摘要
提出一种新颖的基于Boosting RBF神经网络的交通事件检测方法。对Boosting算法进行改进,采用更有效的参数求解方法,即弱分类器的加权参数不但与错误率有关,还与其对正样本的识别能力有关。以上下游的流量和占有率作为特征,将RBF神经网络作为分类器进行交通事件的自动分类与检测。为了进一步提高神经网络的泛化能力,采用Boosting方法进行网络集成。最后运用Matlab进行了仿真分析,结果表明提出的交通事件检测算法利用较少样本数据即可快速实现交通事件检测。
A new method was proposed for traffic incidents detection based on Boosting RBF neural network. The improved Boosting adopted a new method to acquire parameters, and the weighted parameters of weak classifiers were determined not only by the error rates, but also by their abilities to recognize the positive samples. The features of flow and occupancy rate were extracted from traffic incidents. Then RBF neural network was used to classify the traffic incidents. In order to improve the precision of the RBF neural network for traffic incidents detection, Boosting algorithm was used to build an integration-neural network. Finally a simulation using Matlab was carried out, and the results show that this algorithm can detect incidents rapidly by using a few samples.
出处
《计算机应用》
CSCD
北大核心
2007年第12期3105-3107,共3页
journal of Computer Applications
基金
河北省自然科学基金资助项目(F2007000682)
