基于聚类预分割和高低精度距离重构的彩色浮选泡沫图像分割
doi: 10.3724/SP.J.1146.2006.01980
A Segmentation Method Based on Clustering Pre-segmentation and High-low Scale Distance Reconstruction for Colour Froth Image
-
摘要: 该文针对矿物浮选过程泡沫图像质量不理想、气泡大小形状灰度不均的问题,提出一种基于聚类预分割和高低精度距离重构的泡沫图像分割方法。首先,利用k-均值聚类进行前景泡沫与背景矿浆彩色图像分割,依据灰度分布和形状分布特征对提取到的泡沫图像进行滤波;然后,基于形态重构提出结合高低精度距离变换对距离图像进行重构,同时利用面积重构h顶改进变换为分水岭变换提取准确的特征标识;最后利用分水岭算法得到分水线,从而完成浮选泡沫的分割。由分割后的泡沫图像可统计分析出气泡个数与尺寸等物理特征参数从而为浮选控制提供依据。仿真结果表明了方法的有效性。Abstract: Due to a large variation in the quality of froth images of ore and inhomogeneity of size, shape and grayscale of bubbles, a new segmentation method based on clustering pre-segmentation and high-low scale distance reconstruction is proposed for froth images. Firstly, the segmentation between foreground froth and background mineral slurry image is achieved by the k-means clustering method and the noises are filtered according to intensity distribution and shape distribution information, and a new reconstruction combined with high-low scale distance transformation based on morphological reconstruction is presented and applied to the froth distance- transformation image. Then the precise region makers for watershed transformation are extracted by area-reconstruction h-dome improved transformation. Finally, the watershed algorithm is used to get waterline for every bubble. Bubble physical characteristics such as the bubble number and bubble size can be obtained from the segmented image,which provide the guidance for flotation control process. The experimental results show its effectiveness.
-
Moolman D W, Aldrich C, and Schmitz G P J. Theinterrelationship between surface froth characteristics andindustrial flotation performance[J].Minerals Enginerring.1996,9(8):837-854[2]Kaartinen J and Hatonen J. Machine-vision-based control ofzinc flotationA case study[J].Control Engineering Practice.2006, 14(12):1455-1466[3]Moolman D W, Aldrich C, and Van Deventer J S J. Theinterpretation of flotation froth surfaces by using digitalimage analysis and neural networks[J].Chemical EngineeringScience.1995, 50(22):3501-3513[4]Sadr-Kazemi N and Cilliers J J. An image processingalgorithm for measurement of flotation froth bubble size andshape distribution[J].Minerals Engineering.1997, 10(10):1075-1083[5]Wang W X, Bergholm F, and Yang B. Froth delineationbased on image classification[J].Mineral Engineering.2003, 16(11):1183-1192[6]冯征. 一种基于粗糙集的k -means 聚类算法. 计算机工程与应用, 2006, 42(20): 141-146.Feng Zheng. A rough-based K-means clustering algorithm.computer. Engineering and Applications, 2006, 42(20):141-146.[7]Salembier P and Flat J S. Zones filtering connected operatorsand filters by reconstruction. IEEE Trans. on ImageProcessing, 1995, 4(8): 1153-1160.[8]Vincent L. Morphological grayscale reconstruction in imageanalysis: Applications and efficient algorithms[J].IEEE Trans.on Image Processing.1993, 2(2):176-201[9]岳振军, 邱望成, 刘春林. 一种自适应的多目标图像分割方法.中国图象图形学报, 2004, 9(6): 674-678.Yue Zhen-jun, Qiu Wang-cheng, and Liu Chun-lin. Aself-adaptive approach of multi-object image segmentation.Journal of Image and Graphics, 2004, 9(6): 674-678.[10]Vincent L and Soille P. Watersheds in digital spaces: Anefficient algorithm based on immersion simulations[J].IEEETrans. on Pattern Analysis and Machine Intelligence.1991,13(6):583-598 -
计量
- 文章访问数: 3342
- HTML全文浏览量: 102
- PDF下载量: 1306
- 被引次数: 0
下载:
