Bauxite Selected Bubble Image Texture Feature Extraction Methods

Vision characteristics of flotation forth are closely related with the production efficiency in the bauxite flotation production. The cleaning forth has characteristics as high mineralization degree, seriously collapse and without apparent shape relative to the roughing and the scavenging. Thus texture is the key feature for characterize forth state. At present, the control of operating variables in the bauxite flotation production is often based on the texture of the froth surface state. However, due to the limitation of the artificial observation, it is hard to guarantee the optimizing flotation process. Based on the exploring the features of flotation forth, the article discusses the texture feature extraction methods of the bauxite cleaning froth images to quantify forth texture. It has important meaning for guiding the flotation production.First of all, based on the analysis the features of cleaning images and geometric structure such as flotation froth edge cannot be neglected in analysis of forth texture, a strategy based on the contourlet transformation domain is presented for image enhancement. This method modifies the contourlet of image to enhance boundaries and remove the noise of the bubble images, which provided high quality bubble images to characteristics extraction, compared to many other image enhancement methods.Then, to extract texture characterization of bubble image, the improved local binary pattern has been put forward. Binary function is improved by introducing the local contrast and variance is used as the adaptive weight of histogram, which overcomes light-sensitive effectively. And then the texture features are used for classification recognition froth status. The results prove that the characteristics can describe different froth status and it has a very good classification effect.Finally, in view of the fact that roughness is used to describe the key features of forth surface state, this paper studies a new measure method based on the improved local binary pattern. The relativity is analyzed between the extracted texture and the concentrate of mineral to obtain optimum feature interval. It will provide guides for improving the operation and control of flotation.