Research On Recognition Technology Based On Reflection Spectrum Of Coal And Rock

Coal resources occupy a dominant position in my country’s energy use and consumption.With the development of the times,traditional artificial mining can no longer meet the needs of modern mining,and mechanized,intelligent,and unmanned mining has gradually become the trend of coal mining.Coal and rock identification is the core technology in the mining process.At present,most of the underground mining is still at the stage of manual identification,which does not meet the needs of intelligent mining.However,unmanned and intelligent coal and rock identification has always been an unsolved problem.One of the most successful directions for using remote sensing to detect mineral resources in hyperspectral remote sensing.Mineral resources have different spectral curves due to the differences in their internal components.The use of hyperspectral remote sensing features of multiple and narrow bands can effectively perform mineral resources.Classification and identification.This paper uses the characteristics of hyperspectral technology to analyze the coal and rock spectrum curves to obtain the difference mechanism,and design a coal and rock identification algorithm model based on this.The main work and research results are as follows:1.A total of 43 coal and rock samples of various types were collected in the three major mining areas in Huainan,and the solid coal and rock samples were crushed and sieved to prepare powder samples with different particle sizes.The powder sample adopts two methods of fixed collection and height-adjusted collection to measure the spectral curve.Select 15 representative typical coal and rock samples for industrial analysis and XRF element analysis,and measure the material content of the coal and rock samples.Based on the measured spectrum data and composition content data,the coal rock hyperspectral database is established,and the coal rock information database query system is designed using python GUI,which provides data support for the analysis and identification of coal rock spectrum and the application under actual working conditions.2.Based on the measured coal and rock spectrum data,perform preprocessing such as envelope removal on the spectrum curve to enhance the absorption characteristics of the spectrum curve.Analyze the relationship between the production of absorption valley characteristics and the content of material components.Among them,the absorption valley of the coal sample before the 1000nm band is caused by the energy level transition caused by the electron absorption of the metal cation.The absorption valley after the 1000nm band is related to the fundamental frequency,frequency multiplication and combined frequency produced by the bending and stretching vibration of the H2O,-OH,and CO32-groups.The coal absorption characteristics are parameterized and the composition content is analyzed.The results show that the content of H2O molecule,AL2O3 and Fe2O3 and the yield of volatiles have a strong,medium and weak linear correlation with the depth of coal absorption valley.The characteristics of the absorption valley of coal-measure rocks are more obvious than that of coal samples.It is affected by the electron absorption of transition metal cations such as Fe2+ and Fe3+ at the visible light-short-wave near-infrared band to produce energy level transitions,and the absorption at the long-wave near-infrared band The valley feature is affected by the fundamental frequency,frequency multiplication and combined frequency produced by the bending and stretching vibration of the H2O,-OH,and CO32-groups.Coal and rock have differences in two angles,the overall waveform and the characteristics of the absorption valley.From the perspective of the waveform,the overall reflectivity of coal-measure rocks is generally higher than that of coal,and has the characteristics of convex waveform;the overall waveform of coal is relatively gentle,showing a slow upward trend.From the perspective of the characteristics of absorption valleys,the characteristics of coal-measure rocks are more obvious,especially in the four wavebands of 1400nm,1900nm,2209nm and 2300nm in the mid-and long-wave near infrared.3.Four representative coals are selected as the research object,the detection distance and coal particle size are used as independent variables,and 9 detection distances and 5 sample particle sizes are designed under the condition of fixed light source incident angle and detection angle.To study the influence of detection conditions on the reflectance spectrum of coal and rock.The experimental results show that the reflectivity of coal and rock samples decreases as a quadratic polynomial function with the increase of the detection distance.The reflectivity and distance have a strong correlation,and anomalies appear in some specific wavebands.With the increase of coal particle size,the overall reflectivity of the coal spectrum curve shows a decreasing trend.The reflectivity of the coal sample at the characteristic points between 0.25mm-0.5mm and 0.15mm-0.25mm has a "diving" drop.The reflectivity of the rock samples at characteristic points between 0.15mm-0.25mm and lmm-2mm also showed a "diving"drop.According to the experimental results,the detection distance of 10 cm and the particle size of 1 mm sample are selected as the external detection conditions to construct the coal rock recognition algorithm.4.The 43 samples are divided into the training set and the test set at a ratio of 0.7,and the algorithm models for coal and rock recognition are designed from the three perspectives of characteristic band extraction,full-band reflectivity characteristics and spectral curve morphological characteristics,and the accuracy,The Kappa coefficient and F1-Score three standard evaluation models are good and bad.Among them,the recognition rate of the SPA-SVM model based on the feature band extraction is 84.6%on the test set,the Kappa coefficient is 0.69,and the F1-Score is 0.83;the recognition rate of the 1D-CNN model based on the full-band reflectance feature is 100 on the test set.%,Kappa coefficient and F1-Score are both 1.The recognition rate of SCA-SID model on the test set based on spectral morphological characteristics is 92.3%,Kappa coefficient is 0.85,and F1-Score is 0.92.Comparing the three coal and rock recognition algorithm models,the SCA-SID model is more practical and more suitable for applications under actual working conditions in terms of model complexity and ease of operation.Figure[46]Table[11]Reference[80]...