基于视觉的鸡蛋缺陷检测

发布时间：2018-05-25 20:16

本文选题：蛋形曲线 + RANSAC拟合曲线　；参考：《广东工业大学》2017年硕士论文

【摘要】：随着我国养殖业的迅速发展,鸡蛋产量和销量占据国际市场的大量份额,鸡蛋品质的检测显得十分重要。鸡蛋表面有很多的微生物和细菌,在鸡蛋运输过程中,会因磕碰、挤压等外在因素而造成外壳破损,微生物和细菌很容易污染鸡蛋内的营养物质,还会感染周围完好的鸡蛋,不仅给厂家带来巨大的经济损失,还给消费者带来健康隐患。蛋类加工业对鸡蛋的缺陷检测研究刻不容缓。基于视觉技术的检测技术因为具有高精度、高效率、无损坏等优点在蛋类加工业中受到越来越多的关注。形状缺陷检测常用基于曲线拟合的方法,通常用于拟合直线、圆、椭圆等,鲜有对鸡蛋外形的拟合研究,原因在于蛋形曲线的准确描述难度较高。常见的基于Hough变换和基于最小二乘法的曲线拟合受噪声影响较大,拟合精度和运行效率低。传统的裂纹缺陷检测常用基于纹理特征的算法和基于形态特征的算法,两者都存在如何最大程度保存裂纹特征的问题。因此,研究鸡蛋缺陷检测具有重要的现实意义。本文搭建基于视觉的鸡蛋缺陷检测系统,比较各种光源的参数确定激光照明器,比较相机性能确定CCD相机采集图片。通过确定蛋形曲线方程以及拟合蛋形曲线条件,采用单模和多模RANSAC算法对鸡蛋形状进行拟合,检测鸡蛋是否存在外形缺陷。通过对预处理后的裂纹图像进行高频增强滤波,然后纹理特征的局部最大差值法和形态学处理相结合,提取裂纹特征,检测鸡蛋是否存在裂纹缺陷。根据实验结果可知,本文提出的鸡蛋外形缺陷检测算法具有鲁棒性,准确率为95.4%,召回率为96.1%;本文提出的鸡蛋裂纹缺陷检测算法的准确率为96.8%,召回率为97.5%。
[Abstract]:With the rapid development of China's aquaculture industry, egg production and sales occupy a large share of the international market, the detection of egg quality is very important. There are a lot of microbes and bacteria on the surface of eggs. During the transportation of eggs, the shell will be damaged by bumping, squeezing and other external factors. Microorganisms and bacteria will easily contaminate the nutrients in the eggs, and they will also infect the eggs in good condition around them. Not only bring huge economic losses to manufacturers, but also bring health risks to consumers. It is urgent to study the defect detection of eggs in egg processing industry. Because of its advantages of high precision, high efficiency and no damage, visual detection technology has attracted more and more attention in egg processing industry. Shape defect detection is usually based on curve fitting, which is usually used to fit straight line, circle, ellipse and so on. There is little research on egg shape fitting, because it is difficult to accurately describe egg shape curve. The common curve fitting based on Hough transform and least square method is greatly affected by noise, and the fitting accuracy and running efficiency are low. The traditional methods of crack defect detection are based on texture feature and shape feature. Both of them have the problem of how to preserve the crack feature to the maximum extent. Therefore, the study of egg defect detection has important practical significance. In this paper, an egg defect detection system based on vision is set up. The parameters of various light sources are compared to determine the laser illuminator, and the performance of the camera is compared to determine the CCD camera to collect pictures. By determining the egg shape curve equation and fitting the egg shape curve condition, single mode and multi mode RANSAC algorithm were used to fit the egg shape to detect whether the egg had any shape defect. After preprocessing the crack image is filtered by high frequency enhancement, then the local maximum difference method of texture feature is combined with morphological processing to extract the crack feature and detect whether the egg has crack defect or not. According to the experimental results, the algorithm proposed in this paper is robust, the accuracy is 95.4and the recall rate is 96.1.The accuracy of the algorithm is 96.8and the recall rate is 97.5.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S879.3;TP391.41

【参考文献】