节点文献

基于图像的煤岩识别方法研究

Study of Coal Rock Recognition Methods Based on Image Processing

【作者】 佘杰

【导师】 孙继平;

【作者基本信息】 中国矿业大学(北京) , 通信与信息系统, 2014, 博士

【摘要】 本文对基于图像的煤岩识别技术进行了系统研究。阐述了影响煤岩图像采集质量的因素、噪声特性分类以及模型,提出将小波和小波包应用于煤岩图像去噪,为对煤岩图像去噪效果进行评估定义了保真度的概念。分析了常见的图像特征抽取方法,研究了煤岩图像在小波分解之后的特点和煤岩网状图的特点,提出将煤岩图像进行多尺度分解并结合灰度共生矩阵的方式进行特征抽取。研究了距离判别法原理,提出了均值纹理导向度和方差纹理导向度。详细研究了支持向量机的基本原理,提出了一种基于支持向量机的煤岩图像分类识别方法和系统以及采集装置。详细研究了BP神经网络和小波神经网络原理,改进了BP神经网络的最小均方误差函数和小波神经网络的结构。在煤岩图像特征抽取的基础上,提出了基于Minkowski距离判别、支持向量机、BP神经网络、改进的BP神经网络和改进的小波神经网络的五种模式分类方法,进行了煤岩识别仿真实验,并针对实验结果进行了比较和分析。

【Abstract】 The key technology of coal rock recognition based on image was systematically studied in thisthesis. Firstly, the influence factors of coal-rock image acquisition, noise characteristicsclassification and model were introduced. This paper proposes that the wavelet and wavelet packetare used in the coal rock image denoising. The fidelity concept is defined in order to evaluate coalrock image denoising effect. The common image feature extraction method is analysised. Thecharacteristics of coal rock image in wavelet decomposition characteristics and coal rock meshfigure are studied. The approach of the coal rock image multiscale decomposition which combineswith gray level co-occurrence matrix is used for feature extraction. The distance discriminantmethod is studied. The mean texture orientation texture orientation and variance textureorientation texture orientation are proposed. The basic principle of support vector machine isdiseussed emphatically and the method and system of coal-rock image classification andrecognition based on support vector machine are proposed. Collecting device is proposed.Thetheory of BP neural network and wavelet neural network are diseussed emphatically and theminimum mean square error function of BP neural network and structure of wavelet neuralnetwork are improved.On the basis of coal-rock image feature extraction, the classifiers based onMinkowski distanee, support vector machine, BP neural network, improved BP neural networkand wavelet neural network are proposed. The simulation experiments of coal-rock imagerecognition were condueted and the test results are compared and analyzed.

节点文献中: