基于图像特征分类和RBF网络的两轴车辆动态称重技术

汽车在行驶过程中的总重信号通常由各轴重信号确定,各轴重信号的测试精度取决于对车辆运动参数和振动信号等的精确分析。本文采用径向基函数(Radial basis function,RBF)网络处理轴重信号,针对该网络的泛化能力与拟合精度的矛盾,将车辆按照重量分为大、中、小3种类型,并进行整车建模和网络训练;实际测试过程中,利用汽车俯视图像提取类型特征,然后根据汽车的类型将测试参数输入不同的神经网络进行处理,以静态测量结果为相对真值。仿真结果表明,分类建模比单一混合建模具有更高的测试精度。

Signal Analysis for Two-Axle Vehicle Weigh-in-Motion Based on RBF and Image Processing

The vehicle weight in motion is decided by the weight distribution to every axis.The measurement accuracy is related to the accurate analysis on motion and vibration factors.The radial basis function(RBF) neural network is used to construct the nonlinear model of the weighing system,including the topological structure and the selection of the RBF center.In allusion to the contravention between wide adaptation and imitating accuracy,vehicles are divided into big,medium and small three types.The model is constructed in whole vehicle.The type of the vehicle is achieved by image processing in which the CCD camera is used to get the platform of the vehicle.The different types of vehicle use corresponding neural network.The weight distribution to every axis of vehicle that passes across the bed-plate in even speed is also analyzed.The static weight signals are used as the relative real value.Simulation result shows that higher precision of measurement can be achieved as long as the vehicle across the bed-plate in longer time and the type can be achieved.