面向参数测量的改进分水岭浮选泡沫图像分割方法  

Improved watershed segmentation method for flotation froth image based on parameter measurement

作　　者：李建奇[1,2] 阳春华[1] 曹斌芳[1] 朱红求[1] 刘金平[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083 [2]湖南文理学院电气与信息工程学院,常德415000

出　　处：《仪器仪表学报》2013年第6期1233-1240,共8页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金重点项目(61134006;61071176);国家科技支撑计划项目(2012BAF03B05);湖南省自然科学基金(11JJ6062)资助项目

摘　　要：矿物浮选过程中泡沫参数与工况密切相关,针对泡沫中特殊区域(透明窗、黑洞和泡沫间光亮带)的存在而导致参数难以测量的问题,提出了一种基于局部纹理特征和自适应形态学的改进分水岭分割方法。首先针对现场光照不均,采用自适应多尺度Retinex算法进行图像增强。然后使用快速自适应阈值法提取高亮区和暗区,并结合先验空间关系实现透明窗标定;基于LBPV纹理算子实现暗区的区域生长以标定黑洞和识别泡沫间光亮带,进而完成对特殊区域的处理并获得透明窗和黑洞指数。最后采用自适应形态学获得多尺度梯度图像和标记点,使用改进的分水岭算法完成泡沫图像的分割。结果表明,该方法能够更准确的处理含特殊区域的粘连不均匀泡沫图像,对不同工况下泡沫参数测量有更好的鲁棒性,浮选优良工况率提高了5%。Froth image parameters in industry field are closely correlated with production status in mineral flotation process. Aiming at the problem that in froth image there exist some special regions （ such as transparent windows, black holes and long, narrow bands of non-peaks） , which leads to the difficulty in parameter measurement, in this pa- per,based on local texture characteristic and adaptive morphology, an improved watershed algorithm is proposed. Firstly, aiming at the uneven field illumination, adaptive multi-scale Retinex algorithm is applied to enhance the blur- ry froth image. Secondly, fast adaptive threshold algorithm is adopted to extract the bright and dark regions, and the priori space relation is used to identify the long and narrow band;meanwhile on the basis of LBPV texture operator, regional growing is realized to recognise black holes and long, narrow band, and then complete the processing of the special regions and obtain the image characteristics of the black holes and long, narrow band. Thirdly, adaptive mor- phology method is used to obtain multi-scale grads images and label points, and improved watershed algorithm is used to complete froth image segmentation. Experimentation results show that the proposed method could more accurately process conglutinant and uneven froth image containing special regions, has better accuracy and robustness;and the flotation condition stable operation ratio is increased by 5%.

关 键 词：浮选 泡沫图像 参数测量 分水岭算法 

分 类 号：TP274.3[自动化与计算机技术—检测技术与自动化装置]