Identification of fish species based on image processing and statistical analysis research
Identification of fish species based on image processing and statistical analysis research
复制标题
基于图像处理和统计分析研究的鱼类种类识别
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
Jinqi Hong
中科院分区：
文献类型：
--
作者：
Lian Li;Jinqi Hong
对于较多的鱼类品种，要进行后续的加工和销售，这就需要对鱼类的种类进行分类。本研究是基于图像处理技术对鱼类的种类进行分类，利用现有的四种鱼类图像采集设备采集样本，通过MATLAB软件对图像进行预处理，如鱼类灰度化、二值化、图像增强、轮廓提取等，提取四种鱼类的11个特征参数，如利用主成分分析（PCA）对11个特征参数进行降维，本研究选取了4个主成分。然后利用SPSS软件建立Fisher和Mahalanobis距离模型，结合四个主成分再利用主成分构建模型对四种不同的鱼类进行分类。通过SPSS软件模拟和识别结果表明，平均识别率为96.67%，可以很好地应用于鱼类种类识别技术。
For more fish varieties, to the subsequent processing and marketing, which is necessary to classify the types of fish. This study is based on image processing technology to classify the types of fish, four fish to make use of the existing image acquisition device for collecting samples, through the MATLAB software for image preprocessing, such as fish gray, binarization, image enhancement, contour extraction to extract the 11 feature parameters of four fish species, such as using the principal component analysis (PCA) to 11 characteristic parameters for dimension reduction, this study took four principal component. Then use SPSS software to establish fisher and mahalanobis distance model, the combination of four principal component reuse component to build a model to classify the four different kinds of fish. Through SPSS software simulation and identification results show that the average recognition rate of 96.67%, which can be well applied to the fish species identification technology.