Research On Vehicle Location And Motion State Recognition Based On Dual Sensor Data Fusion

At present,China's expressways are developing rapidly,and mileage is increasing.Due to large regional differences and meteorological conditions,road traffic safety hazards are increased.Especially under low visibility conditions,traffic accidents occur frequently,and the difficulty of detecting traffic incidents increases.Therefore,real-time and accurate collection of road dynamics and diverse traffic information is needed to provide a basis for traffic incident detection and induction schemes.So far,traffic information collection mainly includes induction coils,video,microwave,ultrasonic,infrared,GPS positioning,and electronic tags.The current detection technology has problems such as high price,complicated system,damage to the pavement structure,and poor visibility and low environmental conditions.The microwave has the characteristics of sensitive to moving objects and fast detection speed;audio is also one of the traffic information detection technologies,and the spontaneous sound of the vehicle is a manifestation of the driving state of the vehicle.The motion state of the vehicle has the dual characteristics of movement and spontaneous sound.The use of audio and microwave to detect the motion state of road vehicles has the advantages of strong complementarity,convenient installation,no damage to the pavement structure,and good performance in low visibility environment.Therefore,this paper adopts two kinds of microwave and audio acquisition technologies to realize the collection of driving vehicle information,and realizes the vehicle positioning and motion state recognition research through the collected vehicle information.In this paper,the pulse signal of microwave is used to realize the research and analysis of vehicle motion state.The acquisition of road-tracked position detection by vehicle is carried out by collecting tagged audio data.The problem of high false detection rate in vehicle detection for single microwave and audio is utilized.Microwave+audio realizes vehicle detection research based on voting layer fusion.The main research work of this paper includes the following:(1)Firstly,the method of detecting the state-related parameters of the moving vehicle is analyzed by microwave sensing.A method based on two-way microwave sensing to detect the distance between the vehicle and the shoulder is proposed.The vehicle motion state detection system is constructed and related to the actual and simulated environment.The experiment verified the detection method and achieved good detection results in the experimental and simulation environments.(2)Secondly,by collecting the sound pickups on the road shoulders,collecting the audio and video data passed by the vehicles,marking the audio by means of artificial annotation in the early stage,preprocessing the extracted audio data of different lanes,and passing the Meer frequency cepstrum coefficient(Mel Frequency Cepstrum Coefficient,MFCC)extracts the characteristics of audio data,and proposes a screening method based on MFCC feature data.Finally,according to the single source no MFCC screening method,the single source MFCC screening method,the dual source MFCC screening method and the dual source MFCC screening method,the Long Short Term Memory(LSTM)network model is used for learning training.It is indicated that the dual-source MFCC screening method has the best recognition result,and the recognition rate reaches 86.5%.(3)Finally,according to the characteristics of the vehicle with both dynamic characteristics and vocalization,combined with the complementarity between microwave and audio,a data fusion method based on voting layer for microwave and audio is proposed.Through experimental simulation and comparative analysis,fusion Compared with a single detection method,the algorithm improves the accuracy of detection and reduces the rate of false detection.