| 研究生: |
吳俊樓 Wu, Chun-Lou |
|---|---|
| 論文名稱: |
運用影像處理技術擷取延繩釣揚繩作業影片中之漁獲資訊 Getting the Information of Catches from the Videos of Hauling Activities on Long-line Fishing Vessels by Image Processing Techniques |
| 指導教授: |
林忠宏
Lin, Chung-Hung |
| 學位類別: |
碩士 Master |
| 系所名稱: |
工學院 - 系統及船舶機電工程學系 Department of Systems and Naval Mechatronic Engineering |
| 論文出版年: | 2012 |
| 畢業學年度: | 100 |
| 語文別: | 中文 |
| 論文頁數: | 103 |
| 中文關鍵詞: | 船上監控系統 、背景抽取演算法 、Sobel遮罩 、影像相減法 、經驗模態分析法 、倒傳遞類神經網路 |
| 外文關鍵詞: | Monitoring systems on fishing vessles, Background extraction algorithm, Sobel mask, Image subtraction algorithm, Empirical mode decomposition, Back propagation nerual network |
| 相關次數: | 點閱:100 下載:0 |
| 分享至: |
| 查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
本論文以漁船上監視系統錄製影片之後續影像處理為研究重點。
安裝於漁船上的監視系統最大的問題為設備必須能夠抵抗海上嚴峻的環境,提供後續影像處理作業人員清晰可辨的影像。避免拍攝影像過於模糊的作法是開發一組程式能夠自動辨識畫面是否模糊而發出警告,通知船員必須擦拭鏡頭。程式使用背景抽取演算法抽取出不同時間之背景圖片,接著以Sobel遮罩計算出影像邊緣銳利度,將不同時間背景圖片進行比較,得到畫面之模糊程度,稱為模糊化百分比。當模糊化百分比高於自訂的門檻值則判斷鏡頭必須擦拭。實際測試的結果認為模糊化百分比曲線能夠表現出拍攝畫面的模糊程度,畫面模糊化百分比大約在50%時便無法看清楚鏡頭前的物體。
後續影像處理很重要的一部分工作是量測畫面中的漁獲長度。為了瞭解各種可能會影響魚體長量測誤差的因素,在實驗室中進行控制條件下之影像量測實驗,探討這些因素對於誤差有何影響,並利用漁船出港期間拍攝之影像進行魚體長量測誤差分析。實驗室實驗結果,魚體長量測誤差在建議的拍攝條件下都在5%以下。
考慮自動化魚長量測為未來趨勢,本論文探討以影像相減法、經驗模態分析法進行自動化魚體輪廓獲取、長度量測之可行性。在實驗室中大略模擬了漁船上監控系統拍攝時有雜訊影響的狀況,以上述兩種方法進行自動化魚長量測。使用經驗模態分析法量測魚體長之成功率為29%,影像相減法為50%。成功量測到魚體長的情況下,兩種方法的魚體長量測誤差都在6%以下。
在自動化魚長量測獲取輪廓後,本論文亦探討了利用魚體輪廓擷取魚體特徵數據,以此特徵數據作為倒傳遞類神經網路辨識鯊魚、旗魚、鮪魚的依據之可行性。以手動方式獲取魚體特徵數據並用來分類鯊魚、旗魚、鮪魚之正確率都在93%以上。
This research focus on the image processing techniques used to process videos recorded by monitoring systems installed on fishing vessels.
The biggest problem of monitoring systems installed on fishing vessels is that it is hard to record clear videos due to the severe environment at sea.To prevent the inconvenients of processing blur videos,we develop a program that can be used to notify the crew it is time to clear the lens on the cameras because of the videos recorded is getting blur.This program extracts background images using background image extraction techniques,then using Sobel mask to calculate the edge sharpness of the background images.Compare the image sharpness of background image of diffenent time to get percentage of blurred.If percentage of blurred is higher than the threshold we decided,it will notify the crew to clear the lens on the cameras.The results shows that the curve of percentage of blurred can quantify the blurnes of images. When percentage of blurred is higher than 60%,it will be determined over-blurred.
A very important task of the image processing process is to measure the length of catches recorded by the videos.To understand the effects of shooting conditions on the fish measuring error,we measure a target object under controlled conditions in the laboratory.We also used videos recorded by monitoring systems installed on fishing vessels to analysis fish measuring error. If the shooting conditions are under recommended control,the error will be under 5%。
Considering the trend of auto-fish-measuring,we used Empirical Mode Decomposition and Image Substraction algorithm to get the complete contours and length of fishes automatically.We measured fish length using the two methods described above under simulated conditions that are similar to the conditions on fishing vessles.The successs rate of Empirical Mode Decomposition and Image Substraction algorithm are 29% and 50%.When we get lengths of fishes using the methods describled above,the errors will be under 6%。
In auto-fish-measuring,we get the the complete contours of fishes.We test the feasibility of classify sharks、tunas and marlins by artificial neural network using features got from these contours.The success rate is above 93%。
[1] Ching-Lu Hsieh、Hsiang-Yun Chang、Fei-Hung Chen、Jhao-Huei Liou、Shui-Kai Chang、Ta-Te Lin,”A simple and effective digital imaging approach for tuna fish length measurement compatible with fishing operations”, Computers and Electronics in Agriculture,Vol.75,p.44-51,2010。
[2] Howard McElderry、Maria Jose Pria、Morgan Dyas、Randy McVeigh,” A PILOT STUDY USING EM IN THE HAWAIIAN LONGLINE FISHERY”,Archipelago Marine Research Ltd,2010。
[3] Jean-Yves Bouguet,”Camera Calibration Toolbox for Matlab”, MRL-Intel Corp,2011。
[4] Julie Bonney,Katy McGauley,”Testing the Use of Electronic Monitoring to Quantify At‐sea Halibut Discards in the Central Gulf of Alaska Rockfish Fishery”,EFP 07‐02 Final Report,2008。
[5] N. E. Huang, " The Empirical Mode Decomposition and The Hilbert Spectrum for Nonlinear and Non-stationary Time Series Analysis," Proc. R. Soc. Lond. A,Vol. 454,p.903-995。
[6] Zhengyou Zhang ,“Flexible camera calibration by viewing a plane from unknown orientations”,The Proceedings of the Seventh IEEE International Conference on,vol.1,p.666-673, 1999。
[7] 王俊明、陳世旺,” 漸進背景影像的建構”,碩士論文,國立臺灣師範大學,2002。
[8] 朱凱、王正林,”精通MATLAB神經網絡”,電子工業出版,2010。
[9] 江福松、倪怡訓、邵廣昭,“重要國際漁業議題發展趨勢分析座談會-探討永續漁業之內涵實錄集”,行政院農業委員會漁業署,2004。
[10] 何宜達,”視覺伺服技術於三維目標軌跡預測與攔截之應用”,碩士論文,國立成功大學,2002。
[11] 吳學良,”以影像為基礎之車輛偵測與計數系統”,碩士論文,國立成功大學,2011。
[12] 施凱軒,”類神經方法應用在渾濁水域影像重建之研究”,碩士論文,國立成功大學,2005
[13] 徐維揚,”門禁監控即時辨識系統之實用設計”,碩士論文,國立中央大學,2003。
[14] 莊守正、程一駿、戴永禔,“國際間對混獲議題之討論與我國之因應座談會”,行政院農業委員會漁業署,2003。
[15] 張允豪,”基於類神經網路於魷魚影像處理分級之研究”,碩士論文,國立成功大學,1996。
[16] 黃南祁,”倒傳遞網路在對接銲鋼材疲勞壽命預測上之應用”,碩士論文,國立成功大學,2001。
[17] 陳建隆,”應用改良式經驗模態分解法於消除文件影像中的不良光照現象”,碩士論文,國立中央大學,2009。
[18] 陳飛宏,“鮪釣船觀察員自動攝影系統之開發”,碩士論文,國立屏東科技大學,2007。
[19] 張德豐,”Matlab數字圖像處理”,機械工業出版社,2009。
[20] 劉瑞禎、于仕琪,”OpenCV教程基礎篇”, 北京航空航天大學出版社,2007。
[21] 劉震昌,”數位影像處理” ,高立圖書,2010。
[22] 鄭昭明,” 人類的圖形辨識”,科學月刊,0154期,1982。
[23] 謝清祿,”強化漁獲統計資料分析品質之研究 (第3年/全程3年)”, 行政院農業委員會96年度科技計畫研究報告,2007。
[24] http://perso.ens-lyon.fr/patrick.flandrin/emd.html,Empirical Mode Decomposition,2011。
[25] http://www.opencv.org.cn/index.php/%E9%A6%96%E9%A1%B5,“OpenCV中文網站”,2011。
[26] http://research.kahaku.go.jp/zoology/uodas/course/taxonomy/measurements/contents/fl.html ,体の各部分の割合,2011。
校內:2017-08-30公開