Research On Vibration Signal Detection And Recognition Of Traffic Accident

With the rapid development of China’s transportation,the number of vehicles and highway mileage rank first in the world.The endless traffic accidents bring huge losses to the economic and social development,and put forward higher requirements for the ability to deal with traffic accidents.Therefore,it is of great practical significance to study the effective monitoring methods of high-speed traffic accidents for timely guiding the traffic and saving the lives of the injured.This paper analyzes the shortcomings of the existing road monitoring methods,and puts forward a method of accident identification based on the vibration signal generated by vehicle driving.The vibration signal acquisition system is designed and implemented,and the vibration signal recognition algorithm is studied,which realizes the detection and recognition of vehicle driving vibration signal.The main work is as follows:1.Realization of vehicle vibration signal detection system.This paper analyzes the propagation model of vehicle vibration signal,compares the sensors according to the experimental requirements,and designs and realizes the vibration signal acquisition system based on ct1500 l acceleration sensor and data acquisition card.2.Vehicle vibration signal recognition based on time domain statistical characteristics.In view of the non-linear and non-stationary characteristics of vehicle vibration signal,the adaptive decomposition algorithm is used to process the signal.The EEMD,CEEMD,LMD and VMD algorithms are compared in terms of signal decomposition effect and computational complexity.The simulation results show that VMD algorithm is superior to other adaptive decomposition algorithms.The VMD is used to decompose the vehicle vibration signal,and the cross-correlation coefficient is used as the index to screen the components.The principal component of the screened component matrix is extracted by using KICA algorithm to further eliminate the environmental noise.The signal is recombined by the denoised components to highlight the time-domain statistical characteristics of the signal.The differences of eight time-domain statistical parameters,such as maximum value,peak value,kurtosis,pulse,root mean square,margin,mean value and slope,of vibration signal in two states of normal driving and collision are studied.The feature vector based on time-domain statistics is constructed,and the recognition model is established by using support vector mechanism optimized by genetic algorithm,The experimental results show that the accuracy of this method can be 95.5%.3.Vehicle vibration signal recognition based on time-frequency analysis.Based on the generalized S-transform,the time-frequency image of vehicle vibration signal is generated.The hog algorithm is used to extract the features of the time-frequency image,and the vehicle vibration signal recognition based on time-frequency analysis is realized.Four time-frequency analysis methods,STFT,WT,ST and GST,are compared in terms of time-frequency focus.It is proved that GST is superior to other algorithms in processing non-stationary signals.The generalized S-transform is used to process the real vehicle vibration signal to get the time-frequency map.The hog algorithm is used to extract the image features of the time-frequency map to construct the feature vector.The support vector machine based on genetic algorithm is used to build the recognition model.The experimental results show that the recognition accuracy of this method can reach 92.7%.