Study On Improvement And Parameter Self-tuning Of California Algorithm For Traffic Incident Detection

Traffic incident detection algorithm is the premise and key technology to grasp the abnormal operation of road and road management.California algorithm is the earliest traffic incident detection algorithm,which can describe the actual running state of traffic flow to a certain extent.However,in the complex scene the algorithm false alarm rate is still slightly high,and when the environment changes the algorithm is poor adaptability.Therefore,it is of great theoretical and practical significance to solve these problems to improve the detection effect of traffic events and improve the level of road management.In this paper,The California algorithm model is improved reduce false alarm rate,and a parameter self-tuning method based on intelligent algorithm is proposed to improve algorithm adaptability.On this basis,the concept of detection sensitivity is introduced,and a self-tuning condition based on incident detection performance is discussed.The main research contents include.(1)California algorithm and model improvement.This paper first described the principle and basic process of California algorithm for traffic incident detection,and summarized the current status of the California algorithm.Some experiments were used to further analyze and improve the existing California algorithm model,and an improved multi-parameter California algorithm was proposed.(2)Algorithm parameter self-tuning method.Aiming at the existing research trends and shortcomings,the existing intelligent learning algorithms were compared.And a California algorithm self-tuning method based on intelligent learning algorithm was proposed,which can adjust the parameters offline and online.In order to help establish a more efficient rescue and reduce waste of resources,a detection sensitivity coefficient was proposed based on self-tuning method.(3)Research on an algorithm threshold self-tuning condition based on incident detection performance.An algorithm threshold self-tuning condition based on incident detection performance was proposed for the self-tuning condition of the California algorithm parameter self-tuning method in the setting process.Considering the influence factors of time and space,a self-tuning condition model based on event detection performance was established by means of multivariate statistical analysis.(4)Analysis of Parameter Self-tuning Method.The parameter self-tuning method was added to the California algorithm and the improved one in this paper,and it was applied to some sections of Yuwu Expressway as case analysis.Through analyzing the results of the event detection,the validity and practicability of the method are verified.Compared with the algorithm that did not join the parameter self-tuning,the method proposed in this paper could detect the performance degradation of the algorithm for a certain period of time and adjusted the threshold parameters of the algorithm when the time and space environment changed,so that the detection rate and false positive rate remained at the original level.Therefore,this method can significantly improve the effectiveness and adaptability of the incident detection California algorithm to help build more efficient rescue and to reduce waste of resources.