Vehicle detection in traffic surveillance images is an important approach to obtain vehicle data and rich traffic flow parameters. Deep learning based methods have been widely used in vehicle detection with high accuracy and efficiency. However, deep learning based methods require a large number of manually labeled ground truths (bounding box of each vehicle in each image) to train the Convolutional Neural Networks (CNN). In the modern urban surveillance cameras, there are already many manually labeled ground truths in daytime images for training CNN, while there are little or much less manually labeled ground truths in nighttime images. In this paper, we focus on the research to make maximum usage of labeled daytime images (Source Domain) to help the vehicle detection in unlabeled nighttime images (Target Domain). For this purpose, we propose a new method based on Faster R-CNN with Domain Adaptation (DA) to improve the vehicle detection at nighttime. With the assistance of DA, the domain distribution discrepancy of Source and Target Domains is reduced. We collected a new dataset of 2,200 traffic images (1,200 for daytime and 1,000 for nighttime) of 57,059 vehicles for training and testing CNN. In the experiment, only using the manually labeled ground truths of daytime data, Faster R-CNN obtained 82.84% as F-measure on the nighttime vehicle detection, while the proposed method (Faster R CNN+DA) achieved 86.39% as F-measure on the nighttime vehicle detection. Traffic flow parameter estimations in both free flow and congested traffic conditions are evaluated in daytime and nighttime, and the results turn out to be very encouraging.

CICTP