基于BP神经网络的行人和自行车交通识别方法

A Study on Pedestrian and Cyclist Recognition Based on BP Neural Network

研究了基于BP神经网络的行人和自行车识别方法.首先对图像提取4个特征,形成特征向量作为BP神经网络的输入;然后设计BP神经网络的结构,网络输出为对行人和自行车的识别;为了确定BP神经网络合理的隐层神经元数目,分别对不同隐层神经元数目的神经网络进行了实验分析.最后利用实测的数据对BP神经网络进行训练、仿真实验,并对实验结果进行分析;结果表明:最佳网络的正确识别率为84%,行人和自行车的正确识别率分别为89%和71%.

A study on the pedestrian and cyclist recognition based on the backpropagation（BP） neural network is presented in this paper. The binary image of moving object contour is processed by the method presented here. The method first draws four features from the binary image and forms the feature vector as the input of BP neural network. The output of BP neural network is the recognition of pedestrians and cyclists. Secondly, the structure of BP neural network is designed. In order to obtain the reasonable number of the hidden layer neuron, the paper performs experiments with the BP neural network with the different number of the hidden layer neuron. Finally, the BP neural network is trained and simulated by using the data of actual measurement and the experiment results are analyzed. For the best BP neural network, the right recognition ratio, the pedestrian right recognition ratio and the cyclist right recognition ratio are 84%, 89% and 71% respectively.