Research On Recognition System Of Coal Flotation Froth Image

Flotation is a technology for carrying out the mineral separation,due to the difference in surface wettability of mineral grains,whose visual features in froth surface directly reflect work conditions of flotation.The judgment of flotation state is still based on manual observation in most domestic flotation mills currently,which produces great restriction on the promotion and development of flotation technology.The introduction of image recognition technology into flotation not only can help to optimize the operation of flotation operation,but also can liberate much labor power and material resources,which carries great importance to the subsequent development of flotation.The thesis takes the project of automatic controlling system for flotation reagent adding in Zhaolou coal mine as the background.The in-depth study on flotation froth identification system has been carried out,which effectively oversteps the inherent weakness of the traditional technological process of flotation.And it can be briefly summarized as the following aspects:Firstly,considering the situation that coal flotation froth images are vulnerable to mixed noise,especially Gaussian-salt and pepper mixed noise,A novel filtering algorithm is proposed,which not only outperforms the traditional algorithm in filtering mixed noise of froth images,but also meets certain real-time needs.Secondly,research on extraction algorithms of feature based on coal flotation froth images has been conducted,and texture feature,size feature and speed feature have been selected as the input of system.Among which,the deficiencies of traditional watershed algorithm have been analyzed and summarized,and a correspondingly improved algorithm has basically achieved precise segmentation of coal flotation froth images.Thirdly,through empirical formulas and related experiments,the optimal number on hidden layer units of BP neural network have been determined,whose performance has been compared with genetic BP neural network based on the same network parameters.And genetic BP neural network has been selected as identification algorithm of the system.Finally,out of consideration for working environment of flotation workshop and so on,reasonable selection of industrial camera is put forward.Subsequently,backend database and front-end interface has been designed according to system requirement.