基于前后端生成概率密度图模型的虾苗自动计数  

Automatic counting of shrimp larvae based on probability density map model generated from front end and back end

作　　者：蔡润基 彭小红[1] 叶双福 张天晨 高月芳[2] 吕俊霖[3] CAI Runji;PENG Xiaohong;YE Shuangfu;ZHANG Tianchen;GAO Yuefang;LYU Junlin(School of Mathematics and Computer,Guangdong Ocean University,Zhanjiang 524088,China;College of Mathematics and Informatics College of Software Engineering,South China Agricultural University,Guangzhou 510642,China;South China Sea Fisheries Research Institute,Chinese Academy of Fishery Sciences,Guangzhou 510300,China)

机构地区：[1]广东海洋大学数学与计算机学院,广东湛江524088 [2]华南农业大学数学与信息学院,广东广州510642 [3]中国水产科学研究院南海水产研究所,广东广州510300

出　　处：《南方水产科学》2025年第1期173-184,共12页South China Fisheries Science

基　　金：农业财政专项项目(NFZX2021);广东省对虾现代种业智慧平台(2022GCZX001);对虾智慧繁育关键技术研究与应用示范(2023ZDZX4012)。

摘　　要：虾苗计数是虾类养殖与销售中生物量估算的基本操作,对养殖管理和提高产量至关重要。由于受虾苗体型小、遮挡严重以及密集程度不一等因素影响,现有的自动计数方法难以在一定密度下准确计数。基于此,提出了一种基于贝叶斯概率分布的前后端结合计数网络(Shrimp counting, SC)模型,用于解决虾苗计数问题。该模型主要由前端模块、注意力模块和后端模块构成。首先,使用前端模块提取具有判别性的表型特征,并采用注意力模块对特征进行重组,以提升对图像的局部注意力;随后,使用后端模块生成虾苗分布预测概率密度图;最后,通过贝叶斯损失函数对模型进行参数调整,以提升虾苗计数的精确度。为了验证方法的有效性,构建了一个包含2种不同密度的虾苗计数数据集,并在该数据集上进行了多组实验对比。结果显示,总体准确率可达93.325%,平均绝对误差和均方误差分别为96.304和154.567。与现有主流的计数方法 [布莱克-利特曼模型(Black-Litterman, BL)、人群密度估计网络(Contextual Scene Recognition Network, CSR-Net)、多维注意力增强人群计数模型(Boosting Crowd Counting via Multifaceted Attention, BCCMA)]相比,SC模型准确率最高、误差最小。该模型适用于孵化场、销售和养殖入塘等多场景的虾苗自动计数。Shrimp larval counting is a fundamental operation for biomass estimation in shrimp farming and sales,and it is cru-cial for aquaculture management and yield enhancement.Due to the factors such as small size of shrimp larvae,significant oc-clusion,and varying density,the current automated counting methods are difficult to achieve accurate larval counting at certain densities.To address this issue,we propose a shrimp counting(SC)model for shrimp larval counting which combined frontend and backend counting network based on Bayesian probability distribution.This model primarily consists of a frontend module,an attention module and a backend module.The frontend module first extracts discriminative phenotypic features,while the at-tention module reorganizes these features to enhance local attention to the images.Secondly,the backend module generates a predicted probability density map for shrimp larval distribution.Thirdly,the Bayesian loss function is utilized to adjust the model parameters and improve the accuracy of shrimp larval counting.To validate the effectiveness of the proposed method,we con-structed a shrimp larval counting dataset with two different density conditions and conducted multiple experimental compari-sons on the dataset.The overall accuracy reached 93.325%,with mean absolute error and mean squared error being 96.304 and 154.567,respectively.Compared with the current mainstream counting methods[Black-Litterman(BL),Contextual Scene Re-cognition Network (CSR-Net),Boosting Crowd Counting via Multifaceted Attention BCCMA],the proposed model exhibits the highest accuracy and the lowest loss.It applies to automated shrimp larval counting in hatcheries,sales and stocking scenarios.

关 键 词：虾苗计数 密度图预测 计数网络模型 注意力模块 贝叶斯损失 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]