環保專案成果報告查詢系統

專案計畫編號	EPA-111-L001-31-125
經費年度	112
計畫經費	7,800 千元
預算科目	地方政府執行空氣污染防制工作
專案開始日期	2023/01/13
專案結束日期	2023/12/31
專案主持人	王應堂
主辦單位	臺南市政府環境保護局
承辦人	王秀惠
執行單位	堃捷工程股份有限公司
專案分類	大氣空氣
中文關鍵字	112年度臺南市逸散源空污AI科技監控計畫、工地AI科技監控、廟會AI監控、火災事故空污事件監控
英文關鍵字	During the implementation period of the project (from 13th January to 15th November 2023)
協同主持人
共同主持人
計畫聯絡人	王應堂
計畫聯絡電話	06-6331955#
計畫聯絡信箱	kingjacktech@gmail.com

中文摘要	計畫執行期間(112年01月13日至112年11月15日)，本計畫已完成6處工地AI科技監控設備及2處污染防制設備智能監控設置；工地AI科技監控方面，4月至11月共計有25,317筆車輛辨識紀錄，其中車牌辨識準確率為74.88%、防塵網覆網準確率約為97.0%，至於揚塵辨識功能初期由於受到環境干擾、鏡頭髒污、水漬等因素影響，導致錯誤差異較大，經優化調整後，目前揚塵告警準確率已提升至95%；經觀察，工地攔停未覆蓋車輛主動性有所提高，另相較於裝設初期，自主灑水也有所提高；9月份揚塵告警案件當中，有9成案件於通報時，工地已自主性進行灑水改善中；智能監控部分則是透過NBIoT回傳數據推估洗車用水量，兩處工地平均用水量分別約在686~691L/min(建築工程)、1086.7~1096.7L/min(疏濬工程)，另車輛經過洗車台時需要清洗30秒時間車輛才會乾淨。至於廟會繞境空污事件監控方面，已搭配環保局執行12場次廟會聯合稽查小組作業，另透過微感器完成6場次廟會遶境活動空品分析，此外配合環保局首創廟會AI監控，本計畫額外於富驛時尚酒店頂樓裝設監視器進行廟會煙火監控，並與本市空品監測網進行資料架接，當有微感器PM2.5濃度大於30μg/m3時，鏡頭會自動轉向該顆方向拍攝與辨識。目前已建立寺廟資訊、AI偵測與告警模式；廟會AI辨識系統於8月14日永華宮及10月14日興濟宮遶境活動共有13筆微感器高值告警，其中3筆微感器告警高值為廟會鞭炮引起；中秋、國慶連假期間AI辨識系統自動偵測到16件煙火施放，經分析比對未影響空品且無民眾陳情。另外有關火災事故空污事件監控方面，運用本市1,400個固定式微感器及連續監測特性導入於火警等空污事件監測，計畫期間共協助執行火災應變作業50件次，FB發布21件次、跨局處通報19件次、LINE發布6件次、跑馬燈發布7件次。當中以工廠火警所佔案件最多，並且於112年06月18日永康區雙O有限公司工廠火警派員攜帶MOT設備進行現場空品監控與數據蒐集，以分析有無影響空氣品質。
中文關鍵字	112年度臺南市逸散源空污AI科技監控計畫、工地AI科技監控、廟會AI監控、火災事故空污事件監控

中文摘要

計畫執行期間(112年01月13日至112年11月15日)，本計畫已完成6處工地AI科技監控設備及2處污染防制設備智能監控設置；工地AI科技監控方面，4月至11月共計有25,317筆車輛辨識紀錄，其中車牌辨識準確率為74.88%、防塵網覆網準確率約為97.0%，至於揚塵辨識功能初期由於受到環境干擾、鏡頭髒污、水漬等因素影響，導致錯誤差異較大，經優化調整後，目前揚塵告警準確率已提升至95%；經觀察，工地攔停未覆蓋車輛主動性有所提高，另相較於裝設初期，自主灑水也有所提高；9月份揚塵告警案件當中，有9成案件於通報時，工地已自主性進行灑水改善中；智能監控部分則是透過NBIoT回傳數據推估洗車用水量，兩處工地平均用水量分別約在686~691L/min(建築工程)、1086.7~1096.7L/min(疏濬工程)，另車輛經過洗車台時需要清洗30秒時間車輛才會乾淨。至於廟會繞境空污事件監控方面，已搭配環保局執行12場次廟會聯合稽查小組作業，另透過微感器完成6場次廟會遶境活動空品分析，此外配合環保局首創廟會AI監控，本計畫額外於富驛時尚酒店頂樓裝設監視器進行廟會煙火監控，並與本市空品監測網進行資料架接，當有微感器PM2.5濃度大於30μg/m3時，鏡頭會自動轉向該顆方向拍攝與辨識。目前已建立寺廟資訊、AI偵測與告警模式；廟會AI辨識系統於8月14日永華宮及10月14日興濟宮遶境活動共有13筆微感器高值告警，其中3筆微感器告警高值為廟會鞭炮引起；中秋、國慶連假期間AI辨識系統自動偵測到16件煙火施放，經分析比對未影響空品且無民眾陳情。另外有關火災事故空污事件監控方面，運用本市1,400個固定式微感器及連續監測特性導入於火警等空污事件監測，計畫期間共協助執行火災應變作業50件次，FB發布21件次、跨局處通報19件次、LINE發布6件次、跑馬燈發布7件次。當中以工廠火警所佔案件最多，並且於112年06月18日永康區雙O有限公司工廠火警派員攜帶MOT設備進行現場空品監控與數據蒐集，以分析有無影響空氣品質。

中文關鍵字

112年度臺南市逸散源空污AI科技監控計畫、工地AI科技監控、廟會AI監控、火災事故空污事件監控

類型	檔名	檔案大小	說明	最新版公開日期
期末報告	112年AI期末定稿本.pdf	88MB		2024/2/11 上午 12:00:00

摘要	During the implementation period of the project (from 13th January to 15th November 2023), AI monitoring equipment was installed at six construction sites, and intelligent surveillance of air pollution prevention equipment was placed at two locations. In terms of AI monitoring equipment at the construction sites, there were a total of 25,317 vehicle detections from April to November, with the accuracy of license plate recognition of 74.88% and the accuracy of dust screen coverage of 97.0%. The dust detection function was initially affected by environmental interference, dirty camera lenses, water stains, and other factors, which led to significant differences between manual and automatic error detection. After optimization and adjustments, the accuracy of the dust monitoring alert has now increased to 95%. The initiative to stop uncovered vehicles at the construction sites has improved. Compared with the initial stage of installation, autonomous watering has also increased. For example, among the dust alert cases in September, 90% of reported cases had already carried out water spraying as a proactive dust control measure upon notification at the construction sites. Regarding intelligent surveillance, NBIoT was used to transmit data for estimating water consumption in car washing. The average water consumption of the construction site was approximately 686-691 L/min, and that of the dredging site was about 1086.7-1096.7 L/min. Besides, it takes about 30 seconds to clean the vehicles when vehicles pass through the car wash palace. For monitoring air pollution during temple procession events, the inspection team has cooperated with the Environmental Protection Bureau to execute twelve joint inspections at temple festivals and conducted air quality analyses with microsensors during six temple procession events. Utilizing AI surveillance during temple festivals was initiated by the Environmental Protection Bureau. An AI surveillance camera integrating with the Tainan air quality monitoring network has been installed on the rooftop of Fu Yi Shi Shang Hotel to monitor temple fireworks. The lens will automatically turn to that direction to shoot and identify when PM2.5 concentrations of microsensor exceed 30μg/m3. Currently, the system has established temple information, AI detection, and alert modes. During procession events on August 14th at Yonghua Temple and on October 14th at Xing Ji Temple, there were 13 high-value alerts of microsensors and three of them were due to firecrackers at the festivals. During the Mid-Autumn Festival and National Day holidays, the AI recognition system automatically detected 16 instances of fireworks being set off. After analysis and comparison, it was found that these activities showed no impact on air quality and there were no complaints from the public. Concerning monitoring air pollution related to fire accidents, the project has used 1,400 fixed microsensors with continuous monitoring characteristics in fire alarms and other air pollution incidents. The assistance for fire emergency response operations was 50 times during the implementation of the project. The notification was released through Facebook 21 times, interdepartmental notifications 19 times, LINE 6 times, and scrolling displays 7 times. Factory fires constituted the majority of cases, and during a factory fire in the O O Limited Company in Yongkang District on 18th June 2023, personnel were dispatched with MOT equipment to conduct on-site air quality monitoring and data collection to analyze its impact on air quality.
英文關鍵字	During the implementation period of the project (from 13th January to 15th November 2023)

摘要

During the implementation period of the project (from 13th January to 15th November 2023), AI monitoring equipment was installed at six construction sites, and intelligent surveillance of air pollution prevention equipment was placed at two locations. In terms of AI monitoring equipment at the construction sites, there were a total of 25,317 vehicle detections from April to November, with the accuracy of license plate recognition of 74.88% and the accuracy of dust screen coverage of 97.0%. The dust detection function was initially affected by environmental interference, dirty camera lenses, water stains, and other factors, which led to significant differences between manual and automatic error detection. After optimization and adjustments, the accuracy of the dust monitoring alert has now increased to 95%. The initiative to stop uncovered vehicles at the construction sites has improved. Compared with the initial stage of installation, autonomous watering has also increased. For example, among the dust alert cases in September, 90% of reported cases had already carried out water spraying as a proactive dust control measure upon notification at the construction sites. Regarding intelligent surveillance, NBIoT was used to transmit data for estimating water consumption in car washing. The average water consumption of the construction site was approximately 686-691 L/min, and that of the dredging site was about 1086.7-1096.7 L/min. Besides, it takes about 30 seconds to clean the vehicles when vehicles pass through the car wash palace. For monitoring air pollution during temple procession events, the inspection team has cooperated with the Environmental Protection Bureau to execute twelve joint inspections at temple festivals and conducted air quality analyses with microsensors during six temple procession events. Utilizing AI surveillance during temple festivals was initiated by the Environmental Protection Bureau. An AI surveillance camera integrating with the Tainan air quality monitoring network has been installed on the rooftop of Fu Yi Shi Shang Hotel to monitor temple fireworks. The lens will automatically turn to that direction to shoot and identify when PM2.5 concentrations of microsensor exceed 30μg/m3. Currently, the system has established temple information, AI detection, and alert modes. During procession events on August 14th at Yonghua Temple and on October 14th at Xing Ji Temple, there were 13 high-value alerts of microsensors and three of them were due to firecrackers at the festivals. During the Mid-Autumn Festival and National Day holidays, the AI recognition system automatically detected 16 instances of fireworks being set off. After analysis and comparison, it was found that these activities showed no impact on air quality and there were no complaints from the public. Concerning monitoring air pollution related to fire accidents, the project has used 1,400 fixed microsensors with continuous monitoring characteristics in fire alarms and other air pollution incidents. The assistance for fire emergency response operations was 50 times during the implementation of the project. The notification was released through Facebook 21 times, interdepartmental notifications 19 times, LINE 6 times, and scrolling displays 7 times. Factory fires constituted the majority of cases, and during a factory fire in the O O Limited Company in Yongkang District on 18th June 2023, personnel were dispatched with MOT equipment to conduct on-site air quality monitoring and data collection to analyze its impact on air quality.

英文關鍵字

During the implementation period of the project (from 13th January to 15th November 2023)