| 英文摘要 |
The primary counties and cities nationwide that have implemented image recognition to monitor stationary air pollution source are Taichung and Kaohsiung. According to the Environmental Protection Bureau (EPB), in practice, image recognition is easily affected by external environmental interferences, leading to misjudgments. Additionally, the inability to effectively distinguish between white smoke and steam through images remains an issue to be addressed. Therefore, the Ministry of Environment aims to enhance the recognition capabilities of the monitoring system through this project, assisting in improving the functionality and application scope of the cloud-based intelligent image monitoring system.
In Taichung, misjudgments are primarily caused by hot air, while in Kaohsiung, they are caused by lighting. The application of this project’s results increased the smoke recognition accuracy rate in Taichung from 68.31% in 2021 to 91.96% in 2023, and in Kaohsiung from 83.36% in 2022 to 89.23% in 2023, significantly improving the accuracy of smoke recognition. Furthermore, this project has conducted a feasibility analysis on distinguishing between white smoke and steam through environmental and process data, achieving preliminary results.
Another important achievement of this project is assisting the Ministry of Environment in promoting intelligent image smoke recognition technology for law enforcement. Our team, in collaboration with the Industrial Technology Research Institute (ITRI), developed an intelligent image smoke recognition law enforcement system using a smoke opacity recognition method. We further integrated digital signature technology to ensure the rigor of the law enforcement results. During the period, we conducted three indoor calibrations and six outdoor verification operations in accordance with the Ministry of Environment's requirements. The system met the technological law enforcement standards recognized by the ITRI team, assisting the Ministry of Environment in establishing the "Method for Testing Opacity of Particulate Matter in Emission Pipes - Image Determination Method (NIEA A221.70B)." In 2023, this project’s intelligent outcomes helped the organizing unit win the 2024 Smart City Innovation Application Award. Subsequently, we organized six promotional seminars, demonstrating the equipment operation and detection report generation on-site. From May 23 to June 17, 2023, a demonstration setup was conducted at the Throne Building in Kaohsiung Park, testing technological law enforcement on fixed-source chimneys. During this period, the system automatically generated 29 detection reports, serving as references for environmental protection units to issue citations.
|