Research On Detection And Classification For Mineral Microparticle Image

In recent years,the micro structure and element distribution of minerals can be determined by the image of ground mineral particles collecting from electron microscope.Although the electron microscopy avoids the influence of individual subjective and improves the efficiency,there are problems of transition zone and particle adhesion in the collected mineral micro images.On the one hand,these problems will result in low segmentation accuracy and affect the subsequent classification of mineral elements.On the other hand,there are many kinds of mineral element data,high dimension and small quantity,which cause the low accuracy of classification.Therefore,on the basis of the existing segmentation and classification methods,two kinds of problems are studied in this paper.One is to deal with the transition region and conglutinated particles in the mineral micrograph image.The other is the fine segmentation of particle image and the classification of mineral elements.To summary,there are some good contributions in this paper:(1)The mineral microgranular image is preprocessed,that is,the transition region and adhesion region of the mineral microgranular image are processed by using the improved k-means algorithm.On the one hand,the improved k-means algorithm and canny algorithm are used to extract the transition region of mineral microscopic particles.Then,the transition region is removed by morphology,and the cavity is filled.On the other hand,a novel adhesion segmentation algorithm is proposed based on watershed transform of distance.Firstly,the adhesion point is extracted by multiple corrosion.Then,the distance between the fixed point and the particles with small relative area is calculated to reduce the adhesion area.Finally,Freeman chain code corner detection is used to check and correct the segmentation line.Experimental results show that the algorithm is effective.(2)The pre-processed image is segmented.This paper proposes an improved image segmentation algorithm based on graph theory.The core of the algorithm is that the merging conditions in the image segmentation algorithm based on graph are changed.At first,the maximum gray value and the minimum gray value in the two regions are made difference,and the two regions less than the set threshold are merged.If the above condition is not satisfied,the average difference of gray level between the two regions is calculated,and the two regions less than the set mean threshold are merged.Experimental results show that the improved algorithm can achieve fine segmentation of mineral particles.(3)The mineral elements of the segmented mineral microparticle images are classified.Due to the variety and quantity of mineral elements,this paper proposes a new classification model of multiple hyperspheres based on support vector machine.At first,the center of each class is obtained by using the support vector data description algorithm.Then a hypersphere with as large a radius as possible is constructed,which contains as many samples as possible.Experiments show that the model can effectively realize the classification of mineral elements.