多分支BP网络模型及其在车型分类中的应用

Multi-branch BP Neural Network and Its Application in Vehicle Classification

通过分析一般的BP网络在解决多模式识别问题时的缺陷,本文根据分而治之和模块化的思想提出了多分支BP网络的模型。多分支BP网络可将N种模式的识别问题转换为N个并行的两种模式识别问题。文章还研究了多分支BP网络模型在车型分类中的应用,并利用G107上的实测数据进行仿真,仿真结果表明,与一般的BP网络相比,多分支BP网络具有更好的分类性能,且显著地减少了训练时间。

Based on the analysis of general BP neural network＇ s limitations in multi-class pattern recognition, this paper proposes a multi-branch BP neural network （MBBPNN） according to the divide-and-conquer and modular principle. A N-class problem can be divided into N parallel two-class problems by MBBPNN. The pattern of the sample will be confirmed by a fusion algorithm finally. This paper also studies MBBPNN＇ s application in vehicle classification and does simulation with some actual data collected from national highway G107. The simulation results confirm that MBBPNN will improve the performance and remarkably reduce the training time against the general BP neural network for pattern recognition.