紅外線自動相機-野生動物觀測影像

目前在林業試驗所生態感測平台http://iesn.tfri.gov.tw/forestDW/針對蛙類聲景監測、珍稀植物菱形奴草線上直播監測、以及野生動物監測在蓮華池研究中心農委會林業試驗所的試驗林一號及二號量水堰之間，設置一處面積25公頃(500m×500m)之蓮華池闊葉林長期動態樣區，並於區域內架設紅外線自動相機進行野生動物的監測。目前經這幾年各個錄影機錄影拍攝到的動物資料哺乳類有：山羌、石虎、獼猴、穿山甲、白鼻心、赤腹松鼠、野豬、食蟹獴，鳥類有：藍腹鷴、竹雞、深山竹雞、黑冠麻鷺、虎鶇、翠翼鳩。資料收集方式以動物經過觸發錄影30秒後停止錄影，錄影間隔為每5分鐘不重複錄影，以防止因動物重複來回觸發錄影造成檔案過多，記憶卡容量不足與電池過度消耗而無電力。目前資料收取間隔以一個月人員上山更換記憶卡與充電電池收取回的影像檔以人工的方式進行自動相機編號歸檔，影像拍攝日期建立檔名。然後再經過人工的方式一個一個影像檔一一觀看，進行人工動物分類辨識作業再將分類好的資料以動物名稱以日經過次數、月經過次數、年總數量作為日後分析的數據。

The dataset is collected via the Ecological Sensing Platform of the Taiwan Forestry Research Institute (TFRI) at http://iesn.tfri.gov.tw/forestDW/. A 25-hectare (500m × 500m) permanent broad-leaved forest dynamic plot was established between Flow Gauging Weirs 1 and 2 of the experimental forest at the Lienhuachih Research Center, Council of Agriculture (COA), TFRI, for three monitoring targets: amphibian acoustic monitoring, live-stream monitoring of the rare parasitic plant *Mitrastemon yamamotoi*, and wildlife monitoring. Infrared automatic cameras were deployed within the plot for wildlife surveillance. Over the past years, the captured wildlife footage includes the following mammalian species: Formosan Reeves's muntjac, leopard cat, Formosan rock macaque, Chinese pangolin, masked palm civet, red-bellied tree squirrel, wild boar, and crab-eating mongoose; and avian species: Swinhoe's pheasant, Chinese bamboo partridge, Taiwan bamboo partridge, black-crowned night heron, tiger thrush, and emerald dove. The data collection protocol specifies that each camera will record 30 seconds of footage after being triggered by a passing animal, with a 5-minute cooldown interval between consecutive recordings to avoid redundant footage caused by repeated visits from the same animal, which would otherwise lead to excessive storage consumption, insufficient memory card space, and rapid battery depletion. Field personnel retrieve the recorded footage monthly by replacing memory cards and recharging the on-site batteries. The retrieved footage is manually archived by camera ID, with filenames named according to the image capture date. Subsequently, each video file is manually reviewed to identify and classify the recorded animals. The classified data are aggregated by species name, with three core metrics for subsequent analysis: daily visit count, monthly visit count, and annual total abundance.