煤泥浮选泡沫图像识别系统研究

发布时间：2018-06-11 16:17

本文选题：煤泥浮选 + 泡沫图像　；参考：《中国矿业大学》2017年硕士论文

【摘要】：浮选是利用矿物颗粒表面润湿性的不同来实现矿物分选的一门技术,其泡沫的表面视觉特征直接反映浮选生产工况。目前,国内大多数选厂对浮选状态的判别仍然停留在人工观测的基础上,极大地限制了浮选技术的推广和发展。将图像识别技术引入浮选不但可以有助于浮选作业的稳定优化运行,而且可以解放出大量的人力和物力,对其的后续发展具有十分重要的意义。本文以“赵楼煤矿浮选加药自动控制系统”项目为背景,对煤泥浮选泡沫图像识别系统进行了深入的研究,有效地克服了传统浮选工艺所固有的弊病,其研究内容主要可概括为如下几个方面:1)针对煤泥浮选泡沫图像易受到混合噪声,特别是高斯-椒盐混合噪声,干扰的情况,提出了一种新型滤波算法,其不仅在去除泡沫图像的混合噪声上要明显优于传统滤波算法,而且可以满足一定的实时性需求。2)对煤泥浮选泡沫图像的特征提取算法进行了研究,并选择纹理、尺寸和速度特征参数作为系统的输入。其中,本文分析和总结了传统分水岭算法的不足之处,并对其进行了相关改进,基本实现了煤泥浮选泡沫图像的精准分割。3)通过经验公式和相关实验,确定了BP神经网络隐含层神经元的最优数目,并在相同网络参数下,对比了BP神经网络和遗传BP神经网络的性能。进一步确定了遗传BP神经网络作为本系统的识别算法。4)出于对浮选车间工作环境等方面的考虑,对工业相机进行了合理的选型,并根据系统需求,对其后台数据库和前台界面进行了设计和介绍。
[Abstract]:Flotation is a technique to realize mineral separation by using the different wettability of mineral particle surface. The surface visual characteristics of the foam directly reflect the flotation production conditions. At present, the discrimination of flotation state in most domestic concentrators still stays on the basis of manual observation, which greatly limits the popularization and development of flotation technology. The introduction of image recognition technology into flotation can not only contribute to the stable and optimal operation of flotation operations, but also liberate a large number of manpower and material resources, which is of great significance to the subsequent development of flotation operations. In this paper, based on the project of "automatic control system for flotation of Zhaolou Coal Mine", the image recognition system of coal slime flotation foam is deeply studied, which effectively overcomes the inherent disadvantages of traditional flotation technology. The main research contents can be summarized as follows: (1) A new filtering algorithm is proposed to solve the problem of mixed noise, especially Gao Si salt and pepper mixed noise, in the slime flotation foam image. It is not only superior to the traditional filtering algorithm in removing the mixed noise of foam image, but also can meet the real-time requirement. 2) the feature extraction algorithm of coal slime floatation foam image is studied, and the texture is selected. The dimension and velocity characteristic parameters are used as the input of the system. Among them, this paper analyzes and summarizes the shortcomings of the traditional watershed algorithm, and improves it, basically realizes the accurate segmentation of slime floatation foam image. 3) through empirical formula and related experiments, The optimal number of hidden layer neurons in BP neural network is determined, and the performance of BP neural network and genetic BP neural network are compared under the same network parameters. Furthermore, it is determined that genetic BP neural network is the recognition algorithm of the system. 4) considering the working environment of floatation workshop, the industrial camera is selected reasonably, and according to the system requirement, The background database and front interface are designed and introduced.
【学位授予单位】：中国矿业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD923;TD94;TP391.41

【参考文献】