基于支持向量机的煤岩图像特征抽取与分类识别

为了尽可能减少作业人员数目，研究了煤岩图像的自动识别技术，介绍了煤岩图像的识别基础、小波变换和支持向量机原理，分析了煤岩图像纹理在多尺度分解情况下的特点以及支持向量机的参数设置，利用煤岩图像基于灰度共生矩阵的纹理统计量角二阶矩、对比度、相关性、均值、方差构造纹理特征子向量P1，利用煤岩图像不同尺度分解下的角二阶矩、对比度、相关、均值、方差构造纹理特征子向量P2，利用不同尺度分解系数构造纹理特征子向量P3，结合3个特征子向量构造纹理特征向量，最后结合支持向量机对煤岩图像进行分类识别。对不同的特征抽取方式以及煤岩的不同分类进行了比较分析。结果表明：该特征抽取以及分类方法能有效的表达纹理信息，对煤岩的识别准确率达到了97.959 2%，与不使用小波的方法相比提高了7.01%。研究结果可为煤岩界面的自动识别提供依据。

Coal-rock imagefeature extraction and recognition based on support vector machine

In order to reduce the number of workers，the automatic identification technique of coal and rock image was researched.The basis of coal rock image recognition and the principle of wavelet transform and support vector machine were proposed.Coal and rock image texture feature in multiscale decomposition and support vector machine’s parameter settings were analysed.Texture vector P1 were structured by angular second moment，contrast，relevance，mean value and variance.Texture vector P2 were structured by angular second moment，contrast，relevance，mean value and variance after coal and rock image were decomposed.Different scale decomposition coefficient was used to construct texture features vector P3.Three characteristic vectors were used to construct texture vector.Finally，classification and identification were carried by support vector machine.Comparison and analysis were made by feature extraction in different ways and coal-rock different classification.The results show that the feature extraction and classification methods can effectively express texture information and the coal rock recognition accuracy achieves 97.959 2%.The testing accuracy is increased by 7.01% compared with not using wavelet.Research results could provide the basis for the coal rock interface automatic recognition.