Research On Recognition Technology Of Coal And Gangue Based On Image Processing

In China, coal is a significant energy resource. There are gangues inevitably in mining. Gangue seriously affects the quality of coal, so the separation of coal and gangue improves the utilization efficiency of coal. Generally speaking, the methods of coal preparation are hand cleaning and machine cleaning. With the rapid development of machine vision technology, people begin to use images to distinguish coal from gangue. Thus, a recognition method which is based on image processing has been studied.Above all, the recognition scheme is discussed. Density is a property of substance, so the result is selecting density as the recognition characteristic. A recognition scheme has been determined. First, it calculates the volume by using the images. Second, it weighs the quality with the help of weighing device. Third, it calculates the density. Finally, it identifies coal and gangue according to their different densities. Besides, a data acquisition scheme has been studied and the required devices have been selected as well.Then the calibration of the recognition system for coal and gangue has been studied. At the same time, the camera system has been calibrated in laboratory conditions. The conversion relationship between pixel width and actual physical length has been got.According to the characteristics of the images, a set of image processing algorithm has been developed. First, it changes brightness and contrast by homomorphic filtering. Second, it conducts image smoothing. By comparing several filtering effects, Gauss filter is adopted. Third, it conducts image segmentation to separate contours from background. And then it optimizes the extracted contours with the help of mathematical morphology.According to the acquired contours of coal and gangue, a cumulative integral method is adopted to calculate the volume. Then the density can be calculated combined with the quality. Finally it identifies the object in use of the established recognition model. Detection has been performed on samples by taking advantage of the experimental device. And the designed recognition method is proved to be correct. The results show that this method is able to rarrv nn real-time detertinn nn rnal and gangue ranirly and effertively...