Kabir2, Ehsan, Lotfizad, mojtaba, Sadoughi Yazdi, Hadi, Fathi, Mahmoud (2007). Determining Individual Normal Behavior of Drivers Using an Interactive Model. , 7(0), 21-38.
Ehsan Kabir2; mojtaba Lotfizad; Hadi Sadoughi Yazdi; Mahmoud Fathi. "Determining Individual Normal Behavior of Drivers Using an Interactive Model". , 7, 0, 2007, 21-38.
Kabir2, Ehsan, Lotfizad, mojtaba, Sadoughi Yazdi, Hadi, Fathi, Mahmoud (2007). 'Determining Individual Normal Behavior of Drivers Using an Interactive Model', , 7(0), pp. 21-38.
Kabir2, Ehsan, Lotfizad, mojtaba, Sadoughi Yazdi, Hadi, Fathi, Mahmoud Determining Individual Normal Behavior of Drivers Using an Interactive Model. , 2007; 7(0): 21-38.

Determining Individual Normal Behavior of Drivers Using an Interactive Model

The Modares Journal of Electrical Engineering
Article 5, Volume 7, Issue 0, 2007, Pages 21-38    PDF (600.65 K)
Authors
Ehsan Kabir21; mojtaba Lotfizad* 2; Hadi Sadoughi Yazdi3; Mahmoud Fathi4
1Tarbiat Modarres Univ
2Tarbiat modares Univ
3Tarbiat Moallem University of Sabzevar
4, Iran University of Science and Technology
Abstract
In this paper, an interactive model for individual normal behaviour of drivers is presented in which the mutual effect of vehicles has been incorporated. Temporal features obtained from vehicles tracking and their motion history is utilized for generating a model of normal behaviour. Because of non-stationarity of behaviour, Hidden Markov Model has been used for interactive model. This model has three main parts. The first part is the history of antecedent trajectory which for this purpose has proposed a Centers Transition Matrix (CTM) that is some type of spatio-temporal information-data bank from motions seen in the old frames. The second part is based on the linguistic features or motion recognition of vehicles, these motions contain forward, turn right and left, lane changing to right and left motion. The third part is constituted from low level features which contain Velocity and distance to neighbor object. Also CTM is efficient in search at similar blob in image sequences and it can determine the radius and region of search. This top-down feedback caused an increment of performance of RLS tracker and object searching. In the presented system, we obtained a 81.2% membership rate to normal model. Also types of motion are recognized using HMM with a recognition rate of up to 82.7%. Prediction error is reduced on many vehicles trajectory by at least 80% using a feedback system.
Keywords
Interactive Model; Hidden Markov Model; Normal Behaviour; Centers Transition Matrix; Motion Recognition; Multi-Object Tracking; RLS Predictor
References
Statistics
Article View: 112
PDF Download: 68
Statistics
Number of Journals45
Number of Issues2,196
Number of Articles24,878
Article View28,797,400
PDF Download18,605,531