Research On Static Parameters Of Flotation Foam Based On Machine Vision

In the research of flotation mechanism,such as particle and bubble collision theory,the morphological parameters of bubbles are used to conduct experimental research and verify the correctness of the theoretical model.The research of extraction method has important research value;In the concentrating mill,it still depends on the judgment of the surface visual state of flotation foam by technicians to adjust the flotation operating parameters.There are some disadvantages,such as strong subjectivity,lack of unified evaluation criteria and lag of adjustment of working condition parameters.Applying machine vision to the flotation process,according to the visual characteristic parameters of the foam realize monitoring of the flotation process.The static parameters such as foam shape,color and texture can be used as a priori knowledge to identify the state of flotation conditions and predict.In this paper,establishing a machine vision system to collect image information of flotation foam(bubbles),and study the static parameters.The main research work was as follows:According to the existence form of discrete bubbles in the slurry,a bubble observation system is established,which was composed of simulating flotation process equipment and machine vision equipment.A bubble edge extraction method based on region segmentation was proposed,and the edge points were extracted using Canny detection.For the bubble overlap,the CSS corner detection and the direction chain code were used to mark the concave point to divide the contour,and then using the least square method to reconstruct the bubble edge.The horizontal placement segmentation method is proposed to measure the volume and surface area of the bubble.Compared with the other three different methods,this method in this paper had higher precision and robustness.The foam image was captured on the flotation machine by a monitoring system which built with LED light source and industrial camera.Making the following improvements to the watershed algorithm based on marker points: homomorphic filtering was used to improve the image brightness unevenness and shadow problem;KFCM was used to cluster similar highlight points in the foam and extract marking points and edge strips;and making the watershed segmentation with the highlighted mark as “catchment basin”.Compared with the traditional watershed method,the number of foams after improved was relatively close to reality.The dual-domain denoising technique was applied to the foam image filtering to remove the noise while preserving the details in the image.An LBP texture extraction method based on the HSV color space model was proposed,three scales LBP features distribution histograms were extracted on the color component image.It had higher feature resolution and texture recognition.Through training to set up the SVM classifier and then classified and tested the images of two different flotation conditions.It had high accuracy and can be used to identify different working condition.Figure 44;Table 4;Reference 57.