基于CNN-SVM的选煤厂浮选泡沫图像识别方法

The image identification method of flotation foam based on CNN-SVM at coal preparation plant

以选煤厂煤泥浮选泡沫为分类对象,提出一种CNN—SVM混合模型,对煤泥浮选泡沫图像进行分类识别。试验采取山东某选煤厂的20000张浮选图像制作数据集,根据灰分不同将图像分成8个类别,并针对图像的噪声特点,对其去除高斯椒盐噪声并做了增强预处理。通过模型试验,相对于单独的CNN和SVM模型来说,这种复合模型更加可靠准确。

Taking the slime flotation foam at the coal preparation plant as the classification object,put forward a kind of CNN—SVM mixed network model for classifying the image of slime flotation foam.Adopting 20000 pieces of flotation image production data set at Shangdong Coal Preparation Plant and classified as 8 categories of images according to different ash content and aiming at the noise characteristics of image,get rid of Gaussian salt and pepper noise and make strong pre-treatment.Through model test,the accuracy rate of mixed model prediction can be at 87.66%.For the CNN and SVM model,this composite model is more accurate and has more strong robustness.