The Research Of Coal Flotation Froth Image Processing Technology

Coal flotation which is under the effect of reagent is a complex physical and chemical process that involves interaction of gas-liquid-solid three phases. By bubbling air to coal slurry continuously in the process of coal flotation, hydrophobic coal particle adheres to the bubble and gradually emerges from bubble layer while hydrophilic gangue sinks in the coal slurry and discharges with the tailings. Finally the clean coal and gangue separate with each other. For many years, the reagent dose of coal flotation is mainly adjusted manually through observation the bubble’s state by the workers’s eyes. The backward automation level of the process and higher of worker labor-intensive hinders the recovery rate of clean coal and optimization of flotation. But bubble layer of coal flotation contains a large number of important information which is related to flotation index. So the recovery rate of clean coal which is got through flotation can be improved by describing visual features of coal flotation quality. Through research of features of coal flotation froth image, we can extract the image features parameters which are able to reflect the coal flotation state. Using these features parameters to reflect process state of coal flotation is very prospective and realistic significance to achieve automation of the coal flotation.This paper analyzes the coal flotation mechanism, and according to the principle that coal particles adheres to coal flotation froth analyze the relationship between the characteristics of Coal slime flotation and state of flotation. Based on the study, this paper preliminarily builds the image vision system of coal flotation froth. The main research work of this paper are listed below.(1)Firstly, we do analyze the coal flotation image’s features, according to the conditions that noise size in coal flotation froth image is different and the needs to avoid the change of coal flotation froth image when it is segmented, combining with the characteristic of morphological reconstruction and filtering area,area reconstruction oprn-closing filter with increasing area structure element is used to filter out the noise.(2)Contrast of foreground and background of coal slurry bubble image is very low. And the image can’t be recognized or produces the phenomenon under segmentation by using the method of watershed segmentation. To solve above deficiencies, this paper proposes the method of watershed segmentation based on morphological gradient. Marker points are extracted by using the method of area reconstruction H-dome transform. Then flotation froth image is segmented based on marker points. Segmentation results of bubble image with lower site foreground and background contrast is very good by using this method.(3)Extraction the histogram statistical features about the coal flotation froth images, by analyzing feature parameters reflecting flotation state,using gray level co-occurence matrix extracts four directions(0°、45°、90°、135°) of coal flotation froth image and analyze variation that eigenvalues show in stage of coal flotation froth image while the fact that histogram can’t reflect spatial pixels of coal flotation froth is considered. Through extracting bubble size of coal flotation froth image, loading rate of bubble, dynamic characteristics of bubble and particle size of coal flotation, this paper analyzes these characteristics with the flotation changes in the regulatory, which can used as a judge flotation process indicators.(4)Based on the coal flotation process, according to the environment and the requirements of flotation, build the hardware platform about coal flotation froth image machine vision,write software programs to achieve real-time image acquisition functions. According to the advantages and disadvantages of Matlab and VC in actual development process, mixed-language programming based on MATLAB and VC is used to collect and handle image. According to actual operational experience,summarize the attention that write the mixed programming,which can offer a reference for the subsequent real-time image monitoring.