Improved watershed segmentation method for flotation froth image based on parameter measurement
Li Jianqi,Yang Chunhua,Cao Binfang,Zhu Hongqiu,Liu JinPing
DOI: https://doi.org/10.3969/j.issn.0254-3087.2013.06.005
2013-01-01
Abstract: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 paper,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 blurry 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 morphology 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%.