張文亮臺灣大學：生物環境系統工程學研究所蕭友晉Shiau, Yo-JinYo-JinShiau2007-11-272018-06-292007-11-272018-06-292006http://ntur.lib.ntu.edu.tw//handle/246246/56021本研究以混合高斯理論判斷靜態與動態影像中之生物數量。 靜態影像實驗以台北縣新海橋人工溼地之浮萍為研究對象，共拍攝五種不同的覆蓋數量，依序為336棵、412棵、1,010棵、1,528棵以及1,924棵。在葉片沒有重疊的條件下，辨識正確率分別為：95.7%、76.8%、68.2%、82.1%以及90%。誤差產生的最大原因在於水面反光。 動態影像實驗分成室內與現地兩部分，室內實驗以三組不同大小魚隻組合以及三種不同水質濁度進行辨識。正確率隨魚隻體長漸小而逐漸下降，依序為79.7%、78.4%及64.2%。此外，正確率亦隨濁度增加而遞減，當濁度為2.9NTU、10NTU及15NTU時，正確率依序為78.4%、60.4%及45.5%。 現地實驗使用水下攝影機實際對台北縣坪林鄉金瓜寮溪當地溪床進行魚隻辨識。計算結果魚隻流通量為 。實驗得知當魚隻相互遮蔽或魚隻貼近河床游動產生陰影時，較易干擾系統之辨識結果。 研究認為，混合高斯理論未來應可作為生態工程物種數量的計算模式之一。This paper discusses the organism-counting ability of the Mixture-Gaussian theory on photograph analysis and video analysis. The test samples of the photograph experiment were the duckweeds in Xin-Hai-Qiao Constructed Wetlands in Taipei County. The actual leaf-covering quantities of duckweeds on five photos were 336, 412, 1010, 1528 and 1928. Without overlapping of leaves, the counting accuracy rates were 95.7%, 76.8%, 68.2%, 82.1 % and 90%, respectively. The major cause of counting errors was water reflection. The video experiment was done in the laboratory and in the field. In the laboratory, three fish sizes were mixed with three kinds of water turbidity in order to find out the counting variables. As a result, when fishes were smaller, the counting accuracy tended to be inaccurate, from 79.7% to 78.4% and 64.2%. Furthermore, the counting accuracy also decreased when water turbidity increased. In fact, the water turbidity was 2.9 NTU, 10 NTU and 15 NTU, and the counting accuracy was 78.4%, 60.4% and 45.5%. In the field, the underwater video camera was placed into Jin-Gua-Liao River in Taipei County to videotape fishes at the riverbed. After the videotape was brought back to the laboratory, it was analysed by the Mixture-Gaussian theory. The counting result of fish-flux was 4.26fish/minm3. As a matter of fact, fish-overlapping and fish-shadows of close swimming at the riverbed truly affected the counting accuracy. The study shows that it is possible to consider the Mixture-Gaussian theory as an evaluation method of organism-counting in the field of ecological engineering.目錄 I 圖目錄 III 表目錄 V 中文摘要 1 英文摘要 2 第一章 前言 3 1.1文獻回顧 4 1.2研究目的 6 第二章 理論與模式 7 第三章 材料與方法 14 3.1靜態影像辨識 14 3.2動態影像辨識 16 3.2.1室內實驗 16 3.2.2室外實驗 18 3.3辨識流程 20 第四章 靜態影像辨識結果與討論 22 第五章 動態影像辨識結果與討論 27 5.1室內實驗 27 5.1.1魚類大小變化辨識結果 28 5.1.2水質濁度變化辨識結果 30 5.2野外實驗 32 第六章 結論 36 第七章 建議 37 參考文獻 38 附錄A A-1 附錄B B-1 附錄C C-11949638 bytesapplication/pdfen-US混合高斯理論魚類族群影像計數浮萍Mixture-Gaussian theoryfish populationphotographic countingduckweed[SDGs]SDG15thesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/56021/1/ntu-95-R93622008-1.pdf