基于SVM的鸡蛋外观品质检测方法研究-机械设计及理论专业论文.docxVIP

摘 要 随着人们生活水平的提高，对鸡蛋外观品质的要求也越来越高。论文改进了实 验装置并采用机器视觉技术对鸡蛋重量、蛋形以及蛋壳颜色分别进行了研究，具体 内容如下： 改进了实验装置，选择环形日光灯作为光源，通过调节摄像头的焦距、载物台 的高度以及Baslar图像采集软件的相关参数，获得了清晰的鸡蛋图片。对实验装置 做了标定，标定精度在0．1mm之内，可以满足鸡蛋外观品质检测的要求。 对采集到的鸡蛋图片进行预处理，包括颜色空间的转换、灰度变换、去除噪声 和图像二值化。提出在HSV颜色空间下，提取鸡蛋图像饱和度分量S，利用Ostu算 法和数学形态学方法获得鸡蛋的二值化图像。 研究了鸡蛋蛋形、重量和蛋壳颜色的特征参数，利用支持向量机算法建立了各 自的分类模型，并与标准BP和GA-BP算法做了对比实验。结果表明，支持向量机算 法具有较好的训练准确率和测试准确率，对蛋形、重量和蛋壳颜色的检测准确率分 别为100％、98．21％和97．14％。 关键词：鸡蛋品质；支持向量机；颜色矩；图像二值化；BP神经网络；遗传算法 Research on Inspection Methods of Egg Exterior Quality Based on SVM Abstract Along with the raising of people living standard，the requirement of egg exterior quality is higher than before．Paper have improved experiment device and egg weight， egg shape as well as eggshell color are studied respectively using machine vision technology．Contents specifically are following： Experiment device is improved．The annular fluorescent lamp is selected as light source．Through adjusting the related parameter of Baslar image gather software as well as the altitude of object table and the focal length of camera，vivid egg images are gotten． Camera calibration is done for experiment device，and its accuracy is less than 0．1 mm which can satisfy the requirement of egg exterior quality inspection． The captured egg images arc pre—processed with the method of conversion of color space，gray transforill，declining noise and binarization．A new method is put forward for binarization．Firstly,the saturation weight of egg image is distilled in HSV color space, and then Ostu algorithm and mathematic morphologic method are used for obtaining binarization image． The feature parameters of egg shape，weight and eggshell color are studied． Classification models are established by using support vector machine algorithm，and standard BP and GA—BP algorithm are used to do contrast experiment． Results demonstrate that the train accuracy and test accuracy of support vector machine algorithm is better．For the detection of egg shape，weight and eggshell colo