本文選題：浮選圖像 + 特征提取　； 參考：《昆明理工大學》2017年碩士論文

【摘要】：浮選在金屬生產(chǎn)中是一個重要的一環(huán),在浮選車間中,通過一定條件下,然后加入適量的藥劑及適量的空氣,然后可以開始攪拌使浮選藥劑與金屬進行充分的接觸,這樣才能在攪拌的過程中出現(xiàn)很多的浮選泡沫,最后通過對這些泡沫的回收利用,從泡沫當中提取出原礦品。浮選泡沫的特征如大小,形狀,顏色。浮選泡沫大小不一致,浮選泡沫之間黏連性較強,表面紋理復雜等特點。浮選過程大多依靠人工觀察浮選泡沫狀態(tài),但因此也會對人體產(chǎn)生不必要的傷害,導致浮選過程難以優(yōu)化。本論文主要介紹了人工控制下的浮選會造成對資源的浪費,人體的危害,以及縮小資源回收率等問題,為解決這些問題,我們考慮通過計算機圖像處理技術(shù)來代替人工控制,這就需要將圖像處理技術(shù)應用到浮選泡沫圖像處理中。同時介紹了浮選泡沫圖像處理過程的復雜性,及泡沫圖像處理的困難性,如果處理成功,將大大提高金屬的回收率,降低生產(chǎn)成本,減少資源浪費,同時降低對人工的需求,這樣也可以避免浮選車間環(huán)境對人體的危害,提高整體的生產(chǎn)效率,提高整個浮選,有色金屬回收行業(yè)的生產(chǎn)效率。從而促進整個國民經(jīng)濟的發(fā)展,進而也能擴大計算機圖像處理技術(shù)的應用發(fā)展空間,促進計算機圖像處理技術(shù)超前邁進。在論文中我們主要從浮選泡沫圖像采集,到浮選泡沫圖像去噪,緊接著浮選泡沫圖像分割然后對進行浮選泡沫圖像特征提取這一過程進行了詳細的描述,并針對浮選泡沫圖像特征提取的三個方面:大小,形狀,顏色來做詳細的解釋。同時又由于浮選過程對實時性的要求,我們考慮引入GPU及多線程來提高浮選泡沫圖像處理技術(shù)的效率。經(jīng)過兩年的努力,在老師及同學的幫助下主要實現(xiàn)了以下工作:本文針對浮選泡沫的特點,分析提取浮選泡沫的大小,形狀及顏色等特征。本論文主要研究工作如下:(1)通過小波分析及多尺度理論,對浮選泡沫的尺寸特征進行提取。(2)通過研究浮選泡沫圖像分割技術(shù),例如分水嶺分割算法,來提取浮選圖像形狀特征。(3)通過分析比較浮選圖像RGB顏色值分布,提取浮選圖像顏色特征。(4)對相關(guān)程序進行優(yōu)化,利用GPU基于CUDA平臺增強圖像處理的實時性。
[Abstract]:Floatation is an important ring in metal production. In the floatation workshop, under certain conditions, appropriate amount of reagents and proper air are added, and then stirring can be started to make the flotation reagents in full contact with the metal.In this way, a lot of floatation foam can appear in the process of stirring. Finally, through the recovery and utilization of these foams, the raw ore can be extracted from the foam.Flotation foam features such as size, shape, color.The size of flotation foam is different, the adhesion between flotation foam is strong, the surface texture is complex and so on.The flotation process mostly depends on the artificial observation of the flotation foam state, but it will also cause unnecessary harm to the human body, resulting in the flotation process is difficult to optimize.This paper mainly introduces the problems that floatation under manual control will cause waste of resources, harm to human body, and reduce the recovery rate of resources. In order to solve these problems, we consider replacing manual control with computer image processing technology.Therefore, it is necessary to apply image processing technology to flotation foam image processing.At the same time, the complexity of flotation foam image processing and the difficulty of foam image processing are introduced. If the processing is successful, the recovery rate of metals will be greatly increased, the production cost will be reduced, the waste of resources will be reduced, and the demand for manpower will be reduced.In this way, the environment of floatation workshop can avoid the harm to human body, improve the overall production efficiency, and improve the production efficiency of the whole floatation and non-ferrous metal recovery industry.Thus, the development of the whole national economy can be promoted, and the application and development space of the computer image processing technology can also be expanded, and the computer image processing technology can be advanced.In the paper we mainly from flotation foam image acquisition to flotation foam image de-noising followed by flotation foam image segmentation and then the flotation foam image feature extraction process is described in detail.Three aspects of feature extraction of flotation foam image: size, shape and color are explained in detail.At the same time, due to the real-time requirement of flotation process, we consider introducing GPU and multi-thread to improve the efficiency of flotation foam image processing technology.With the help of teachers and students, the following work has been realized: according to the characteristics of flotation foam, the size, shape and color of flotation foam are analyzed and extracted in this paper.The main research work of this thesis is as follows: (1) by wavelet analysis and multi-scale theory, the size characteristics of flotation foam are extracted. (2) by studying flotation foam image segmentation techniques, such as watershed segmentation algorithm,By analyzing and comparing the distribution of RGB color value of flotation image, extracting the color feature of flotation image, we optimize the relevant program, and use GPU to enhance the real-time performance of image processing based on CUDA platform.
【學位授予單位】：昆明理工大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TD923;TP391.41

【參考文獻】

相關(guān)期刊論文 前10條

1 劉小波;;浮選泡沫圖像主成分的分類與識別[J];云南冶金;2014年03期

2 許景波;袁怡寶;劉泊;林海軍;李建新;;時域測量中高斯低通濾波器的有理逼近設(shè)計方法[J];固體電子學研究與進展;2011年06期

3 陳志強;郭永亮;陳詩哲;;基于小波變換與改進的Roberts算子融合的圖像邊緣檢測[J];傳感器世界;2011年04期

4 何桂春;馮金妮;吳藝鵬;鄭錫聯(lián);;浮選泡沫圖像處理技術(shù)研究現(xiàn)狀與進展[J];有色金屬科學與工程;2011年02期

5 劉金平;桂衛(wèi)華;牟學民;唐朝暉;李建奇;;基于Gabor小波的浮選泡沫圖像紋理特征提取[J];儀器儀表學報;2010年08期

6 周開軍;陽春華;牟學民;桂衛(wèi)華;;基于泡沫特征與LS-SVM的浮選回收率預測[J];儀器儀表學報;2009年06期

7 陽春華;周開軍;牟學民;桂衛(wèi)華;;基于計算機視覺的浮選泡沫顏色及尺寸測量方法[J];儀器儀表學報;2009年04期

8 楊衛(wèi)莉;郭雷;;基于分水嶺算法和圖論的圖像分割[J];計算機工程與應用;2007年07期

9 劉明艷;趙景秀;孫寧;;用Prewitt算子細化邊緣[J];光電子技術(shù);2006年04期

10 顧欽;基于GPU計算的虛擬現(xiàn)實仿真系統(tǒng)設(shè)計模型[J];微處理機;2005年02期

相關(guān)碩士學位論文 前3條

1 趙彥;基于GPU技術(shù)的增強現(xiàn)實系統(tǒng)[D];電子科技大學;2009年

2 胡珊;圖像檢索中基于SVM的相關(guān)反饋技術(shù)研究[D];西北工業(yè)大學;2007年

3 侯黎黎;利用圖像處理技術(shù)實現(xiàn)浮選泡沫分類的研究[D];吉林大學;2006年

，

本文編號：1745978