摘要
利用小波多尺度分析提取的铜浮选泡沫图像传统统计特征,具有多尺度特性,但没有形态学意义,无法直接反映泡沫图像的表观特点。对泡沫灰度图像进行二值化,能得到泡沫尺寸特征,但不具有多尺度特性。针对以上问题,提出一种基于小波多尺度二值化的泡沫图像特征提取方法。首先在对泡沫灰度图像进行小波变换的基础上,分别对不同尺度的小波逼近子图进行二值化;然后根据二维小波变换的空间-频率关系,对各二值图像的白色区域面积进行统计计算,得到一种新的与泡沫形态直接相关的特征——等效尺寸特征;最后,根据获得的等效尺寸特征,利用支持向量机对泡沫图像进行分类。实验结果表明,所提取的特征具有很好的可分性,可以很好地对铜浮选现场泡沫工况进行识别。
Traditional statistical features of copper flotation froth image extracted with wavelet multi-scale analysis have multi-scale property,but fail to reflect the froth morphology characteristics directly.Bubble size features are easy to be obtained through froth gray image binarization,yet lacking of multi-scale property.Aiming at the above problems,a novel method is proposed for the feature extraction of froth image based on wavelet multi-scale binarization.Firstly,the froth gray image is decomposed to approximation subimages and detail subimages with wavelet transform.The approximation subimages are binarized at different scales.Then the total area of the white regions of each binary image is calculated according to the space-frequency relationship of two-dimensional wavelet transform,thus a new feature named equivalent size feature is obtained,which is directly related to the froth morphology.At last,the obtained equivalent size feature is adopted for froth image classification with support vector machine （SVM）.Experiment results demonstrate that the classification has achieved favorable effect,so that the froth performance of copper flotation can be well recognized.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2014年第3期586-592,共7页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(61273169
61134006)
国家科技支撑计划(2012BAK09B04)资助项目
关键词
小波多尺度二值化
泡沫图像
等效尺寸特征
工况识别
wavelet multi-scale binarization
froth image
equivalent size feature
performance recognization
