【作者】 杨丹丹；

【导师】 杨洁明；

【作者基本信息】 太原理工大学 ， 机械电子工程， 2012， 硕士

【摘要】 煤泥浮选主要是依据煤粒与矸石颗粒表面的物理化学性质的差异进行分选的,通过加入化学药剂,使可浮性较好、疏水的煤粒上升至表面形成浮选泡沫层,可浮性较差、亲水的矸石颗粒留在煤浆中成为浮选尾煤。泡沫表层粘附精煤颗粒的多少可以由泡沫视觉信息表征出来,所以通过提取浮选泡沫层的表面特征可以判断出浮选状态。目前不同状态下的浮选泡沫主要以图像形式来保存,采用工业CCD相机连续摄取浮选泡沫图像,通过提取图像中的泡沫特征来识别浮选状态。在浮选泡沫图像特征提取过程中需要进行一系列的预处理,方可提取出所需的参数。本文针对煤泥浮选泡沫图像噪声较大、对比度低、有气泡阴影等特点,分别研究了在图像特征提取过程中适合泡沫图像的各种处理方法,通过对这些方法的分析和比较,提出了一些改进算法,并通过仿真实验证明了本文所提方法的准确性。论文主要从以下几个方面作了研究：首先研究了图像去噪方法,分别对目前应用较多的面积重构开闭滤波方法和高斯滤波方法进行了改进,提出了一种基于面积重构开闭滤波与交替顺序滤波结合的去噪方法和一种自适应高斯滤波方法,并对两种去噪方法进行了对比,分析了每种方法适用的范围。然后对图像分割方法进行了研究,通过结合两种分割效果互补的方法——基于高斯滤波的标记分水岭分割和基于FCM的分水岭分割方法,提出了一种基于标记叠加的改进分水岭分割方法,经过仿真实验验证了这种方法的有效性,并将该方法与基于梯度低频成分中提取标记的分水岭分割和基于形态预处理和标记提取的分水岭分割方法进行了对比,进一步证实了这种方法的准确性。最后对浮选泡沫图像进行了特征提取,主要分析了纹理特征、尺寸特征、泡沫稳定度和泡沫承载量四种特征,并通过实验室浮选实验验证了这四种特征量的有效性。更多还原

【Abstract】 The main basis for coal flotation is the differences in physical and chemical properties of coal particles and gangue particles surface. Hydrophobic coal particles with better flotability rise to the surface to form the froth layer, while hydrophilic gangue particles with poor flotability remain in the coal slurry become flotation tailing by adding chemicals. The number of coal particles adhered in the bubble surface can be characterized by the foam visual information, so the flotation state can be judged by the surface characteristics of flotation foam layer. Currently, flotation foam under different state is saved in the form of image, flotation froth images are continuous intaked by Industrial CCD camera, the flotation states are identified by foam image’s characteristics. In order to extract the required parameters, a series of pretreatment should be done in the process of feature extraction of flotation froth image.In this paper, various image processing approaches for bubble images which have a large of noise, low contrast, bubble shadow and so on are researched in the image feature extraction process, some improved algorithms are imposed after analyzing and comparing these methods, and the accuracy of the proposed method is verified by simulation. The paper mainly studied the following aspects:Fistly, the image denoising method is researched, The area reconstruction by opening and closing filter method and the Gaussian filtering method are improved, it proposed two denoising methods-n algorithm combining with area reconstruction by opening and closing filter and alternating sequential filter and an adaptive Gaussian filtering method, and the scope of application of each method is analysed by comparing the two methods.Secondly, the image segmentation method is researched, it proposed an improved watershed segmentation method based on the mark superimposed by combining a marker watershed segmentation based on Gaussian filtering and a watershed segmentation method based on the FCM. It verifies the effectiveness of this method after simulation experiments. Besides, a watershed segmentation method by extracting the markers from low frequency components of the gradients and a watershed segmentation method based on morphological pre-processing and markers extraction were compared with this improved method, it further confirmed the accuracy of this method.Finally, the features of flotation froth images are extracted, which mainly include texture characteristics, size characteristics, foam stability and foam carrying capacity, The validity of the four characteristic quantities is verified by laboratory flotation tests.更多还原