Effect Of Different Illumination On Characteristic Parameters And Identification Accuracy Of Coal And Gangue

The identification technology of coal and gangue based on image processing was studied systematically in this thesis.The important position of coal mining in our country and the research significance of coal gangue identification for improving intelligent coal mining were introduced.Traditional manual sorting has disadvantages such as health hazards and environmental pollution.In order to realize intelligent and unmanned mining.The characteristic parameters and recognition accuracy of coal and gangue with different surface characteristics under different illuminance using image recognition methods are studied in this thesis.Aiming at the complex environment of actual mining and sorting,a coal gangue image collection platform with different light intensities was built in the laboratory,and coal and gangue patterns with different surface characteristics were collected when the light intensity was 1150,2220,3120,4270,and 5260 Lux to establish a coal gangue image data set.In view of the interference in the coal gangue image acquisition platform and the actual environment,the collected images were grayed and filtered.The minimum mean square error and peak signal-to-noise ratio were used to evaluate the image filtering effect,and Gaussian filtering was selected for image removal.In order to explore the effect of light intensity on the characteristic parameters of coal and gangue,the gray-level histogram and gray-level co-occurrence matrix were used to extract multiple gray-scale and texture features of coal and gangue,and the extracted characteristic parameters were statistically analyzed.It is found that the difference in mean value,variance,energy,entropy and skewness of surface dry coal and gangue is large,and has a significant degree of discrimination.The gray-scale kurtosis difference value is small,and the degree of discrimination is not obvious.The texture contrast,angle second-order moment,and inverse difference moment of coal and gangue have large difference values and obvious discrimination,but the difference value of texture correlation is small.The gray-scale variance,energy and entropy of coal and gangue with wet surface have large differences.The gray-scale kurtosis of the two have a certain difference,while the gray-scale mean and skewness have small differences.The texture contrast,the inverse difference moment and the angular second moment are quite different,but the texture correlation is relatively close.Set up a support vector machine as a coal gangue recognition model,select various characteristic parameters and multi-gray and texture features as input parameters,and find the recognition rate of coal gangue with different surface characteristics under different illuminances.By comparing and analyzing the classification accuracy of coal gangue under different illuminances,it can be obtained that when the light intensity is5260 Lux,the recognition rate of coal gangue is higher.At this time,the coal gangue recognition rate based on multi-gray feature fusion is 94%,and the coal gangue recognition rate based on multi-texture feature fusion is 85%.When the light intensity is5260 Lux,the recognition rate of coal and gangue with wet surface is 79% based on multigray feature fusion,and that of gangue based on multi-texture feature fusion is 60%.