computers and electronics in agriculture，2023，205：107560［2025-09-29］.https：//doi.org/10.1016/j.compag.2022.107560.

HE Z Q，XU X Q，LUO J L，et al.A measurement model for aquatic animals based on instance segmentation and 3D point cloud［J］.IEEE access，2024，12：156208-156223.

BARRELET C，CHAUMONT M，SUBSOL G.Automatic coral reef fish identification and 3D measurement in the wild［PP/OL］.arXiv，2023：2312.07357［2025-09-29］.https：//arxiv.org/abs/2312.07357.

SU H，MAJI S，KALOGERAKIS E，et al.Multi-view convolutional neural networks for 3D shape recognition［C］//2015 IEEE International Conference on Computer Vision （ICCV）.Santiago：IEEE，2016：945-953.

HAMDI A，ALZAHRANI F，GIANCOLA S，et al.MVTN：learning multi-view transformations for 3D understanding［J］.International journal of computer vision，2025，133（4）：2197-2226.

MENG H Y，GAO L，LAI Y K，et al.VV-net：voxel VAE net with group convolutions for point cloud segmentation［C］//2019 IEEE/CVF International Conference on Computer Vision （ICCV）. Seoul：IEEE，2019：8499-8507.

CHARLES R Q，HAO S，MO K C，et al.PointNet：deep learning on point sets for 3D classification and segmentation［C］//2017 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）.Honolulu：IEEE，2017：77-85.

QI C R，YI L，SU H，et al.PointNet++：deep hierarchical feature learning on point sets in a metric space［PP/OL］.arXiv，2017：1706.02413［2025-09-29］.https：//arxiv.org/abs/1706.02413.

LI Y Y，BU R，SUN M C，et al.PointCNN：convolution on X-transformed points［J］.Neural information processing systems，2018，31： 820-830.

PHAN A V，LE NGUYEN M，NGUYEN Y L H，et al.DGCNN：a convolutional neural network over large-scale labeled graphs［J］.Neural networks，2018，108：533-543.

GUO M H，CAI J X，LIU Z N，et al.PCT：point cloud transformer［J］.Computational visual media，2021，7（2）：187-199.

SNAVELY N， SEITZ S M， SZELISKI R. Photo tourism： exploring photo collections in 3D［J］. ACM transactions on graphics， 2006， 25（3）： 835-846.

SEITZ S M， CURLESS B， DIEBEL J， et al. A comparison and evaluation of multi-view stereo reconstruction algorithms［C］//Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Washington， DC： IEEE Computer Society， 2006： 519-528.

RUBLEE E， RABAUD V， KONOLIGE K， et al. ORB： an efficient alternative to SIFT or SURF［C］//2011 International Conference on Computer Vision.Piscataway：IEEE， 2011： 2564-2571.

HUANG G，LIU Z，VAN DER MAATEN L，et al.Densely connected convolutional networks［C］//2017 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）.Honolulu：IEEE，2017：2261-2269.

MISRA D，NALAMADA T，ARASANIPALAI A U，et al.Rotate to attend：convolutional triplet attention module［C］//2021 IEEE Winter Conference on Applications of Computer Vision （WACV）.Waikoloa：IEEE，2021：3138-3147.

YU X M，TANG L L，RAO Y M，et al.Point-BERT：pre-training 3D point cloud transformers with masked point modeling［C］//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）.Piscataway： IEEE， 2022： 19291-19300.

QIAN G C，LI Y C，PENG H W，et al.PointNeXt：revisiting PointNet++ with improved training and scaling strategies［PP/OL］.arXiv，2022：2206.04670［2025-09-29］.https：//arxiv.org/abs/2206.04670 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于3D LiDAR感知的大田花生长势信息获取

复合菌剂对猪粪水重金属的钝化效果及沙地燕麦的吸收特征

农业生物质增值化利用技术与装备团队

不同因素对猪粪-木质素共水热碳化产物理化特性的影响

畜禽废水来源地衣芽孢杆菌WS4-CotA漆酶的异源表达、酶学性质及底物选择性