Image Sharpening Of Zinc Flotation Froth Based On Blurry Feature Classification

Froth flotation is one of the moset widely used methods in the flotation of non-ferrous metal ores.Its main process is that bubbles rise in the slurry with selective but not tightly bonded ore particles at the gasliquid interface and then scrape away the froth formed on the slurry surface.As an important parameter of flotation performance,froth characteristics can directly reflect the real-time trend of flotation conditions.The traditional operation of froth flotation relies on artificial observation of the froth states to adjust.This operation mode has strong subjectivity,large measurement error and low efficiency,and cannot objectively evaluate and quantify the real state and working condition of flotation froth.At present,the froth state can be accurately measured by digital image processing technology.However,due to the poor flotation environment,uneven illumination in the flotation plant,excessive dust and fog at low temperature,the froth images collected in real time usually contain a lot of noise,and some of the images are blurry.Aiming at the problem that it is difficult to obtain high quality froth image samples in flotation plant,this doctoral thesis analyzes the causes of noise and blur for froth image,identify the blurry type of the froth by the multiple blur features,design a variety of imge sharpening algorithm based on blur feature classification,and successfully applied to the sharpening system in the mineral flotation froth.The main research work and innovation of this doctoral thesis are as follows:(1)In view of the fact that some froth image samples collected were blurry to a certain extent in the flotation plant,the clear and blurry characteristics of froth images were analyzed in detail,the clear froth image library and blurry image library were established,and an image screening system was designed to screen out abnormal froth images and mixed blurry images.On this basis,an automatic image classification method based on multi-blur feature analysis and support vector machine(SVM)is proposed.The method has high classification accuracy,which is suitable for the real-time demand of flotation industry.(2)Aiming at the problem that the flotation froth image is blurred by the influence of noise,this doctoral thesis proposes a spatio-temporal joint denoising algorithm for flotation froth images based on the bubble stability analysis.In this method,the current froth image frame is denoised by timedomain or spatial-domain filtering according to the stability state of the reference image-block,and then the time-space joint denoising of froth sub-blocks is obtained by combining the two filtering results by weighted average.Experimental results show that this method has good noise reduction effect on froth video images without background,and can restore clear texture features of froth images,which provides a clear image samples for the accurate extraction of froth visual features.(3)Aiming at the motion blur problem caused by camera shake or fast irregular motion of the froth,a spatiotemporal dual-stream information fusion deblurring algorithm is proposed.Aiming at the complex motionblurred froth image with variable space,the output of 2D and 3D convolution flow is calculated by using mask in the generator model,and the final deblurring result is obtained.Experiments on zinc flotation froth image data set show that this method has a good deblurring effect,the edge of the restored froth image is clear,and it can adapt to different conditions of foam image deblurring.(4)Aiming at the phenomenon of fogging and blurring in froth images collected in low temperature environment,a real-time defogging algorithm based on linear transformation of the model and image enhancement is proposed.In this method,the initial defogging image is obtained by improving the traditional atmospheric scattering model and the estimated value of atmospheric light,and the edges of initial defogging image are enhanced by variation-fractional differential operator.Experiments show that the algorithm has good defogging performance for this kind of industrial image,and the algorithm has high efficiency,which can meet the application requirements of real-time flotation monitoring.(5)In this doctoral thesis,the zinc flotation process of a large leadzinc mine in Guangdong is taken as the application object.Based on the original flotation froth image monitoring system,the image pretreatment function is added to realize the real-time collection of zinc flotation froth image,and the collected image data is screened,classified and clarified.In this system,it is verified that the pretreated froth image can provide more accurate flotation condition information,improve the flotation production efficiency,and lay a foundation for the optimal control of flotation.