%0 Journal Article
%T 基于荧光响应的罗非鱼去鳞质量视觉检测方法研究<br>Research on Visual Detection Method of Tilapia Descaling Quality Based on Fluorescence Response
%A 朱鑫宇
%A 李朋朋
%A 邱新静
%A 张春旭
%A 王慧慧
%J Computer Science and Application
%P 1896-1905
%@ 2161-881X
%D 2021
%I Hans Publishing
%R 10.12677/CSA.2021.117194
%X <div style=\"text-align:justify;\">
	<span style=\"font-size:14px;\">罗非鱼前处理过程中去鳞残留直接影响罗非鱼外观品质和商业价值，现有的检测方法具有线下分散性操作以及依赖人工经验的特点，无法满足大规模生产线作业要求。为实现快速、无损的罗非鱼去鳞质量检测，利用去鳞区域与未去鳞区域在360~370 nm波段紫外线下的不同荧光响应，通过罗非鱼加工质量在线监测系统采集图像信息，提取去鳞区域与未去鳞区域的颜色特征，采用主成分分析进行特征的降维与融合。将主成分值作为模型输入，选择卷积神经网络(CNN)，构建罗非鱼去鳞质量检测模型，模型准确度97.5%、精确率98.3%、灵敏度96.7%。该方法在罗非鱼去鳞质量检测应用中具有一定潜力，可为开发实时在线检测装备提供参考。<br />
The descaling residues in the pretreatment of tilapia directly affect the appearance quality and commercial value of tilapia. The existing detection methods have the characteristics of decentralized operation and relying on manual experience, which cannot meet the requirements of large-scale production lines. In order to achieve fast and nondestructive detection of tilapia descaling, the different fluorescence responses of descaling area and undescaled area under ultraviolet range of 360~370 nm band were used to collect image information through the online monitoring system of tilapia processing quality, and the color characteristics of descaled area and undescaled area were extracted. Principal component analysis was used for feature dimensionality redution and fusion. The principal component values were taken as the input of the model, and the convolutional neural network (CNN) was selected to construct the tilapia descaling detection model. The detection Accuracy can reach 97.5%, Precision is 98.3% and Sensitive is 96.7%. This method has certain potential in tilapia descaling detection application, and can provide a reference for the development of real-time online detection equipment.</span> 
</div>
%K 罗非鱼，荧光响应，颜色特征，主成分分析，卷积神经网络<br>Tilapia
%K Fluorescence Response
%K Color Characteristics
%K Principal Component Analysis
%K Convolutional Neural Network
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=43934