基於虛擬路口關聯圖之車號過濾與軌跡分析系統

蔡昆樺 ； 黃子峻 ； 楊創發 ； 楊鋒彬

《TANET2019 臺灣網際網路研討會》 2019年 (2019/12) Pp. 366-371

https://doi.org/10.6924/TANET.201909.0066

摘要

近年來路口影像監控系統已經成為智慧城市中不可或缺的一環。警政單位除了持續擴大監視器部署範圍外，更進一步導入車牌辨識與車輛軌跡分析功能，期望降低警方在犯罪線索的查找時間，加快破案速度。然而，由於種種環境因素影響（例如、下雨、光線、刮風等因素），造成大量錯誤辨識資料產生，而這些錯誤資料又無法透過人工方式進行過濾。本研究採用SPARK分散式運算架構，提出基於虛擬路口關聯圖之車號過濾與軌跡分析系統，可藉由虛擬路口關聯圖資訊，從龐大辨識車號資料中（每日約2000萬筆車號資料），自動將錯誤資料過濾，提升辨識資料品質，並進一步提出軌跡分析功能，包含出沒熱區、異車同號與同夥車等分析，可主動分析車輛行車動向，提供可疑車輛預警功能。

關鍵字

影像監控 ； 車號過濾 ； 虛擬路口關聯圖 ； 行車軌跡分析

並列摘要

With the development of video surveillance and image recognition technologies rapidly, and in response to urban transportation, security and other applications, the intelligent video surveillance system (IVS) has become an important part of urban infrastructure. Many cities continue to expand the scope of the deployment of intersection cameras and build IVS to assist the police to accelerate the efficiency of crime investigation. However, faced with a large number of identification data generated daily, auto-filtering out the wrong data is a big challenge. In this paper, it proposes a plate number filtering and trajectory analysis system based on the virtual intersection map. Under the SPARK decentralized computing architecture, the plate number filtering mechanism adopts the virtual intersection map to filter out wrong plate numbers automatically from 20 million data per day, and the trajectory analysis function can convert the plate number data into a series of vehicle tracks and analyzes the suspicious vehicle driving trend information, including the infested hot zone, A /B car and associate car and other analysis information.

並列關鍵字

IVS ； plate number filter ； virtual intersection map ； vehicle trajectory analysis