基于图像特征匹配的煤泥浮选泡沫速度特征提取方法

郭中天; 王然风; 付翔; 魏凯; 王宇龙

doi:10.13272/j.issn.1671-251x.17991

基于图像特征匹配的煤泥浮选泡沫速度特征提取方法

doi: 10.13272/j.issn.1671-251x.17991

郭中天^1,,
王然风^1, ,,
付翔^{1, 2},
魏凯¹,
王宇龙¹

1.
太原理工大学矿业工程学院, 山西太原　030024
2.
山西焦煤集团有限责任公司博士后工作站, 山西太原　030024

基金项目: 国家自然科学基金项目(52274157)；内蒙古自治区重点专项项目(2022EEDSKJXM010)；山西省重点研发计划项目(202102100401015)

详细信息

作者简介:
郭中天（1998—），男，山西长治人，硕士研究生，主要研究方向为煤泥浮选智能化及图像处理，E-mail：583246098@qq.com

通讯作者:
王然风（1970—），男，山西长子人，副教授，博士，主要研究方向为智能化开采与分选，E-mail：wrf197010@126.com

中图分类号: TD948
计量
- 文章访问数: 157
- HTML全文浏览量: 77
- PDF下载量: 31
- 被引次数: 0
出版历程
- 收稿日期: 2022-08-02
- 修回日期: 2022-09-25
- 网络出版日期: 2022-10-13

Method for extracting froth velocity of coal slime flotation based on image feature matching

GUO Zhongtian^1
,,
WANG Ranfeng^{1
, ,},
FU Xiang^{1, 2},
WEI Kai¹,
WANG Yulong¹

1.
College of Mining Engineering, Taiyuan University of Technology, Taiyuan 030024, China
2.
Postdoctoral Workstation, Shanxi Coking Coal Group Co., Ltd., Taiyuan 030024, China

摘要

摘要: 煤泥浮选泡沫图像局部静态特征相似，一些较为复杂的工况判断需要用到浮选泡沫图像的动态特征，而现有的针对煤泥浮选泡沫速度动态特征的提取方法存在准确性、实时性和稳定性不足问题。针对上述问题，提出了一种基于图像特征匹配的煤泥浮选泡沫速度特征提取方法。首先，采用限制对比度自适应直方图均衡化（CLAHE）和三维块匹配滤波（BM3D）对浮选泡沫图像进行预处理，以提高图像质量，突出图像的边缘细节特征。其次，采用加速KAZE（AKAZE）算法对浮选泡沫特征进行特征点检测。然后，在利用暴力匹配（BF）对泡沫图像特征进行粗匹配的基础上，采用基于网格的运动统计（GMS）算法快速可靠地区分正确与错误的特征匹配。最后，根据特征匹配结果计算煤泥浮选泡沫速度，并以此为测量值，利用卡尔曼运动估计方法对得到的测量值进行迭代修正，得到更稳定的煤泥浮选泡沫速度特征。实验结果表明：① AKAZE−GMS算法较好地解决特征点簇集的同时又尽量保留了更多数量的特征点，这是因为预处理后图像受噪声影响降低、对比度增强、边缘特征更突出。② 与SIFT（尺度不变特征转换）、SURF（加速稳健特征）、AKAZE算法相比，AKAZE−GMS算法匹配对分布更为均匀，保留了更多正确的匹配对，匹配精度达99.99%，且运行时间仅需3.73 s。③ 直接经过特征匹配结果计算得到的泡沫速度测量值波动幅度较大，测量值经过卡尔曼运动估计修正后的速度估计值较为平稳，更符合真实工况。
- 煤泥浮选 /
- 浮选泡沫速度 /
- 浮选泡沫动态特征 /
- 图像特征匹配 /
- 快速鲁棒特征匹配过滤 /
- 卡尔曼运动估计
Abstract: The local static characteristics of coal slime flotation foam image are similar. The dynamic characteristics of flotation foam image are needed for judging some complex working conditions. The existing extraction method for the dynamic features of the froth velocity of coal slime flotation has the problems of insufficient accuracy, real-time performance and stability. In order to solve the above problems, a feature extraction method of froth velocity in coal slime flotation based on image feature matching is proposed. Firstly, the contrast limited adaptive histogram equalization (CLAHE) and block-matching and 3D filtering(BM3D) are used to preprocess the flotation froth image to improve the quality of the image and highlight the edge details of the image. Secondly, the accelerated-KAZE (AKAZE) algorithm of accelerated features in nonlinear scale space is used to detect the feature points of flotation froth features. Thirdly, on the basis of rough matching of froth image features by brute-force matching (BF), a grid-based motion statistics (GMS) algorithm is used to quickly and reliably distinguish correct and wrong feature matching. Finally, the method calculates the slime foam velocity according to the feature matching results. The foam velocity is taken as the measured value. The Kalman motion estimation method is used to iteratively modify the measured values to obtain more stable foam velocity characteristics of coal slime flotation. The experimental results show the following points. ① The AKAZE-GMS algorithm can solve the problem of feature point clustering well and keep more feature points as much as possible. This is because the preprocessed image is less affected by noise, has better contrast, and has more prominent edge features. ② Compared with SIFT (scale-invariant feature transform), SURF (speeded up robust features) and AKAZE, the AKAZE-GMS algorithm has a more uniform distribution of matching pairs, retains more correct matching pairs. The method achieves a matching accuracy of 99.99%. The running time is only 3.73 s. ③ The measured value of froth velocity directly calculated from the feature matching results fluctuates greatly. The velocity estimated value of the measured value corrected by Kalman motion estimation is more stable, which is more consistent with the real working condition.
- coal slime flotation /
- flotation froth velocity /
- dynamic features of flotation froth /
- image feature matching /
- fast robust feature matching filtering /
- Kalman motion estimation

HTML全文

图 1 相邻帧泡沫位置变化

Figure 1. Change of froth position in adjacent frames

下载: 全尺寸图片幻灯片

图 2 煤泥浮选泡沫速度特征提取流程

Figure 2. Coal slime flotation froth velocity feature extraction process

下载: 全尺寸图片幻灯片

图 3 煤泥浮选泡沫图像采集系统

Figure 3. Image acquisition system for coal slime flotation froth

下载: 全尺寸图片幻灯片

图 4 各算法特征检测结果对比

Figure 4. Comparison of the feature test results of each algorithm

下载: 全尺寸图片幻灯片

图 5 各算法特征匹配结果对比

Figure 5. Comparison of the feature matching results of each algorithm

下载: 全尺寸图片幻灯片

图 6 煤泥浮选泡沫速度计算结果

Figure 6. Calculated coal slime flotation froth velocity results

下载: 全尺寸图片幻灯片

表 1 匹配精度及运行时间

Table 1. Matching accuracy and running time

算法	总匹配对	正确匹配对	正确率/%	时间/s
SIFT	556	540	97.12	15.74
SURF	308	301	97.73	12.84
AKAZE	736	732	99.46	9.76
AKAZE−GMS	2585	2584	99.99	3.73

下载: 导出CSV

表 2 不同算法提取速度特征统计分析

Table 2. Statistical analysis of speed features extracted by different algorithms 像素/s

算法	测量值		卡尔曼运动估计修正后测量值
算法	速度均值	标准差	速度均值	标准差
SIFT	9.356 5	7.314 7	12.419 2	5.153 3
SURF	12.309 8	7.602 8	12.311 6	1.408 7
AKAZE	12.752 5	7.008 0	12.318 7	1.383 0
AKAZE−GMS	10.974 2	5.045 6	11.181 3	1.188 2

下载: 导出CSV

参考文献(16)

[1]	曹文艳,王然风,樊民强,等. 基于半监督聚类的煤泥浮选泡沫图像分类方法[J]. 工矿自动化,2019,45(7):38-42,65. doi: 10.13272/j.issn.1671-251x.17437 CAO Wenyan,WANG Ranfeng,FAN Minqiang,et al. Coal slime flotation foam image classification method based on semi-supervised clustering[J]. Industry and Mine Automation,2019,45(7):38-42,65. doi: 10.13272/j.issn.1671-251x.17437
[2]	成佳. 基于浮选泡沫图像处理的精矿品位建模[D]. 成都: 电子科技大学, 2020. CHENG Jia. Based on image processing of flotation foam modeling of concentrate grade[D]. Chengdu: University of Electronic Science and Technology of China, 2020.
[3]	王然风, 高建川, 付翔. 智能化选煤厂架构及关键技术[J]. 工矿自动化, 2019, 45（7）: 28-32. WANG Ranfeng, GAO Jianchuan, FU Xiang. Framework and key technologies of intelligent coal preparation plant[J]. Industry and Mine Automation, 2019, 45（7）: 28-32.
[4]	桂卫华,阳春华,徐德刚,等. 基于机器视觉的矿物浮选过程监控技术研究进展[J]. 自动化学报,2013,39(11):1879-1888. doi: 10.3724/SP.J.1004.2013.01879 GUI Weihua,YANG Chunhua,XU Degang,et al. Machine-vision-based on online measruing and controlling technologies for mineral flotation-a review[J]. Acta Automatica Sinica,2013,39(11):1879-1888. doi: 10.3724/SP.J.1004.2013.01879
[5]	程翠兰. 基于颜色与纹理特征的矿物浮选精选泡沫分类[D]. 长沙: 中南大学, 2010. CHENG Cuilan. Mineral float-selected foam classification based on color and texture characteristics[D]. Changsha: Central South University, 2010.
[6]	桂卫华,廖茜,阳春华,等. 一种新的浮选泡沫图像纹理特征提取方法[J]. 中国科技论文,2012,7(4):277-281. doi: 10.3969/j.issn.2095-2783.2012.04.007 GUI Weihua,LIAO Xi,YANG Chunhua,et al. A new texture extraction method for flotation froth images[J]. China Sciencepaper,2012,7(4):277-281. doi: 10.3969/j.issn.2095-2783.2012.04.007
[7]	冯仁光. 基于自适应权重FCM的浮选泡沫图像分割及应用[D]. 大连: 大连理工大学, 2014. FENG Renguang. A flotation froth image segmentation based on an improved adaptive weight FCM and its application[D]. Dalian: Dalian University of Technology, 2014.
[8]	LIU Jinping,GUI Weihua,TANG Zhaohui. Flow velocity measurement and analysis based on froth image SIFT features and Kalman filter for froth flotation[J]. Turkish Journal of Electrical Enectrical Engineering and Computer Sciences,2013,21:2378-2396.
[9]	刘颖,张平,赵珺,等. 基于尺度不变特征变换的浮选泡沫图像动态特性提取方法[J]. 控制理论与应用,2016,33(6):718-726. doi: 10.7641/CTA.2016.50648 LIU Ying,ZHANG Ping,ZHAO Jun,et al. Dynamic characteristic extraction method of flotation froth image based on scale invariant feature transform[J]. Control Theory & Applications,2016,33(6):718-726. doi: 10.7641/CTA.2016.50648
[10]	龚云,杨庞彬,颉昕宇. 结合同态滤波与直方图均衡化的井下图像匹配算法[J]. 工矿自动化,2021,47(10):37-41,61. doi: 10.13272/j.issn.1671-251x.2021070018 GONG Yun,YANG Pangbin,JIE Xinyu. Underground image matching algorithm combining homomorphic filtering and histogram equalization[J]. Industry and Mine Automation,2021,47(10):37-41,61. doi: 10.13272/j.issn.1671-251x.2021070018
[11]	朱成德,李志伟,王凯,等. 基于改进网格运动统计特征的图像匹配算法[J]. 计算机应用,2019,39(8):2396-2401. doi: 10.11772/j.issn.1001-9081.2018122590 ZHU Chengde,LI Zhiwei,WANG Kai,et al. Image matching algorithm based on improved RANSAC-GMS[J]. Journal of Computer Applications,2019,39(8):2396-2401. doi: 10.11772/j.issn.1001-9081.2018122590
[12]	智宁,毛善君,李梅. 基于照度调整的矿井非均匀照度视频图像增强算法[J]. 煤炭学报,2017,42(8):2190-2197. ZHI Ning,MAO Shanjun,LI Mei. Enhancement algorithm based on illumination adjustment for non-uniform illuminance video images in coal mine[J]. Journal of China Coal Society,2017,42(8):2190-2197.
[13]	肖佳,张俊华,梅礼晔. 基于改进BM3D算法的椒盐噪声去噪[J]. 计算机工程与应用,2018,54(21):170-175,207. doi: 10.3778/j.issn.1002-8331.1806-0362 XIAO Jia,ZHANG Junhua,MEI Liye. Salt and pepper noise denoising based on improved BM3D algorithm[J]. Computer Engineering and Applications,2018,54(21):170-175,207. doi: 10.3778/j.issn.1002-8331.1806-0362
[14]	张润梅,宦思琪,张媛,等. 基于改进AKAZE的特征匹配算法及应用[J]. 重庆理工大学学报(自然科学),2021,35(7):266-275. ZHANG Runmei,HUAN Siqi,ZHANG Yuan,et al. Feature matching algorithm of Huizhou architecture based on improved AKAZE[J]. Journal of Chongqing University of Technology(Natural Science) ,2021,35(7):266-275.
[15]	ALCANTARILLA P F. Fast explicit diffusion for accelerated features in nonlinear scale spaces[C]. British Machine Vision Conference （BMVC）, London, 2013.
[16]	刘江,郭荣春,王燕妮. 基于卡尔曼滤波的高斯混合模型目标检测算法[J]. 探测与控制学报,2022,44(2):79-84. LIU Jiang,GUO Rongchun,WANG Yanni. A Gaussian mixture kalman filter algorithm of target detection[J]. Journal of Detection & Control,2022,44(2):79-84.