基于计算机视觉和XGBoost的虾体活力检测  

Shrimp vitality detection based on computer vision and XGBoost

作　　者：冯国富[1,2] 汪峰 陈明 FENG Guofu;WANG Feng;CHEN Ming(School of Information,Shanghai Ocean University,Shanghai 201306,China;Key Laboratory of FisheryInformation,Ministry of Agriculture and Rural Affairs,Shanghai 201306,China)

机构地区：[1]上海海洋大学信息学院,上海201306 [2]农业农村部渔业信息重点实验室,上海201306

出　　处：《湖南农业大学学报（自然科学版）》2023年第2期218-222,共5页Journal of Hunan Agricultural University(Natural Sciences)

基　　金：江苏省科学技术厅项目(CX(20)2028)。

摘　　要：以南美白对虾为研究对象,提出一种基于计算机视觉和XGBoost的虾体活力检测方法:跟踪对虾应激前后的运动轨迹,提取运动行为特征参数;根据应激性红体现象提取对虾的颜色特征,通过灰度共生矩阵(GLCM)提取虾体应激形成水面波动的纹理特征;运用XGBoost算法筛选出评价因子,通过加权融合确定评价因子的最佳权重;根据融合后特征对虾体活力强度进行检测。结果表明,提出的方法决定系数为0.9056,识别准确率为98.61%,较单一颜色、单一纹理以及光流与纹理相结合的方法,识别准确率分别提高6.63%、2.05%和1.61%。Based on computer vision and XGBoost,a method of shrimp vitality detection was proposed by taking Penaeus white shrimp as the research object.Firstly,track the movement trajectory of shrimp before and after stress to extract the movement behavior parameters.The color characteristics of shrimp were extracted according to the stressful red body phenomenon.Secondly,extract the texture characteristics of shrimp with water surface fluctuation forming under stress by using gray scale co-generation matrix,and use XGBoost algorithm to filter the evaluation factors,and determine the best weights of the evaluation factors by weighted fusion.Finally,the shrimp vitality intensity was detected according to the fused features.The results showed that the decision coefficient of the proposed method was 0.9056 and the recognition accuracy was 98.61%,which improved by 6.63%,2.05%and 1.61%compared with the single color,single texture and combined optical flow and texture methods,respectively.

关 键 词：虾体活力检测 计算机视觉 XGBoost 特征融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]