本研究以電子觀察員系統為基礎，研究如何以漁船上視訊監視系統之影片，智慧化偵測及定位影像中的魚體位置。基於梯度方向直方圖 (Histogram of Oriented Gradient, HOG)的理論，我們提出三種特徵計算及判別的方法，對區塊影像進行有無魚的判別及完整影像的魚體偵測；方法一主要是計算影像梯度角度直方圖，對一張影像取出單一種特徵直方圖，並且利用K-means分群法得到10種有、無魚之模板特徵及建構倒傳遞類神經網路判別機制，對切割區塊影像(Bounding Box)進行判別。其中，分群法對於測試影像之魚體偵測正確率達到94.29%，類神經網路對測試影像魚體偵測正確率則為98.37%。但應用在完整影像作魚體偵測掃描時，結果卻是十不理想的。
對此，我們在研究作法二中提出「魚體全域特徵」及「魚體區域特徵」的特徵計算方式以及模板比對之判別機制，將影像之特徵由單一種擴增至7種，同時限縮模板種類為鮪魚，可改善研究方法一的魚體偵測正確率與無魚誤判率。以完整張數為計算單位，再給予一些特定條件後，85張完整測試影像中可得到魚體偵測率為85.9%。
為了消除並改進研究方法二中給予的限定條件及做法，在研究方法三中結合了Bagging演算法的概念並擴增了多種鮪魚模板及影像特徵之種類。以完整張數為計算單位的100張鮪魚及100張無魚影像中，可得到96.0%的有魚偵測率、89.71%的精確率及11.0%的無魚誤判率；若以Bounding Box觀點計算有、無魚誤判率的話，無魚誤判率甚至可以降到0.06%，而精確率也能達到89.93%，但鮪魚之偵測率會降至76.10%。最後嘗試利用鮪魚模板對鬼頭刀50張、旗魚和鯊魚各10影像進行偵測，其中以整張影像為計算單位，旗魚可達到100.0%、鬼頭刀有56.0%，而鯊魚只有50.0%的偵測率；但以Bounding Box觀點計算的話，旗魚有78.62%、鬼頭刀有53.07%，而鯊魚只有9.34%的偵測率，但鬼頭刀和鯊魚精確率皆可達100%，而旗魚也有99.39的精確率。

In this study, we try to detect and locate a fish in the image that is captured from the
Electronic Observer System. Based on Histogram of Gradient (HOG) theory, we present three methods to detect the position of fish. First, we fetch a single feature from an image, creating 10 types of templates by k-means clustering and building neural network to separate the features of fish from non-fish area. However, the result of experiment is very bad.

Therefore, we present the second method to improve the last result. We come up with the concept of global and local feature, increasing the species of feature for the image from one to seven, and narrow down the type of template to tuna. This method can effectively reduce the false alarm, and increase the fish detection rate at the same time, we can get 85.9% fish detection rate. Nevertheless, there are many limiting conditions in this experiment.

To remove and improve the limiting conditions of the second method, so we present the third method. We increasing the species of feature and the type of tuna template, combining the concept of Bagging theory with global and local feature. In the fish detection experiment, we can get 96.0% fish detection rate, 89.71% precision rate and 11.0% false alarm. In the other species of fish experiment, the fish detection rate is 56.0% in dolphin, 50% in shark, and 100% in swordfish.

[1] Breiman, L., “Bagging Predictors,” Machine Learning, Vol. 24, No. 2, pp.123-140, 1996.
[2] Chan, D., Hockaday, S., Tillett, R., & Ross, L.”A trainable n-tuple pattern classifier and its application for monitoring fish underwater”. Image Processing And Its Applications, 1999. Seventh International Conference on (Conf. Publ. No. 465), 1999.
[3] 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.
[4] D.M. Gavrila, V. Philomin, “Real-time object detection for “smart” vehicles,”IEEE Int. Conf. on Computer Vision, Vol. 1, pp. 87-93, 1999.
[5] D.M. Gavrila, “A Bayesian, Exemplar-Based Approach to Hierarchical Shape Matching,” IEEE Trans. on Pattern Analysis and Machine Intelligence, Vol. 29, Issue 8, pp. 1408-1421, Aug. 2007.
[6] Duc Thanh Nguyen, Wanqing Li, Philip Ogunbona, “A part-based template matching method for multi-view human detection,” Int. Conf. on Image and Vision Computing New Zealand, pp. 357-362, Nov. 2009.
[7] D. Geronimo, A.M. Lopez, A.D. Sappa, T. Graf, “Survey of Pedestrian Detection for Advanced Driver Assistance Systems,”IEEE Trans. on Pattern Analysis and Machine Intelligence, Vol. 32, Issue 7, pp. 1239-1258, July 2010.
[8] Gunilla Borgefors, “Distance Transformations in Digital Images,” Computer Vision, Graphics, and Image Processing, Vol. 34, Issue 3, pp. 344-371, June 1986.
[9] Howard McElderry、Maria Jose Pria、Morgan Dyas、Randy McVeigh,” A PILOT STUDY USING EM IN THE HAWAIIAN LONGLINE FISHERY”,Archipelago Marine Research Ltd, 2010。
[10] J. Koenderink and A. J. van Doorn, “Surface shape and curvature scales,” Image and Vision Computing, vol. 10, pp. 557-565, 1992.
[11] 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。
[12] Lee, D., Redd, S., Schoenberger, R., Xu, X., & Zhan, P. An automated fish species classification and migration monitoring system. Industrial Electronics Society, 2003. IECON'03. The 29th Annual Conference of the IEEE, 2003.
[13] Lee, D.-J., Schoenberger, R. B., Shiozawa, D., Xu, X., & Zhan, P. Contour matching for a fish recognition and migration-monitoring system. Optics East,2004
[14] Lee, J.-H., Wu, M.-Y., & Guo, Z.-C. A tank fish recognition and tracking system-using computer vision techniques. Computer Science and Information Technology (ICCSIT), 2010 3rd IEEE International Conference on, 2010.
[15] N. Dalal, B. Triggs, “Histograms of Oriented Gradients for Human Detection,”IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, Vol. 1, pp. 886-893, June 2005.
[16] N. Dalal, B. Triggs, C. Schmid, “Human Detection Using Oriented Histograms of Flow and Appearance,”In European Conf. on Computer Vision, pp. 7-13, 2006.
[17] P. Viola, M. Jones, “Rapid object detection using a boosted cascade of simple features,” IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, Vol. 1, pp. 511-518, 2001.
[18] P. Sabzmeydani, G. Mori, “Detecting Pedestrians by Learning Shapelet Features,”IEEE Conf. on Computer Vision and Pattern Recognition, pp. 1-8, June 2007.
[19] Qiang Zhu, Mei-Chen Yeh, Kwang-Ting Cheng, S. Avidan, “Fast Human Detection Using a Cascade of Histograms of Oriented Gradients,” IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, Vol. 2, pp. 1491-1498, 2006.
[20] Qing Jun Wang, Ru Bo Zhang, “LPP-HOG: A New Local Image Descriptor for Fast Human Detection,” IEEE Int. Symposium on Knowledge Acquisition and Modeling Workshop, pp. 640-643, Dec. 2008.
[21] S. Belongie, J. Malik, “Matching with Shape Contexts,” IEEE Workshop onContent-based Access of Image and Video Libraries, pp. 20-26, 2000.
[22] Spampinato, C., Giordano, D., Di Salvo, R., Chen-Burger, Y.-H. J., Fisher, R. B., & Nadarajan, G. Automatic fish classification for underwater species behavior understanding. Proceedings of the first ACM international workshop on Analysis and retrieval of tracked events and motion in imagery streams, 2010.
[23] Thanh Nguyen Duc, P. Ogunbona, Wanqing Li, “Human detection based on weighted template matching,”IEEE Int. Conf. on Multimedia and Expo, pp. 634-637, June 2009.
[24] White, D., Svellingen, C., & Strachan, N. Automated measurement of species and length of fish by computer vision. Fisheries Research, 80(2), 203-210, 2006.
[25] Xiaoyu Wang, Tony X. Han, Shuicheng Yan, “An HOG-LBP Human Detector with Partial Occlusion Handling,” IEEE Int. Conf. on Computer Vision, pp. 32-39, Sept. 2009.
[26] Yao, H., Duan, Q., Li, D., & Wang, J. An improved K-means clustering algorithm for fish image segmentation. Mathematical and Computer Modelling, 58(3), 790-798, 2013.
[27] Zion, B., Alchanatis, V., Ostrovsky, V., Barki, A., & Karplus, I. Real-time underwater sorting of edible fish species. Computers and Electronics in Agriculture, 56(1), 34-45, 2007.
[28] Zhihui Hao, Bo Wang, Juyuan Teng, “Fast Pedestrian Detection Based on Adaboost and Probability Template Matching,” Int. Conf. on Advanced Computer Control, Vol. 2, pp. 390-394, Mar. 2010.
[29] 毛銓毅，“語意驅動式HOG行人偵測”，碩士論文，國立高雄大學， 2011。
[30] 吳俊樓，“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”，碩士論文，國立成功大學，2012。
[31] 吳庭瑋，“適用於中小型延繩釣漁船之自動監控與資訊擷取系統之開發”，碩士論文，國立成功大學，2014。
[32] 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。
[33] 蔡欣明，“結合形狀特徵之增強型方向梯度直方圖行人偵測”，國立高雄第一科技大學電腦與通訊工程系，台灣，2009。
[34] 蕭培墉，莊振勛，黃世勳，蔡雅涵，傅立成，“整合形狀特徵之增強型方向梯度直方圖行人偵測”，影像辨識，Vol. 15，No. 2，2009。