基於機器視覺之車牌辨識與車輛追蹤

License Plate Recognition and Vehicle Tracking Based on Computer Vision

李佳謙(Jia-Cian Li)

指導教授 ： 丁肇隆

共同指導教授 ： 張瑞益

國立臺灣大學/工學院/工程科學及海洋工程學研究所/碩士(2017年)

https://doi.org/10.6342/NTU201702836

摘要

車牌辨識對於治安維護以及車輛管理、追蹤有著相當的必要性，隨著科技蓬勃發展，高畫質的監視器非常普及。本研究的目的是希望藉由這些路口的監視器，發展出一套車牌定位、車牌辨識與車牌追蹤的系統，我們以多個路口固定式攝影機的方式拍攝行駛中車輛，利用各個路口拍攝的影片，追蹤出特定車輛，計算車速，預估車輛到達每一個路口的時間，在預估的時間進行車輛比對與車輛追蹤，規劃出行徑路線，進而追蹤犯人動向。經由大量的實驗測試本研究方法，得知，白天車牌定位準確度達99%，夜間車牌定位準確度接近98%，車牌字符辨識率87%，車輛特徵比對無失誤。

關鍵字

連通體 ； 車牌定位 ； CNN字元辨識 ； SURF特徵比對 ； 車輛追蹤

並列摘要

The recognition of license plates in the field of traffic management is necessary for managing the overall cars number and tracking the cars regarding public security. As the development of technology flourished, cameras with high quality resolution becomes common among society. The goal of this research is to use the monitors on the crossroad to build a system that can locate, recognize, and track the car licenses. We used the fixed camera to record the videos of the driving cars and calculated the velocity of the car to estimate the time each car reaching the next crossroad, then make the car pairing and car tracking according to the estimated time. We can then predict the route of this car and track the moving direction of the criminals. By large amount of the testing data from the research method we concluded that the accuracy of the car locating is at least 99% during the day and 98% during the night, the accuracy of the license recognition is 87%, and the car pairing is totally correct.

並列關鍵字

image connectivity ； license locating ； CNN number recognition ； SURF features pairing ； cars tracking

參考文獻

[24] 陳韋成,"工地安全監控系統,"國立台灣大學工程科學及海洋工程學系碩士論文,2015.

[4] Ashourian, M., et al. "Real time implementation of a license plate location recognition system based on adaptive morphology." International Journal of Engineering-Transactions B: Applications 26.11 (2013): 1347-1356.

[5] Xu, Hongke, et al. "Detection of vehicle beating trajectories using color license plate location." Control and Decision Conference (CCDC), 2010 Chinese. IEEE, 2010.

[6] Wang, Yuh-Rau, Wei-Hung Lin, and Shi-Jinn Horng. "A sliding window technique for efficient license plate localization based on discrete wavelet transform." Expert Systems with Applications 38.4 (2011): 3142-3146.

[7] Caner, Hakan, H. Selcuk Gecim, and Ali Ziya Alkar. "Efficient embedded neural-network-based license plate recognition system." IEEE Transactions on Vehicular Technology 57.5 (2008): 2675-2683.