【作者】 赵娇洁；

【导师】 杨姝；

【作者基本信息】 沈阳师范大学 ， 计算机应用技术， 2011， 硕士

【摘要】 汽车轮毂自动识别系统主要是研究了一种机器视觉系统,这种系统能够识别流水生产线上不同款式、不同型号的多种汽车轮毂。这种识别系统的优势在于：非接触性、在线实时性、高精度性、快速性以及较高的抗干扰性等特点,能够按照实际需求实现铸件产品的准确分类,符合当前快速发展的经济发展速度和产业需求,其应用前景开阔、可观。本文的研究思路是将机器视觉技术和图像处理技术进行有机结合,以实现不同种类、不同形状的汽车轮毂分类识别为研究对象,在利用CCD传感器和图像处理卡等硬件获得所需图像的基础上,对轮毂图像先进行图像预处理,再到图像分割和特征提取,最后进行匹配分类,实现识别。主要的研究内容包括：基于matlab将真彩图像转换成8位灰度图像,对于图像中存在的噪声(主要是轮毂内侧的毛刺),根据它的特点,采用中值滤波对图像进行去噪处理,达到了较好的效果。在此基础上,采用全局阈值法对已去噪图像进行分割。结合降噪、分割后图像特点,采用二值形态学将其边缘的毛刺去掉,进一步平滑图像,进行边缘提取。综合运用选点拟合法、二值形态学等处理方法,提取了轮毂图像的9个特征：外圆半径、图像的中心孔、轮毂图像周边孔洞数、中心孔的面积与轮毂面积比、轮毂孔的面积与轮毂面积比、四个转动惯量等理想特征,为后续的图像识别奠定了坚实的基础。在此基础上,对以上相关步骤涉及到的算法进行简单的验证。轮毂形状自动识别技术能够很好的适应当前经济高速发展时期大批量的生产和快速的在线识别分类的实际需要,真正克服了传统手动分类的弊端,有效地降低误判现象,开展了面向形状识别的机器视觉系统的研究工作,为实现工业生产现代新型的识别技术和手段作一些有益的尝试和探讨。该识别方法可以实现多种轮毂在传送带上随机混流的状态下的自动识别分类,有效识别正确率在90%以上。更多还原

【Abstract】 Auto-hub automatic recognition system is mainly a kind of machine vision system, which can recognize different types of various auto-hubs on the assembly lines. It has many advantages, such as non-contact, online real-time, high precision, fast speed and higher anti-interference. It can classify the casting products accurately in terms of actual demands. The system fits the demand of rapid development of economic growing and industrial development. It has the prosperous application future.In the research, we use machine vision technology and image processing technology to classify different types and different shapes auto-hubs. Basing on the capture of auto-hubs images using CCD sensors and image processing cards, we preprocess auto-hub images firstly, then segment images and extract features, and finally match the classification to realize the recognition process. The main research contents consist of a few parts:converting true-color images into eight gray images based on matlab in the preprocessing stage, dealing with the noises which mainly exist in the burr of auto-hub inside using median filter. Besides these, we use the method of global threshold to segment the denoised images, combined with the characteristics of noise reduction and segmentation image, use the characteristics of the binary morphology to eject the noise of the burr, smooth the image further and then extract the edge of image. Comprehensive using fitting method、binary morphology, nine features are extracted:the circle radius, the center of the image of the holes, central hole area, wheel image aperture count, wheel hub hole inertia area and four characteristics of ideal images. In the research, we also verify the algorithm with the more relevant steps involved.Auto-hub automatic recognition system can well meet the actual needs of the current rapid economic development period of mass production and fast speed for on-line recognition and classification. It really overcomes the disadvantages of the traditional manual classification, effectively reduces misjudgment phenomenon. The research work for shape recognition of machine vision systems can realize the automation of some industrial production manufactured. The recognition method can achieve a variety of hub in the conveyor belt random mixed flow under the condition of the automatic recognition and classification, effective recognition accuracy in 90%.更多还原