本文選題：藻類 + 全息成像　； 參考：《南昌航空大學(xué)》2017年碩士論文

【摘要】：藻類是所有植物中最古老的,大多數(shù)藻類生活在水中。藻類不僅具有為水域漁業(yè)生產(chǎn)提供營養(yǎng)基礎(chǔ)的重要意義,而且可以通過水體中藻類細(xì)胞的數(shù)量為判斷水質(zhì)是否污染提供依據(jù)。在對藻類識別技術(shù)的研究中,傳統(tǒng)的研究步驟由為顯微觀測、形態(tài)分析以及計數(shù)統(tǒng)計組成。主要通過人眼觀測,易導(dǎo)致視覺疲勞、效率低、且數(shù)據(jù)不能保存。隨著現(xiàn)代數(shù)字圖像處理技術(shù)的發(fā)展,實現(xiàn)了基于數(shù)字全息對藻類細(xì)胞進(jìn)行識別計數(shù)與分析。數(shù)字全息與傳統(tǒng)光學(xué)全息相比具有制作成本低,成像速度快,記錄和再現(xiàn)靈活等優(yōu)點,其記錄與再現(xiàn)過程都可以通過數(shù)字化處理。對全息再現(xiàn)的藻類樣本圖進(jìn)行HOG特征提取,結(jié)合SVM監(jiān)督學(xué)習(xí)模型實現(xiàn)對藻類細(xì)胞高效、便捷的藻類細(xì)胞分類計數(shù)。本文首先對數(shù)字全息無透鏡成像的基本理論進(jìn)行分析。針對藻類細(xì)胞的特點,基于無透鏡全息成像理論制作出簡單易操作的無透鏡全息成像裝置。在此裝置的基礎(chǔ)上進(jìn)行藻類的計數(shù)與分析。本論文具體內(nèi)容如下:1.針對目前傳統(tǒng)藻類檢測的不足,本文采用數(shù)字圖像處理技術(shù)對藻類細(xì)胞進(jìn)行研究。2.介紹了無透鏡全息成像原理以及展示了無透鏡全息成像的光路圖。對無透鏡全息再現(xiàn)技術(shù)進(jìn)行詳細(xì)論述,根據(jù)無透鏡全息理論與本論文對藻類細(xì)胞研究需要,設(shè)計了一個可以易于攜帶且質(zhì)量輕、體積小的全息無透成像裝置,并且利用3D打印機(jī)做出實物裝置。介紹具體使用操作步驟以及對比了此裝置與傳統(tǒng)裝置相比具有的優(yōu)勢。3.為了對藻類細(xì)胞進(jìn)行快速、高效的特征提取,運用HOG描述算子對藻類進(jìn)行研究。HOG僅在圖像的局部單元上操作,因此它對圖像幾何與光學(xué)的形變都能保持較好的不變性。此外,只要在粗的空域抽樣、精細(xì)的方向抽樣以及較強(qiáng)的局部光學(xué)歸一化等條件下,藻類細(xì)胞檢測效果就不會受其他因素影響。HOG特征適合用于圖像中的藻類檢測。4.利用SVM監(jiān)督學(xué)習(xí)模型對藻類細(xì)胞進(jìn)行目標(biāo)識別且進(jìn)行計數(shù)與分析。SVM與傳統(tǒng)學(xué)習(xí)方法(如模式識別、神經(jīng)網(wǎng)絡(luò))相比,它基于結(jié)構(gòu)風(fēng)險最小化原則,泛華能力強(qiáng)。它是一個凸優(yōu)化問題,因此局部最優(yōu)解一定是全局最優(yōu)點。此外,SVM解決了線性與非線性的分類問題。
[Abstract]:Algae are the oldest of all plants, and most of them live in water. Algae not only has the important meaning of providing nutritive basis for fishery production in water area, but also provides the basis for judging whether the water quality is polluted or not by the number of algae cells in the water body. In the research of algal recognition technology, the traditional research steps consist of microscopic observation, morphological analysis and counting statistics. Mainly through the human eye observation, easy to lead to visual fatigue, low efficiency, and data can not be preserved. With the development of modern digital image processing technology, algal cell recognition counting and analysis based on digital holography is realized. Compared with traditional optical holography, digital holography has the advantages of low cost, fast imaging speed, flexible recording and reproducing, and can be digitally processed. The holographic reconstruction of algae sample map was performed with HOG feature extraction and the SVM supervised learning model was used to realize the efficient and convenient classification and counting of algal cells. In this paper, the basic theory of digital holographic lensless imaging is analyzed. Based on the lensless holographic imaging theory, a simple and easy to operate lensless holographic imaging device was developed according to the characteristics of algae cells. On the basis of this device, algae counting and analysis are carried out. The content of this thesis is as follows: 1. Aiming at the deficiency of traditional algal detection at present, this paper uses digital image processing technology to study algal cells. 2. 2. The principle of lensless holographic imaging and the optical path of lensless holographic imaging are introduced. Based on the theory of lensless holography and the need of algae cell research in this paper, a holographic imaging device is designed, which is easy to carry, light in weight and small in volume. And using a 3D printer to make a physical device. The operation steps are introduced and the advantages of this device compared with the traditional device are compared. In order to extract the algal cells quickly and efficiently, the HOG description operator is used to study the algae. Hog only operates on the local unit of the image, so it can keep good invariance to the geometric and optical deformation of the image. In addition, under the conditions of coarse spatial sampling, fine direction sampling and strong local optical normalization, the effect of algal cell detection will not be affected by other factors. Compared with traditional learning methods (such as pattern recognition, neural network), SVM supervised learning model is based on structural risk minimization principle. It is a convex optimization problem, so the local optimal solution must be the global optimal. In addition, SVM solves the problem of linear and nonlinear classification.
【學(xué)位授予單位】：南昌航空大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：Q949.2;TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 王道明;魯昌華;蔣薇薇;肖明霞;李必然;;基于粒子群算法的決策樹SVM多分類方法研究[J];電子測量與儀器學(xué)報;2015年04期

2 徐超;高夢珠;查宇鋒;曹利民;;基于HOG和SVM的公交乘客人流量統(tǒng)計算法[J];儀器儀表學(xué)報;2015年02期

3 張楚金;王耀南;盧笑;王珂娜;王國峰;;基于假設(shè)驗證和改進(jìn)HOG特征的前車檢測算法[J];電子測量與儀器學(xué)報;2015年02期

4 劉威;段成偉;遇冰;柴麗穎;袁淮;趙宏;;基于后驗HOG特征的多姿態(tài)行人檢測[J];電子學(xué)報;2015年02期

5 喬靈博;王迎新;趙自然;陳志強(qiáng);;毫米波全息成像的空間采樣條件[J];清華大學(xué)學(xué)報(自然科學(xué)版);2014年11期

6 萬源;李歡歡;童恒慶;吳克風(fēng);;分層CS-LBP和HOG特征融合的人臉識別研究[J];武漢理工大學(xué)學(xué)報(交通科學(xué)與工程版);2014年04期

7 向征;譚恒良;馬爭鳴;;改進(jìn)的HOG和Gabor,LBP性能比較[J];計算機(jī)輔助設(shè)計與圖形學(xué)學(xué)報;2012年06期

8 楊敏;張樹生;洪歧;伍新中;;利用小波多尺度變換和矩進(jìn)行形狀特征提取[J];北京生物醫(yī)學(xué)工程;2008年01期

9 劉俊梅;;基于小波變換的圖像紋理特征提取技術(shù)[J];計算機(jī)工程與設(shè)計;2007年13期

10 侯艷麗;;基于改進(jìn)的小波變換的紋理特征提取算法[J];科技信息(科學(xué)教研);2007年17期

相關(guān)博士學(xué)位論文 前2條

1 周皓;數(shù)字全息成像技術(shù)應(yīng)用研究[D];蘇州大學(xué);2015年

2 賴小敏;雙球面波自全息成像[D];華中科技大學(xué);2014年

相關(guān)碩士學(xué)位論文 前1條

1 鄭睿;藻類細(xì)胞圖像的特征提取與分類方法研究[D];杭州電子科技大學(xué);2009年

，

本文編號：1939434