【作者】 胡明明；

【导师】 温强；

【作者基本信息】 哈尔滨工程大学 ， 光学工程， 2007， 硕士

【摘要】 由于色选机独特而有效筛选方式，它在大米等粮食筛选加工工业中得到了广泛的应用。目前我国的颜色分选机研制技术与国外产品相比还有差距，很多厂家需要靠进口昂贵的国外产品。因此作为一个农业大国，研制适合我国国情的颜色分选机，并广泛应用于粮食及产品加工多种行业，将会带来巨大的经济效益和社会效益。本文根据色选机的色选过程，利用PCI数据采集卡及彩色线阵CCD摄像头，搭建了一个图像数据采集实验装置。通过VB编写的控制界面操作该装置，实现对不同大米线阵图像数据的采集，并将采集的数据以文本文档的形式储存，作为后续大米色选算法研究的数据源。通过对每一帧图像数据的研究与分析，发现图像数据中的高频噪声对于提取大米异色粒的特征信息产生较大影响，为了减小或消除高频噪声的干扰，本文尝试用小波滤波的处理方法以得到平滑的信号。并根据数字图像处理中特征提取的相关理论，结合实际图像数据特征，本文采用k-均值聚类分析方法将图像数据分为三类：背景类、正常米类和异色米类。实现从图像数据中提取异色点的目的。与现有色选算法原理不同，该算法不需要人为操作或提供某种先验值，而是由聚类分析算法根据输入数据的内部特征，直接确定异色米所在的类。最后，本文利用对20组不同大米采集的2000多帧图像数据，进行算法验证，从整个实验的结果和分析可知该算法是实际有效的。更多还原

【Abstract】 Because of its unique and effective screening method, color sorting apparatus is widely used in sorting and processing industry of rice and other foodstuff. At present, our country lags behind some foreign countries in the technology of color sorting apparatus, most manufacturers have to rely on expensive imports of foreign products. Therefore, for china, a large agricultural country, it will bring great economic and social benefits to develop our own color sorting apparatus, and use it widely in food processing and other industries.Based on the sorting processing of the color sorting apparatus, an image data acquisition device is built by the use of PCI data acquisition card and the color line CCD camera. And by operating the device through a control interface programmed by VB, the image data of different rice array are acquired, and the sampled data are stored in text files form, as a data source for the study of color-based sorting algorithm followed by.By the research and analysis of each frame of the image data, we find that the high frequency noise in the image date has a greater impact on the feature extraction of the bad rice. In order to reduce or eliminate this disturbance, the paper attempts to get smooth signals by wavelet filtering method. And according to the relevant theory of digital image processing feature extraction and actual image data characteristics, the k-means clustering analysis is applied to sorting the image data into three categories: the background category, normal rice category and the bad rice category, to get the purpose of extracting the bad points off the image date.Different from the principle of existing color sorting algorithms, this algorithm does not rely on any artificial operation or transcendental value, but directly determines which category the bad rice particle should be in, by the clustering algorithm based on the internal features of the data.Finally, we use 3600 frame image data collected from 20 groups of different rice to test the algorithm. The analysis of entire experimental results shows that the algorithm is practical and effective.更多还原