Research of vehicle type recognition based on multi-BP networks with D-S fusion

To improve the recognition of vehicle type with BP neural network, overcome the overload of training samples existed in single BP neural network, we try to take more BP neural network corresponding to numerous training samples respectively, and take the networks which are trained to identify the vehicle type. The results of each BP neural network with the D-S method were used to improve the vehicle recognition results. The experimental results show that the method of multi-BP network with D-S fusion has a higher recognition rate than the single BP neural network in the overload of training samples.