Research On Improved Algorithm Of Edge Detection Based On Fractal Feature Of Mineral Phase

Image edge contour feature is an important basis for image recognition.The key to improve the accuracy of image recognition is to extract clear edge contour characterization parameters accurately.As the essential raw materials for blast furnace charging,the microstructure of mineralogical phase determines pellet metallurgy properties.It is an urgent problem of pellet preparation process to establish the quantitative relationship between the microstructural characteristics of mineralogical phase and its metallurgical properties.This paper focuses on the fractal dimension and multifractal theory.Based on this,the edge features of mineral phase are detected.The superiority of the algorithm is tested by empirical comparison.The relationship between the multifractal spectral width and the key indicators(alkalinity)of the pellets metallurgical properties and surface uniformity is discussed.Firstly,mineral phases were obtained through experiments,and the RGB three-color weighting method was used to grayscale processing,and the adaptive histogram equalization was used to process the grayscale image to enhance the contrast effect of mineral phase,and lay the foundation for subsequent mineral processing.Secondly,in the DEBEFM,an improved algorithm with low time cost is designed through the analysis of existing mineral phase edge detection methods.By comparing the moving windows with the length of 5,and the algorithm judges whether the gray values of the pixels in the window are all 0.If they are all 0,the average value of the gray values difference under the three distances calculated in the window are 0.Otherwise,the average value of the gray values difference under different distances is calculated separately.This algorithm reduces the number of iterations of the algorithm and effectively reduces the execution time of the program.Compared with Canny operator and Log operator in traditional image edge detection method,the mineral phase edge extracted by the improved algorithm is clear and complete,and the algorithm is proved to be noise-resistant.Finally,the relationship between partition function and window size in multifractal theory is applied to verify that the mineral phase has fractal structure and can be analyzed by the multifractal theory parameters.Multifractal theory is used to calculate the singular index and singular spectrum of mineral phase.The edges of mineral phase are extracted according to these two important parameters,and compared with the edges of mineral phase extracted by improved algorithm based on fractal dimension.The research results show that the edge of mineral phase extracted by multifractal algorithm is relatively clear and complete,which can be used as an effective and intuitive expression of the mineral phase visual characteristics of pellets.There are regularities between the multifractal spectral width,the alkalinity and surface uniformity of pellets,which provides a basis for establishing quantitative relationship between them.Figure 48;Table 5;Reference 69.