感应线圈信号曲线的处理方法及其在车型识别中的应用

为提高车型识别的准确率，提出信号处理的方法：首先采用快速小波变换抽取原始曲线的近似轮廓，从而削弱了噪声等细节信息的干扰，并实现2^n倍的数据压缩；其次利用五点三次滤波器对上一步获得的近似曲线进行平滑，使得离散信号更接近于真实的连续信号；在此基础上提出基于峰值时间差的强鲁棒性车长检测方法；最后，根据形态匹配的要求，采用Akima插值法进行信号尺度的归一化处理。车型识别试验采用观测点的实测数据，并将结果与其他文献进行比较，说明所提出的方法的有效性。

Processing Methods of Signal Trace in Inductive Loop Detector and Its Application in Vehicle Recognition

Some methods of signal processing were proposed for improving the precision of vehicle recognition.First,fast wavelet transform(FWT) which realizes 2n compression ratio was employed to extract the approximate profile of original traces to reduce the molesting of detailed information,such as noises.Second,5-point-3-time filter which makes the discrete signal series close to the real continuous one was used to smooth the approximate trace obtained by FWT.Furthermore,robust vehicle-length detection which adopt...