The invention discloses a method for determining a vehicle type and a driving direction, which comprises the following steps of: 1, acquiring a vehicle noise signal on a roadbed beside a public road through a sound recorder, and clockwise or anticlockwise forming an angle of 30-60 degrees from an inner normal of an outer traffic lane adjacent to the sound recorder to a microphone direction of the sound recorder; 2, marking a vehicle driving direction by taking an inner normal of an outer traffic lane adjacent to the sound recorder as a reference, and marking a vehicle type at the same time; 3, calculating the Mel spectrum of the vehicle noise signal, and converting the vehicle noise signal into a noise Mel spectrogram; 4, a data set composed of a plurality of noise Mel spectrograms is constructed, and each noise Mel spectrogram is marked with the vehicle type and the driving direction; 5, selecting training data, verification data and a deep learning neural network; and 6, training the deep learning neural network, and using the trained deep learning neural network to identify the vehicle type and the driving direction. According to the method, a single sound recorder is used for collecting vehicle noise signals, the vehicle type and the driving direction are recognized at the same time, influences of environmental visibility are avoided, and the method is more hidden.
本发明公开了一种确定车辆类型和行驶方向的方法,包括第1步,在公共道路旁边的路基上通过声音记录仪采集车辆噪声信号,声音记录仪近邻的外行车道内法线到声音记录仪的麦克风方向的角度为顺时针或逆时针30°~60°;第2步,以声音记录仪近邻的外行车道内法线为参照标记车辆行驶方向,同时标记车辆类型;第3步,计算车辆噪声信号的梅尔频谱,将车辆噪声信号转换为噪声梅尔频谱图;第4步,构建由若干张噪声梅尔频谱图组成的数据集,每张噪声梅尔频谱图都标记了车辆类型和行驶方向;第5步,选取训练数据、验证数据和深度学习神经网络;第6步,对深度学习神经网络进行训练,将训练后的深度学习神经网络用于识别车辆类型和行驶方向。该方法利用单个声音记录仪采集车辆噪声信号,实现了车辆类型和行驶方向的同时识别,不受环境能见度影响,更具隐蔽性。
Method for determining vehicle type and driving direction
一种确定车辆类型和行驶方向的方法
2024-07-16
Patent
Electronic Resource
Chinese
Vehicle Driving Direction Determination System and Vehicle Driving Direction Determination Method
European Patent Office | 2023
|European Patent Office | 2018
|European Patent Office | 2018
|