基于Curvelet变换的低分辨率煤岩识别方法

Recognition method of low-resolution coal-rock images based on curvelet transform

针对小波变换仅能有效表达图像中的点奇异性,难以提取煤岩图像曲线特征的弱点,以及高分辨率煤岩图像计算量大,难以满足煤岩识别实时性要求的问题,提出了一种基于曲波变换的低分辨率煤岩识别方法。该方法通过曲波变换对煤岩图像进行曲波分解,得到各尺度层曲波系数,利用主分量分析进行降维,并将结果分别输入不同k-NN分类器中,对分类结果加权融合,实现煤岩图像的分类识别。实验表明:通过曲波分解提取的特征能够有效地表达煤岩图像的曲线特征,与现有方法相比较,所提出方法具有更高的识别率,平均识别率达95.0%,在煤岩图像分辨率较低情况下也可以获得很高的识别率,满足煤岩识别实时性的要求。

Considering the limitations of wavelet in image representation—that it is only optimal in representing point singularities and difficult to extract curve features of coal and rock images,a new recognition method for low-resolution coal-rock images based on curvelet transform was proposed. The method used curvelet transform to decompose images into curvelet coefficients in different scales. Then,PCA was applied to obtain a lower dimensional representation that was put into a k-NN classifier. Finally,the final recognition result was obtained via weighted fusion of classification results. Experimental results showed that the features extracted by curvelet decomposition could effectively express the curve features of coal-rock images. Compared with several other existing methods,the proposed method had higher recognition accuracy rate,with the average recognition rate reaching 95.0%. Under the condition of low image resolution can it also get high recognition rate and meet the real-time requirements of coal-rock recognition.