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.
