无限场景中的矿物浮选泡沫图像形态抽样表征方法研究

项目名称： 无限场景中的矿物浮选泡沫图像形态抽样表征方法研究

项目编号： No.61304253

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 周开军

作者单位： 湖南商学院

项目金额： 25万元

中文摘要： 矿物浮选气泡以气、液、固三相共存，受工艺条件影响，气泡具有数量多、流速快及形态随机等特点，特别是泡沫层表面延伸至无限远处，使得无限场景中泡沫形态统计分布的客观表征十分困难，严重影响了浮选过程的生产效率。本项目拟系统地研究气泡颜色、形态及纹理等知识获取方法，并以此作为气泡底层知识；探讨基于图论的泡沫图像形态学枝剪方法，提取泡沫图像多维模式谱，设计时间与空间变化的形态结构元素，利用高效的进化算法进行优化求解，进一步地提取泡沫标识图像；研究气泡泛化知识学习方法，构建包含气泡细节特征的高层知识库；提出无限场景中的混态气泡图像区域分割方法，以高层知识指导底层分割，完成不同工况条件下的泡沫图像自学习分割；依据抽样的累积气泡样本，实现浮选槽总体泡沫形态分布参数估计。形成较为系统的浮选泡沫形态抽样表征理论与方法，为浮选泡沫形态表征提供新的思路和理论依据。成果的工业应用将有效促进我国矿产资源的可持续发展。

中文关键词： 矿物浮选；结构元素；仿生变换；图像分割；特征提取

英文摘要： Mineral flotation froth exists in three-phase states: solid, liquid and gas. large quantities, high flow speed and random morphological are the major characteristics of bubble due to the process conditions effect. Especially, bubble surface extends to infinity. Thus, it is difficult to characterize froth morphological statistic distribution objectively, which has seriously affected the production efficiency of flotation process. Firstly, the proposal intends to systematically research on knowledge acquisition method of froth color, morphology and texture, then, these characteristics are represented as bottom knowledge. Furthermore, bubble images multidimensional pattern spectral are extracted by morphological pruning algorithms based on graph theory, and the spatial-temporal varing morphological structural elements are designed. Meanwhile, a fast and efficient evolutionary algorithm is used to optimize the structure elements, on the basis, froth marker images are extracted. Secondly, bubbles generalization knowledge learning methods are put forward to build a high-level knowledge base with containing bubbles details features. Thirdly, a region segmentation method for hybird states bubble images is proposed under infinite-scene. According to high-level knowledge, the froth images self-learning segmentation algori

英文关键词： mineral flotation；structural element；biologically inspired transform；image segmentation；feature extraction