車牌辨識結合最快路徑規劃於車輛追蹤(以涉案車輛監控查緝網為例)

License Plate Recognition with the Fastest Path Planning in the Vehicle Tracking(An Example for the Involved Vehicles of Monitoring and Investigation Network)

林永弘(Yung-Hung Lin)

指導教授 ： 林芳昌

朝陽科技大學/資訊學院/資訊管理系/碩士(2012年)

摘要

現今汽、機車的普及率高，若要對個別車輛進行商業上或治安上的監控與管理時，可以運用車輛追蹤技術。在本研究中以涉案車輛監控查緝網為例，提出改善現今路口贓車辨識系統的流程，如果要以車牌來追蹤贓車，車牌辨識正是整個系統成敗的關鍵。本研究對於車牌辨識系統提出階段式倒傳遞類神經網路改善其相似字元辨識錯誤的問題，以達到增加車牌辨識率、提高贓車追蹤的成效，並藉由Police Navigator結合舊有的警車定位系統，使用最快路徑規劃產生一條路徑讓警車以最快的速度抵達現場追緝贓車。實驗結果顯示車牌字元辨識的結果為車牌字元辨識正確率97.98%、車牌辨識正確率為89%，與單一倒傳遞類神經網路比較可提升10%車牌辨識率；在路徑規劃實驗中證實本研究提出的最快路徑規劃所花費的CPU執行時間比Dijkstra’s演算法少，且旅行時間近似於全域最佳解。

關鍵字

階段式倒傳遞類神經網路 ； 車牌辨識 ； 涉案車輛監控查緝網 ； 最快路徑規劃

並列摘要

Currently, car and motorcycle penetration is high in Taiwan. To use vehicle tracking technology is needed for monitoring and management individual vehicles in the commerce or concerned security. In this study, based on the case of Involved Vehicles of Monitoring and Investigation Network, this thesis proposes a scheme to improve the current process to Intersection Stolen Vehicle Recognition Systems. Focusing on two parts in the system, one is License Plate Recognition and the other is Police Navigator. The License Plate Recognition (LPR) is a major technique to a successful system for Involved Vehicles of Monitoring and Investigation Network. Multiple-stage Back Propagation Neural Network (Multiple-stage BPNN) is proposed to resolve the wrong reorganization in the similar characters. On the other hand, Police Navigator with Fastest Path Planning is proposed to produce a path to reach the stolen vehicle. Finally, the experiments are given to prove the proposed scheme. Results show the performance of Multiple-stage BPNN is better 10% compared with Single BPNN. Besides, the license plate character recognition accuracy rate is 97.98 and the license plate accuracy rate is 89%. Also, the proposed Fastest Path Planning could have a less CPU time than that with Dijkstra''s algorithm.

並列關鍵字

Fastest Path Planning ； Multiple-Stage BPNN ； License Plate Recognition ； Involved Vehicles of Monitoring and Investigatio

參考文獻

[10]張財榮、賴建庭(2007)，「混合類神經網路應用於影像車號辨識」，資訊科技國際期刊，第1 卷第1期，第100-111頁。

[8]林宗明(2007)，RFID技術應用於停車管理之規劃，碩士論文，國立中央大學通訊工程研究所，桃園。

[12]張傑(2009)，以改良的A*演算法規劃較佳導引路徑之研究，碩士論文，大同大學資訊工程系，台北。

[21]魏銪志(2000)，動態多標的車牌辨識系統之研究，碩士論文，元智大學資訊研究所，桃園。

[23]E. W. Dijkstra(1959), “A Node on Two Problems in Connexion with Graphs,” Numerische Mathematik, pp. 269-271.