CNN-based object detectors have achieved great success in recent years. The available detectors adopted horizontal bounding boxes to locate various objects. However, in some unique scenarios, objects such as buildings and vehicles in aerial images may be densely arranged and have apparent orientations. Therefore, some approaches extend the horizontal bounding box to the oriented bounding box to better extract objects, usually carried out by directly regressing the angle or corners. However, this suffers from the discontinuous boundary problem caused by angular periodicity or corner order. In this paper, we propose a simple but efficient oriented object detector based on YOLOv4 architecture. We regress the offset of an object’s front point instead of its angle or corners to avoid the above mentioned problems. In addition, we introduce the intersection over union (IoU) correction factor to make the training process more stable. The experimental results on two public datasets, DOTA and HRSC2016, demonstrate that the proposed method significantly outperforms other methods in terms of detection speed while maintaining high accuracy. In DOTA, our proposed method achieved the highest mAP for the classes with prominent front-side appearances, such as small vehicles, large vehicles, and ships. The highly efficient architecture of YOLOv4 increases more than 25% detection speed compared to the other approaches.

Traffic emission is one of the major contributors to urban PM_2.5, an important environmental health hazard. Estimating roadside PM_2.5 concentration increments (above background levels) due to vehicles would assist in understanding pedestrians’ actual exposures. This work combines PM_2.5 sensing and vehicle detecting to acquire roadside PM_2.5 concentration increments due to vehicles. An automatic traffic analysis system (YOLOv3-tiny-3l) was applied to simultaneously detect and track vehicles with deep learning and traditional optical flow techniques, respectively, from governmental cameras that have low resolutions of only 352 × 240 pixels. Evaluation with 20% of the 2439 manually labeled images from 23 cameras showed that this system has 87% and 84% of the precision and recall rates, respectively, for five types of vehicles, namely, sedan, motorcycle, bus, truck, and trailer. By fusing the research-grade observations from PM_2.5 sensors installed at two roadside locations with vehicle counts from the nearby governmental cameras analyzed by YOLOv3-tiny-3l, roadside PM_2.5 concentration increments due to on-road sedans were estimated to be 0.0027–0.0050 µg/m³. This practical and low-cost method can be further applied in other countries to assess the impacts of vehicles on roadside PM_2.5 concentrations.