计算机视觉技术在鱼新鲜度检测中的应用研究

Study on detection method for fish freshness based on computer vision

针对传统的新鲜度检测方法存在着操作复杂、耗时和具有破坏性等缺点,研究尝试利用计算机视觉技术对鱼的新鲜度进行快速无损检测。利用自行设计的图像采集装置采集鲫鱼在4℃恒温条件下储藏不同天数的图像,运用数字图像处理技术从采集图像中分别分割提取出鱼眼虹膜、鱼鳃、体表的颜色以及体表的纹理等感兴趣区域图像特征信息,对这些信息采用PCA降维,建立BP神经网络模型对鱼储藏天数进行预测,最佳主成分数为8,训练集样本的分类正确率可达到94%,测试集的达到85%。研究结果表明,计算机视觉技术应用于鱼新鲜度的检测具有可行性。

Freshness is an important indicator of the quality of fish products.Traditional methods for freshness detection are operating complex,time-consuming and destructive.The method based on computer vision technology is employed to detect fish freshness.The self-designed image acquisition device is used to capture the images of carp stored at 4℃C temperature.Color information of fish-eye iris,gills and fish surface,and texture of fish surface are extracted as features of fish images using the digital image processing technology.These attributes are integrated and reduced dimensionality using PCA algorithm.The BP neural network is used to establish identification model for prediction of storage time of carp,the experiments show that the best number of principal component factor is eight,and correct classification rate for training set reaches 94％,for prediction set is 85％.The results show that it is feasible to detect the fish freshness using computer vision technology.