The Particle Size Detection Of Crushed Ore Based On Image Segmentation

The granularity information of the crushed ore is the main indicator which availability react crushers working condition at all levels in the crushing process, and directly determines the productivity of the entire ore dressing process.Through detecting the particle size information of crushed ore, we can obtain the size information in real time, and timely feedback the information.Through the particle size distribution of crushed ore, we adjust the width of crusher discharge mouth to ensure the adjustment of particle size distribution of crushed ore. Accurately detect the size distribution of crushed ore to ensure the require of particle size distribution, at the same time, ensure the optimized control by full ore implementation process. Thereby, enhancing the dressing efficiency and reducing energy consumption, maximizing the use of mineral resources.We should have the segmentation and positioning to crushed ore of collected ore image before the segmentation of ore image. But because the ore image have the following characteristics,such as a large amount,irregular shape,stacked in serious condition, and size differences,which make a big difficult for ore segmentation. At the same time, due to the complexity of the scene of crushed ore environments, which results in serious noise on the ore image surface.For over various reasons, currently single ore segmentation algorithms can't segment complex ore image.The segmentation effect is not ideal.This paper put forward a image segmentation algorithm which based on canny edge detection and watershed merging.At first, we select the pre-process method of ore image according to the noise impact of the ore image. Then the binarization operation is carried on the pre-processed image to obtain a binary image. Then we do distance transform for binary image.Give the bilateral filtering proce ss for the distance image.Using watershed algorithm for distance im age. It got the dividing lines.Directly extract the edge of binary by canny edge detection algorithm,due to the adhesion between crushed ores, leading to many under-segmentation phenomenon on extracted edge image.For less divided problem of ore particles, using watersh ed segmentation algorithm to compensate its shortcomings.In other words, we merged Canny edge detection algorithm and watershed segmentation algorithm. Finally, quantitative description methods of ore size parameters were studied and ultimately we get size informa tion of ore particles. Specific research work and innovation are as follows:1, Pre-treatment methods study of ore image. Since the surface of selected ore image has serious noise and adhesion phenomenon, so choose a reasonable noise filter method is the top priority of image pre-processing. Bilateral filtering as a non-linear spatial filter, inconsidering the relationship between the space pixels also considering the relationship between pixels gray.Therefore,the use of bilateral filtering algorithm retaining the edge of ore particles while filtering the noise which can greatly reduce the problem of image edge blurring. More help of extracting the edge contour of ore image. The method of integral image and local threshold merging is carried on the image,we get a binary image to achieve a target extraction of ore image.The algorithm has high light irradiation and high efficiency. Much noise and small “holes” exist in the binary image, so the optimization for binary image is necessary. We use morphological method to optimize the binary image which is simple and can remove the burrs. Smooth the edge of broken particle image and in addition remove the noise.2, The research of distance transform of binary image. Distance transform is a transform algorithm for binary image. There is a serious over-segmentation phenomenon existing in the dividing lines if we directly do the watershed segmentation for distance image. Therefore we use bilateral filtering to filter out those redundant extreme points of distance image, removing the noise. After filtering process, then do watershed segmentation for distance image, extremely reducing the over-segmentation phenomenon.3, Merge Canny edge detecting algorithm with watershed algorithm.Optimize the dividing lines of watershed segmentation, merging thedividing line of watershed segmentation with the edge of extracted binary image to obtain the final crushed ore image which effectively reduce the under-segmentation and over-segmentation regions.4, The size parameter extraction of ore particle. The size of crushed particle is difficult to quantitative description. We choose the size parameters applied to ore particles. We make the pixel calibration for ore segmentation regions on the basis of image segmentation,then we obtain the ore perimeter and area, finally we get crushed ore particle size,volume and other parameters according to Kemeny empirical formula. At last, the size distribution detection of ore is completed.