基于旋转分类模糊纹理谱的矿物浮选泡沫平滑度分析.pdf

第 30 卷第 9 期 2013 年 9 月 控 制 理 论 与 应 用 Control Theory 纹理特征; 模糊控制; 旋转分类; 机器视觉 中图分类号 TP29文献标识码 A Froth homogeneity analysis using rotate classifi cation fuzzy texture spectrum for mineral fl otation process monitoring ZHOU Xuan, YANG Chun-hua, REN Hui-feng†, CHEN Zhi-peng, ZENG Zi-jiao School of Ination Science and Engineering, Central South University, Changsha Hunan 410083, China Abstract The concentration grade in mineral fl otation is closely related to the froth texture complexity. To deal with the defectsofrigidcomparisonofbasictexturespectrum, wepresentanovelrotateclassifi cationfuzzytexturespectrum for mineral separation process from engineering aspects. The frequency-statistics curve of neighborhood pixels grayscale difference is fi tted by using an exponential function, and the optimal threshold value of the fuzzy internal is determined. According to the grayscale difference between neighborhood pixels and seed pixel in local region, the fuzzy texture unit is constructed, which is used to calculate the fuzzy texture spectrum of various fuzzy texture unit numbers. Meanwhile, the original 6561 texture units are simplifi ed to 834 classes. The homogeneity and the coarseness are introduced to describe the froth texture feature. The analysis results of froth images with different grades in spot show that the texture homogeneity can refl ect the concentration grade in real-time. Key words froth fl otation; texture feature; fuzzy control; rotate classifi cation; computer vision 1Introduction Copper, aluminum, lead, zinc and other metals are closelyrelatedtohumanlife. However, astheconsumption is growing gradually, corresponding mineral resources are diminishing. At present, most of the existing ores are lean or symbiotic mine, so it’s severely signifi cant to extract and recycle the valuable metals effectively. Froth fl ota- tion is the application of surface science to separate and purify valuable minerals from gangue according to differ- ent hydrophobic of the surface properties of various min- erals[1–2] . Because of its great adaptability and high effi - ciency, it now becomes the most widely used of separation and purifi cation of the low grade or complex ore[3–4] . In recent years, fl otation has been extended to medical and health, environmental protection[5], resource recycling[6] and many other fi elds. Researches have shown that visual characteristics of the surface froth in fl otation cell are the comprehensive re- fl ection of mineral physical properties, fl otation reagents, and mechanical and operating conditions[7]. With the ad- vantages of computer science and image processing tech- nology, computer-vision is widely used in industry pro- cess monitoring[8–10]. Experts studied bubbles color[11–13], size[14–15], velocity[16], and other features[17–18] of fl otation froth under computer-vision, deduced the relationships be- tweenbubblesappearance and fl otation perance[19–21]. The homogeneity of bubbles surface can be described by image texture, which can refl ect the enrichment of min- eral on the froth, so the froth texture tied to concentrate grade closely[22].Marais[23]and Aldrich[24]studied the texture of froth based on gray level co-occurrence ma- trix GLCM and proved the relationship between mineral grade and recovery. But GLCM is beyond the model of human vision, and the global use of ination is miss- ing. Hy otyniemi has extracted the froth texture based on GGHA and power spectrum using fast fourier trans FFT[25]. However this is compute-intensive for simulation only. Bartolacci used Wavelet to extract tex- ture features of froth images[20]. But the Wavelet for extraction of texture quality is limited to the selection of fi lter bank issues. The requirement of multiple fi lter banks causes a number of parameters need to be deter- Received 9 March 2012; revised 17 May 2013. †Corresponding author. E-mail renhf316; Tel. 86 13548681086. This work was supported by the National Natural Science Foundation of China No.61134006, the National Science Foundation for Distinguished Young Scholars No.61025015, and Foundation Sciences Central South University. 1154Control Theory 阳阳阳春春春华华华1965–, 女, 博士, 教授, 目前研究方向为复杂工业过 程的建模与优化、 实时系统容差计算等, E-mail ychh; 任任任会会会峰峰峰1981–, 男, 博士, 目前研究方向为图像处理与模式识 别、 基于视觉图像的软测量等, E-mail renhf316; 陈陈陈志志志鹏鹏鹏1987–, 男, 硕士, 目前研究方向为图像处理, E-mail 466628563; 曾曾曾子子子骄骄骄1988–, 男, 硕士研究生, 目前研究方向为图像处理, E-mail 239452054.