Segmentation And Statistics Research Of Complex Image

Ore mining and dressing needs the size distribution information in mining industry. In the process of control and optimize mining and ore dressing, ore particle size distribution determines the quality of crushing, production rate and energy consumption. Using image analysis to measure mineral particles can improve real-time and accuracy of measurement. As the ore particles image is relatively complicated, such as ore image has noticeable edges sometimes, and when the ore particles accumulate, it is difficult to identify the boundary of each ore use a single traditional segmentation, so that ore image can not be accurately divided on conveyor belt.In this paper, focus on the segmentation of ore target in video image on a conveyor belt, the target analysis and morphological analysis of segmented image, the average particle size and particle size distribution of ore and other issues.First, preprocessing is needed to eliminate noises before segmentation, which include image filtering and image binarization processing. By analyzing image features, adopt bilateral filtering method and the double-threshold binarization method. Next, according to the knowledge of morphological, carry out distance transformation and morphological reconstruction for binarization image and obtain the local maximum of ore image. Then the reconstructed image is segmented, that is, the proposed segmentation based on seed boundary growing method, which extract seed regions and edges according to characteristics of reconstructed image at the first, and formulate boundary growth conditions, at the same time formulate stop growing conditions to determine whether it is the ore boundaries. Finally, make statistics to segmented image and count number and area of the ore target.In this paper, do real-time processing for ore video on belt. Two-sided filter reduced the noise information, distance transformation and image reconstruction extracted local maximum of the image, that is, the extraction of seed points. The proposed segmentation method takes each separate object as its processing unit rather than the whole image to segment ore image, so that segmentation result is more accurate. By analyzing image segmentation result, the number and morphology of ore are consistent with the visual results. Experimental results show that the proposed system and segmentation method suitable for particle size analysis of ore image on belt.