English  |  正體中文  |  简体中文  |  Items with full text/Total items : 28611/40649
Visitors : 641353      Online Users : 70
RC Version 4.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
Scope Adv. Search
LoginUploadHelpAboutAdminister

Please use this identifier to cite or link to this item: http://ntour.ntou.edu.tw:8080/ir/handle/987654321/54121

Title: 應用深度學習物件偵測於改進水下測量魚體尺寸精準度
On Accuracy Improvement of Underwater Fish Size Measurement Using Deep Learning for object detection
Authors: Tu, Zhe-Wei
涂哲瑋
Contributors: NTOU:Department of Electrical Engineering
國立臺灣海洋大學:電機工程學系
Keywords: 深度學習;物件偵測;立體視覺;人工智慧;智慧養殖;水下影像處理
Deep Learning;Object Detection;Stereo Vision;Artificial Intelligence;Smart Cultivation;Underwater Image Processing
Date: 2019
Issue Date: 2020-07-09T03:02:09Z
Abstract: 本論文分別採用深度學習(Deep Learning)物件偵測、影像處理(Image processing)及立體視覺(Stereo Vision)技術建置水下量測系統,應用於水下各種魚類尺寸之自動量測。 在AI物件偵測上,藉由Faster-RCNN、Yolo_v3等深度學習網路可以產生具有穩定長寬比例的Bounding box,因此本論將此一特性應用於定位水下影像中魚隻之座標,以獲得左右影像包含魚隻物件的RoI(Region of Interest)進行立體匹配(Matching)並取得魚隻的三維世界空間座標,並透過比例推算及線性回歸演算法來計算體長、重量。因為水下影像的魚體特徵不明顯,不容易獲得可用的視差圖(disparity map),因此吾人使用邊緣偵測方法來增強匹配的效果進而提高尺寸量測的精確度。此量測系統之特點係將AI深度學習技術導入於建構台灣的智慧養殖產業,可減少人力資源、降低成本,提升在國際養殖產業技術的競爭力。 關鍵詞:深度學習、物件偵測、立體視覺、人工智慧、智慧養殖、水下影像處理
Keywords ─Deep Learning, Object Detection, Stereo Vision, Artificial Intelligence, Smart Cultivation, Underwater Image Processing. This thesis applies Deep Learning, Image Processing and Stereo Vision technology to the implementation of an underwater measurement system for automatic measurement of various fish sizes under water. In AI object detection, deep learning is widely known as being capable of generating a bounding box with a stable aspect ratio. This work utilizes this property to help locate the coordinates of fish in the underwater image, thus obtaining the RoI (Region of Interest) of the left and right images containing the target fish. These coordinates are fed to the stereo matching for acquiring the disparity map and hence the 3D world space coordinates of the fish, and the body length and weight are estimated through proportion calculation and linear regression. Because the texture feature of the underwater fish is not salient due to scattering and hazing phenomenon, it is not easy to obtain an accurate disparity map. Therefore, image processing techniques such as edge detection is applied to enhance the matching effect and improve the accuracy of the size measurement as well. This work is characterized by introducing AI deep learning techniques to the implementation of an underwater measurement system, which is useful in reducing human resources, costs and enhancing the international competitiveness of Taiwan aquaculture industry.
URI: http://ethesys.lib.ntou.edu.tw/cgi-bin/gs32/gsweb.cgi?o=dstdcdr&s=G0010653020.id
http://ntour.ntou.edu.tw:8080/ir/handle/987654321/54121
Appears in Collections:[電機工程學系] 博碩士論文

Files in This Item:

File SizeFormat
index.html0KbHTML20View/Open


All items in NTOUR are protected by copyright, with all rights reserved.

 


著作權政策宣告: 本網站之內容為國立臺灣海洋大學所收錄之機構典藏,無償提供學術研究與公眾教育等公益性使用,請合理使用本網站之內容,以尊重著作權人之權益。
網站維護: 海大圖資處 圖書系統組
DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - Feedback