Real-Time Vehicle Color Recognition Based on YOLO9000
In this paper, we proposed a real-time automated vehicle color recognition method using you look only once (YOLO)9000 object detection for intelligent transportation system applications in smart city. The workflow in our method contains only one step which achieves recognize vehicle colors from original images. The model proposed is trained and fine tuned for vehicle localization and color recognition so that it can be robust under different conditions (e.g., variations in background and lighting). Targeting a more realistic scenario, we introduce a dataset, called VDCR dataset, which collected on access surveillance. This dataset is comprised up of 5216 original images which include ten common colors of vehicles (white, black, red, blue, gray, golden, brown, green, yellow, and orange). In our proposed dataset, our method achieved the recognition rate of 95.47% and test-time for one image is 74.46 ms.
KeywordsVehicle color recognition YOLO9000 Intelligent surveillance
This work is supported by National Natural Science Foundation of China (Project 61471066) and the open project fund (No. 201600017) of the National Key Laboratory of Electromagnetic Environment, China.
- 2.Yang M, Han G, Li X, et al. Vehicle color recognition using monocular camera. In: International conference on wireless communications and signal processing; 2011. p. 1–5.Google Scholar
- 3.Dule E, Gökmen M, Beratoğlu MS. A convenient feature vector construction for vehicle color recognition. 2010.Google Scholar
- 6.Rachmadi RF, Purnama IKE. Vehicle color recognition using convolutional neural network. 2015.Google Scholar
- 9.Redmon J, Farhadi A. YOLO9000: better, faster, stronger. 2016.Google Scholar
- 12.Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. Comput Sci. 2014.Google Scholar
- 13.He K, Zhang X, Ren S, et al. Identity mappings in deep residual networks. 2016:630–45.Google Scholar