環境資源報告成果查詢系統

113年北部地區環境污染運用科技工具與AI智能判讀追蹤計畫

中文摘要 本計畫透過陳情案件、重大空氣及水污染案件之稽查裁處資料以及環管署提供相關資料,分析指定區域之潛在污染源進行熱區圖層套疊與繪製,並分別於大園工業區提供2處空污判煙架設參考地點、觀音工業區提供2處空污判煙架設參考地點以及於桃園市河川提供3處水色辨識架設參考點為,共計7處可架設位置供環管署參考。 已於113年4月26日完成2組空污及1組水色之雲端影像監視設備架設,另於113年5月2日完成4組異味感測器之設備架設。以及於113年5月16日提供水色影像智慧辨識與異常影像資訊擷取功能正式上線。使用者可透過手機、平板、電腦進入平臺觀看即時影像、歷次異常影像以及近一個月之歷史影像調閱。而Line推播通報功能亦持續提供服務。 統計自113年1月1日至113年11月30日止,共計辨識出2,307則異常冒煙事件,其辨識成功率達87.43%;而水色異常事件部分,則是統計自113年5月16日至113年11月30日止,共辨識出189則異常水色事件,其辨識成功率達84.13%;而異味偵測達警戒值之告警事件,統計自113年6月1日至113年11月30日止共計有114則異味告警事件,而比對影像後,確實有發現其異常冒煙情形,達到「聞的到且看的到」之成效。 另已於計畫期間內,執行2式揮發性有機物成分分析檢測與現場排放檢測,其中包含12瓶揮發性有機物成分分析、2天現場排放檢測及分析報告。並於113年11月27日完成1場次之成果發表會。
中文關鍵字 固定污染源、影像辨識、環境監控

基本資訊

專案計畫編號 經費年度 113 計畫經費 4600 千元
專案開始日期 2024/01/10 專案結束日期 2024/11/30 專案主持人 王御安
主辦單位 北區環境管理中心 承辦人 王慶元 執行單位 振興發科技有限公司

成果下載

類型 檔名 檔案大小 說明
期末報告 113EA008.pdf 21MB

Environmental Pollution Monitoring and Tracking Project with Technology Tools and AI Intelligence Interpretation in the Northern Region in 2024

英文摘要 This project utilizes public nuisance petition cases, inspection and penalty data from major air and water pollution incidents, along with relevant information provided by the Environmental Management Administration (EMA), to analyze potential pollution sources in designated areas. Heatmap layering and mapping are carried out for analyzing these regions. Specifically, reference locations for air pollution smoke detection are provided at two sites in the Dayuan Industrial Zone and two sites in the Guanyin Industrial Zone, while three reference points for water color identification are provided at rivers in Taoyuan City. In total, seven locations are recommended for installation, which are provided for the EMA's consideration. On April 26, 2024, two sets of air pollution monitoring systems and one set of water color monitoring systems have been successfully installed. On May 2, 2024, four sets of odor sensors have been also set up. Additionally, on May 16, 2024, the cloud-based smart image recognition and abnormal image extraction functions for water color monitoring were officially launched. Users can access real-time images, historical abnormal images, and the past month’s image data through mobile phones, tablets, or computers. The platform also continues to provide Line notification feature messages. From January 1 to November 30, 2024, 2,307 abnormal smoke events were identified, achieving a recognition success rate of 87.43%. For water color abnormalities, 189 events were identified from May 16 to November 30, 2024, with a recognition success rate of 84.13%. Regarding odor detection, 114 odor alarm events were recorded from June 1 to November 30, 2024. After cross-referencing images, abnormal smoke occurrences were confirmed, successfully achieving the “smell and see” effect. Additionally, two volatile organic compound (VOC) analysis tests and on-site emission measurements were conducted during the project period. This included the analysis of 12 bottles of VOC samples and two days of on-site emission testing, with the corresponding analysis reports completed. The presentation of project execution results was held on November 27, 2024.
英文關鍵字 Stationary sources, Image recognition, Environmental monitoring