駕駛安全輔助系統之分析子系統

Dangerous Event Analysis Subsystem of Driver Assist System

邱建富(Chiou Chien-fu)

指導教授 ： 方瓊瑤

國立臺灣師範大學/理學院/資訊工程研究所/碩士(2007年)

摘要

隨著科技的進步、汽車工業的蓬勃發展，許多車載型行車安全輔助系統(Driver assistance system DAS)也因應而生，不管是已普遍裝設的安全氣囊、ABS(antilock braking system)、倒車雷達…等，或是較新穎的酒精濃度偵測器、跨越車道警告器，以及尚處於研發階段的交通標誌標線偵測器。各式各樣的裝置於行車中會適時給予駕駛者特定的訊息，但各裝置的單一運作、各自發出危險警告，不但使駕駛者需分心且反而更容易疲勞。於此，本研究希望能開發一危險行車事件分析系統，目的在整合分析各裝置所收集到的資料，若分析結果本車處於危險情況，便適時給予警告，提醒駕駛注意安全，以避免意外事故的發生。
由於各種新的裝置不斷被開發出來，裝置愈來愈多，提供的資料也愈來愈多，但並非所有的資料都對分析行車危險事件有幫助。因此，本論文使用fuzzy-rough sets技術做特徵維度的降維，在盡可能保留完整資訊的情況下挑選出具代表性的特徵，剔除掉高度相依性或是和危險程度值不相關的特徵，以減少特徵向量的維度，加快分析動作的執行速度，且避免雜訊的干擾，提升分析結果的精確度。
降維後的特徵向量，將依據其代表性和確定性來尋找有關危險程度的推論法則，符合條件的法則皆挑選出來組成一組行車危險事件法則，此組法則即為隱含於原資料中的知識法則。最後，將法則轉成fuzzy Petri nets的形式，於系統上線時進行平行推論分析行車之危險情況。

關鍵字

自動法則選取

並列摘要

To help provide safety for drivers, many driver assistance systems (DAS) have been developed. Various kinds of devices are used in DAS. Some devices have been installed in cars, like air bags, antilock braking system (ABS), and backup radar. Some descriptions of other devices have recently been published, such as the crossing-lane sensor and a alcohol sensor. And some devices are still in research, such as road sign sensors.
Such devices work independently of each other and, as a result, indicate when drivers become distracted from their primary task of driving or are tired. For this reason, this paper proposes a system to integrate the outputs of these devices and to provide a warning to drivers.
First, the dangerous driving event analysis system using fuzzy rough sets to reduce the attributes is presented. Then the system selects the important rules depending on the representation and confirmation of the rules from the reduced data. Finally, the system using fuzzy Petri nets to form the reasoning module from a set of rules we derive determines if there is a danger. Thereby, the driver is warned, and the accident can be prevented.

並列關鍵字

Fuzzy Rough Sets ； Rule Selection ； Fuzzy Petri Nets ； DAS

參考文獻

method for anomaly intrusion detection in computer network systems,” Expert Systems, Vol. 20, pp. 251–259, 2003.

[Chi08] J. H. Chiang and S. H. Ho, “A Combination of Rough-Based Feature Selection and RBF Neural Network for Classification Using Gene Expression Data,” IEEE Transactions on Nanobioscience, Vol. 7, pp. 91-99, 2008.

[Far05] M. D. Farrell, Jr., and R. M. Mersereau, “On the Impact of PCA Dimension Reduction for Hyperspectral Detection of Difficult Targets,” IEEE Geoscience and Remote Sensing Letters, Vol. 2, pp.192-195, 2005.

[Fio07] C. Fiot, A. Laurent, and M. Teisseire, “From Crispness to Fuzziness: Three Algorithms for Soft Sequential Pattern Mining,” IEEE Transactions on Fuzzy Systems, Vol. 15, pp. 1263-1277, 2007.

[Jen04] R. Jensen and Q. Shen, “Semantics-Preserving Dimensionality Reduction: Rough and Fuzzy-Rough-Based Approaches,” IEEE Transactions on Knowledge and Data Engineering, Vol. 16, pp. 1457-1471, 2004.