【作者】 王巧华；

【导师】 文友先；

【作者基本信息】 华中农业大学 ， 农业机械化工程， 2004， 硕士

【摘要】 鸡蛋具有很高的营养价值，深受消费者喜爱。蛋的内部品质的好坏直接影响食用品质。在销售流通及加工方面，如能做到按新鲜度分级并按质论价，则既保护了消费者利益，又有利于生产经营者采取科学的管理，保证蛋的品质。因此，如何高效快速地检测鸡蛋的新鲜度，是亟待解决的课题。 鸡蛋新鲜度研究原理为：根据蛋的光学特性可知，鲜鸡蛋对光的透射率会随着贮存时间的延长而下降。光的透射率变化，表明蛋的内部品质发生了变化，这种变化可以用鸡蛋内容物光信息的变化反映出来。 本研究用机器视觉及神经网络技术建立鸡蛋的新鲜度(哈夫值)与鸡蛋光信息参数之间的相关关系，进行鸡蛋新鲜度无损检测与分级。鸡蛋新鲜度模型建立过程是：先在试验装置上提取鸡蛋的颜色信息(H、I、S)，试验设备由光源与光室、摄像头与图像采集卡及计算机组成；光源与光室提供检测鸡蛋图像的环境；摄像头摄取被检测鸡蛋透射光图像并由图像采集卡数字化后送入计算机；通过编制的鸡蛋图像分析程序，提取鸡蛋图像参数(H、I、S)，再把鸡蛋打破，测量并计算出鸡蛋的真实新鲜度(哈夫值)，然后把哈夫值与试验提取的鸡蛋信息参数(H、I、S)对应起来，并作为样本数据。用MATLAB为工具，创建BP神经网络模型，该网络结构为：输入层3个神经元；隐含层20个神经元，传递函数为tansig；输出层3个神经元，传递函数为purelin。在建立鸡蛋新鲜度检测的BP神经网络模式基础上，本文设计出自动检测系统，对于外来鸡蛋的颜色数据，在网络初始化后，可以立即显示出判别结果。 用基于MATLAB的BP神经网络技术检测鸡蛋新鲜度，方法高效可行；鸡蛋新鲜度(哈夫值)与其内容物颜色信息(H、I、S)有关，以二者为样本数据建立的BP神经网络，具有很好的泛化功能和鲁棒性，经检验，网络识别正确率分别为：浅色白壳鸡蛋—88.7％，深色褐壳鸡蛋—90.8％，混合蛋—88.57％；从试验结果看，深色鸡蛋检验的准确率最高，浅色蛋次之，混合蛋稍低；此方法还可用于蛋品的其它内部品质检测上。更多还原

【Abstract】 Poultry egg is very popular with consumers because of its high nutrition. The inside quality of egg influences edible quality directly. In circulating of sale and processing, if it can be graded and priced depends on the quality according to fresh degree and the weight, it may not only protect consumers’ benefit but also help the managers and operators to adopt scientific management, to guarantee the quality of egg. So how to high-efficiently detect the fresh degree of egg, is an underlying subject.The research principle to establish the model of egg fresh degree is: According to the optics characteristic of egg, the longer the store time, the transmission rate of the fresh egg is lower. When the transmission rate changes, it indicates that the inside quality of the egg has changed, which can reflect with the change of color information within the egg.The relevant relation is established between egg fresh degree (Haff value) and the egg information parameter with machine vision and neural network technology in this research. The model of egg fresh degree is set up according to the course: Draw the color information (H, I, S) of the egg on the experimental rig first. The testing equipment is made up of light source, light room, video capture card, CCD camera and the computer. Light source and light room offer the environment of testing the egg image; The CCD camera takes a photograph of the egg, then the video capture card sends it into the computer after digitization the picture; After analysing the egg picture with the procedure that was worked out, the egg picture parameter (H, I, S ) was drawn. And then the egg was broken, the true fresh degree of the egg (Haff value) was measured and calculated. The Haff value and the egg information parameter (H, I, S) were corresponded as the sample data. The BP neural network model was established by Using MATLAB, whose structure is: Input layer has 3 neurons; Imply layer has 20 neurons, shift function is tansig; Export layer has 3 neurons, shift function is purelin. On the basic of setting up BP neural network model of detecting the egg fresh degree, the automatic detection system was designed in this article, which can immediately show the differentiating result according to the egg’s color data after the network initialize.Detecting the egg fresh degree with BP neural network technology based on MATLAB, the method is high-efficient and feasible; The egg fresh degree (Haff value) has something to do with the color information(H , I , S ) within the egg. BP neural network is established according to the two sample data, which has wonderful function and robust property. Through experiment, the correct discerning rate with network is:white shell egg of light color -88.7%, brown shell egg of dark color -90.8%, mix eggs -88.57%; From the test result, it shows that the correct discerning rate of the dark egg is highest, the light-colored egg takes second place, the mix eggs is slightly lower. The method can be used to detect the other inside quality of egg product.更多还原