搪瓷生产线在线统计图像识别算法研究与软件开发

发布时间：2019-01-26 20:21

【摘要】：图像识别技术是实现未来智能制造的关键技术之一,对于传统生产线改造也有广泛应用前景。本文以某搪瓷生产线产品在线统计系统为研究对象,对其所涉及的图像预处理、特征检测、图像识别等技术进行了研究,并利用Matlab的GUI工具包实现了一个简单的产品在线统计原型系统。本文所做的工作主要从以下几方面展开:首先,对图像预处理算法进行了研究,通过对直方图均衡化、均值滤波、分段线性变换、伽马变换等图像增强算法的处理效果进行对比分析,确定采用分段线性变换对单品生产情况下所采集的图像进行预处理。对于混品生产情况下产品的分类与计数,研究了基于RGB颜色空间和HSV颜色空间色彩提取方法,对比两种方法的提取效果,选择了基于HSV颜色空间的色彩提取方法提取混合产品图像中的目标区域。其次,研究了搪瓷产品特征检测方法,针对单品生产时所采集的图像,本文提出了一种快速椭圆检测方法,实现了对产品特征的快速识别。针对混品生产情况下图像识别所遇到的问题,研究了目标区域位置和尺寸检测、图像中相连区域分割、透视变换等方法,实现了对混品生产条件下的图像识别与产品分类统计。最后,完成了软硬件系统设计工作,开发了基于Matlab的产品在线统计图像识别原型系统,能够准确快速地完成对生产线上搪瓷产品的统计。
[Abstract]:Image recognition technology is one of the key technologies to realize intelligent manufacturing in the future. In this paper, an online statistical system of enamel production line is used as the research object, and the image preprocessing, feature detection, image recognition and other techniques involved in the system are studied. A simple prototype system of product online statistics is implemented by using Matlab's GUI toolkit. The main work of this paper is as follows: first, the image preprocessing algorithm is studied, through histogram equalization, mean filter, piecewise linear transformation, The processing effect of image enhancement algorithm such as Gamma transform is compared and analyzed. The piecewise linear transform is used to preprocess the image collected under the condition of single product production. The methods of color extraction based on RGB color space and HSV color space were studied for the classification and counting of products under the condition of mixed product production, and the results of the two methods were compared. The color extraction method based on HSV color space is selected to extract the target region from the mixed product image. Secondly, the feature detection method of enamel products is studied. A fast ellipse detection method is proposed for the images collected in the production of individual products, which realizes the rapid recognition of product features. Aiming at the problems of image recognition under the condition of mixing production, this paper studies the methods of location and size detection of target area, segmentation of connected area in image, perspective transformation and so on, and realizes image recognition and product classification statistics under the condition of mixing production. Finally, the software and hardware system is designed, and the prototype system of online statistical image recognition based on Matlab is developed, which can accurately and quickly finish the statistics of enamel products on the production line.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TQ173.6

【参考文献】