Accurate Detection and Recognition of Dirty Vehicle Plate Numbers for High-Speed Applications
Abstract:
This paper presents an online highly accurate system for automatic number plate recognition (ANPR) that can be used as a basis for many real-world ITS applications. The system is designed to deal with unclear vehicle plates, variations in weather and lighting conditions, different traffic situations, and high-speed vehicles. This paper addresses various issues by presenting proper hardware platforms along with real-time, robust, and innovative algorithms. We have collected huge and highly inclusive data sets of Persian license plates for evaluations, comparisons, and improvement of various involved algorithms. The data sets include images that were captured from crossroads, streets, and highways, in day and night, various weather conditions, and different plate clarities. Over these data sets, our system achieves 98.7%, 99.2%, and 97.6% accuracies for plate detection, character segmentation, and plate recognition, respectively. The false alarm rate in plate detection is less than 0.5%. The overall accuracy on the dirty plates portion of our data sets is 91.4%. Our ANPR system has been installed in several locations and has been tested extensively for more than a year. The proposed algorithms for each part of the system are highly robust to lighting changes, size variations, plate clarity, and plate skewness. The system is also independent of the number of plates in captured images. This system has been also tested on three other Iranian data sets and has achieved 100% accuracy in both detection and recognition parts. To show that our ANPR is not language dependent, we have tested our system on available English plates data set and achieved 97% overall accuracy.
Existing System:
An ANPR system consists of three different modules: a) Monochrome/Color cameras, b) IR projector, and c) the processing board. In addition to compatibility of interfaces, each section must be chosen properly for a specific application. In this paper, a detailed exploration on the important parameters of an ANPR module has been done. Most of the key parameters are discussed and a comparison between final cost and the desired accuracy is shown. A monochrome camera is usually used with an IR projector for plate detection and recognition. Monochrome camera sensors are capable of providing higher details and sensitivity compared to color camera sensors. An IR projector increases the contrast between plates characters and plates backgrounds. The IR projector is mostly useful at night to brighten the plates. We also need to synchronize the camera exposure time with the IR projector pulses in order to capture images with clearer plates. IR projector which operates invisibly, is an important alternative to the flash lights.
Proposed System:
A monochrome camera with a synchronous IR projector and a color camera are employed in a multi-purpose industrial ANPR system. The monochrome camera with IR projector is responsible for plate detection during the night or other low illumination conditions. It is worthwhile to note that for the IR projector to be effective the vehicles plates should have been coated with IR reflective materials. The role of IR projectors is also important in detecting dirty plates even in daylight by taking care of the camera exposure time. IR projector power has a close relation with the camera exposure time and the exposure time plays an important role in the final clarity of the vehicles plates. Since vehicles move swiftly, high values of exposure time lead to blurred images while low exposure time values produce dark images. Therefore, it is important to tune the output power of IR projector with respect to the exposure time of the monochrome camera. It is also necessary to have an adaptive procedure to fine-tune the exposure time based on the lighting conditions. Modifying the exposure time is performed in an adaptive procedure that gets its feedback from the thickness of plate characters. Having thin characters is a sign of high ambient light. In this case, we must decrease the exposure time. On the other hand, achieving thick characters shows that the environmental light is low and we must increase the exposure time. The modification steps are dependent on the setup and application and must be found experimentally. For example, at sunrise, sunlight reflects from vehicles that move from east to west. In such cases, exposure time should be lowered down to a value that eliminates the reflections. In Fig. 3, a comparison between fixed and variable exposure time algorithms is demonstrated. Color cameras are needed to provide visual evidences for the violation scenes in order to support the corresponding traffic tickets.
Conclusion:
In this paper, an industrial, robust and reliable ANPR system for high speed applications is proposed. The main advantage of our system is its high detection and recognition accuracies on dirty plates. To achieve reliable evaluations, two new data sets were created and used in this paper: one for violation detection called “Crossroad Data set” and the other for vehicle counting in highways called “Highway Data set.” The accuracies of our system on the Crossroad Data set are 98.7%, 99.2%, and 97.6% for plate detection, character segmentation, and plate recognition parts, respectively. In vehicle counting application, the detection rate and false alarm rate over the Highway data set are 99.1% and 0.5%, respectively.We have tested this system on a publicly available English plate data set as well and achieved an overall accuracy of 97%. The proposed system is compared to many reported ANPR systems from different point of views. By considering the practical aspects, several copies of our ANPR system have been installed in different intersections and highways of Tehran, capital city of Iran. These systems have been tested day and night over a year and presented robust and reliable performances, in different weather conditions, such as rainy, snowy, and dusty. The character recognition part of our system has been tested separately over the mnist data set and achieved 98.5% accuracy, with comparably low computational requirement. The presented techniques, algorithms and parameter setting procedures, along with our data sets and related evaluations, provide a complete set of solutions to issues and challenges involved in incorporating ANPR systems in various ITS applications.
References:
[1] H. Caner, H. S. Gecim, and A. Z. Alkar, “Efficient embedded neuralnetwork- based license plate recognition system,” IEEE Trans. Veh. Technol., vol. 57, no. 5, pp. 2675–2683, Sep. 2008.
[2] S. D. Palmer and O. N. Aharoni, “System for collision prediction and traffic violation detection,” U.S. Patent 20 130 093 895, Apr. 18, 2013.
[3] G. Liu, Z. Ma, Z. Du, and C. Wen, “The calculation method of road travel time based on license plate recognition technology,” in Advances in Information Technology and Education. New York, NY, USA: Springer, 2011, pp. 385–389.
[4] V. Abolghasemi and A. Ahmadyfard, “An edge-based color-aided method for license plate detection,” Image Vis. Comput., vol. 27, no. 8, pp. 1134–1142, Jul. 2009.
[5] B. Hongliang and L. Changping, “A hybrid license plate extraction method based on edge statistics and morphology,” in Proc. IEEE 17th ICPR, 2004, vol. 2, pp. 831–834.
[6] A. Mousa, “Canny edge-detection based vehicle plate recognition,” Int. J. Signal Process., Image Process. Pattern Recognit., vol. 5, no. 3, pp. 1–8, 2012.
[7] W. Gao, X. Zhang, L. Yang, and H. Liu, “An improved Sobel edge detection,” in Proc. IEEE 3rd ICCSIT, 2010, vol. 5, pp. 67–71.
[8] T. D. Duan, D. A. Duc, and T. L. H. Du, “Combining Hough transform and contour algorithm for detecting vehicles’ license-plates,” in Proc. IEEE Int. Symp. Intell. Multimedia, Video Speech Process., 2004, pp. 747–750.
[9] K. Deb, A. Vavilin, and K.-H. Jo, “An efficient method for correcting vehicle license plate tilt,” in Proc. IEEE Int. Conf. GrC, 2010, pp. 127–132.
[10] S. Du, M. Ibrahim, M. Shehata, andW. Badawy, “Automatic License Plate Recognition (ALPR): A state-of-the-art review,” IEEE Trans. Circuits Syst. Video Technol., vol. 23, no. 2, pp. 311–325, Feb. 2013.