基於低成本懸浮微粒感測器與影像辨識車流資料之交通相關空氣汙染物分析

Abstract

與交通相關的空氣污染 (TRAP) 由氣體和顆粒物的混合物組成，包括 PM2.5、PM1.0、超細懸浮微粒 (UFP)、黑碳 (BC)、一氧化碳 (CO) 和二氧化氮 (NO2) 等，主要由車輛排放。本研究旨在分析影響住宅區道路附近空氣污染程度的因素，並提供住戶監測污染物的低成本方法。利用低成本攝影機和基於影像的車流辨識模型，我們實現了汽車、機車、公車超過 90%，和卡車約66%的交通量辨識準確度。此外，我們還利用監測站的 PM2.5 數據校準了微型感測器，確定將兩個微型感測器與氣象資料結合可產生最佳結果。即使感測器數據有限，模型仍保持高精度（MAPE 17-18%），證實了經濟高效的空氣品質監測的可行性。本研究表明，風速、風向、溫度和濕度顯著影響 TRAP 水平，污染物濃度在強風下降低，並隨氣象條件變化。我們也發現，汽車、機車、公車和卡車的交通量，對不少 TRAP汙染量起著至關重要的作用。而車輛在紅燈處等待也會加劇污染。這些發現強調了微型感測器在廣泛、經濟實惠的空氣品質監測方面的潛力，並為交通繁忙道路附近的居民提供了可行的建議，以減少接觸有害污染物。

Traffic-related air pollution (TRAP) consists of a mixture of gases and particulate matter, including PM2.5, PM1.0, ultrafine particles (UFP), black carbon (BC), carbon monoxide (CO), and nitrogen dioxide (NO2), primarily emitted from vehicles. This study aims to analyze the factors affecting air pollution levels near residential roads and provide low-cost methods for residents to monitor pollutants. By utilizing low-cost cameras and an image-based traffic recognition model, we achieved traffic volume recognition accuracies of over 90% for cars, motorcycles, and buses, and about 66% for trucks. Additionally, we calibrated low-cost PM sensors (LCPMS) using PM2.5 data from monitoring stations, determining that combining two LCPMS with meteorological data yields the best results. Even with limited sensor data, the models maintained good accuracy (MAPE 17-18%), confirming the feasibility of cost-effective air quality monitoring. Our findings indicate that wind speed, wind direction, temperature, and humidity significantly influence TRAP levels, with pollutant concentrations decreasing under strong winds and varying with meteorological conditions. We also found that traffic volumes of cars, motorcycles, buses, and trucks play crucial roles in TRAP levels, with vehicles waiting at red lights exacerbating pollution. These findings underscore the potential of LCPMS for widespread, affordable air quality monitoring and provide actionable insights for residents near high-traffic roads to reduce exposure to harmful pollutants.