Research On The Technology Of Passive Acoustics Detection Of Expressway Traffic Flow Characteristic Parameters

Expressways play an important roll in comprehensive traffic and transportation system in China, and contribute to the national and social progress. But some problems such as traffic jam, traffic accident and environment pollution still exist. How to take advantage modern science and technology to manage expressways scientifically and efficiently, establish mordent expressway system, and insure expressway safety and expedite become the pressing issues. At present, China is engage in research on ITS (Intelligent Transportation System) actively, and looks forward to realize systematic and information expressway management. Traffic flow characteristic parameters are radical in all traffic information of ITS. Detected datum of traffic flow characteristic parameters are the most scientific and objective foundation that traffic manage department institute policy, adopt measure and plan and design the traffic infrastructure. This dissertation supported by national science fund item which is research on direct detection technology of expressway traffic and traffic flow characteristic parameters based on acoustic spectrum of vehicles, combining with the development of detection technology of traffic flow characteristic parameters home and abroad, bring forward detection technology of expressway traffic flow characteristic parameters based on acoustics feature of vehicles. Aiming at application in practice, this dissertation did extensive research. This dissertation analyzed the basic function that commonly vehicle detector should possess, realized vehicle detection and counting, speed measurement, automatic vehicle classification and automatic traffic accident detection. This dissertation analyzed the yawp of vehicle and ground vibration signals, makes sure the frequency spectrum of the signals. Based on the character and characteristic, this dissertation designed and constructed experimental acoustics detection system. This system adopts commercial element, not only save time but also decline the cost. It can realize four channels synchronization signals collection, not only achieve Multi-Data Fusion recognition, but also settle base for array signal processing farther. It has large function development. This dissertation studied signal processing technology and pattern recognition theory and made sure use wave-let and AR model to extract the feature of collected acoustics signals. For the feature of collision signals mostly centralized in high frequency, adopt wave-let package to extract the feature of signals. Based on energy detection, bring forward vehicle detection by combining short energy spectrum and short zero-pass ratio. This method ensure system can automatic detect the jumping-off of the signals. Based on anglicizing the theory of hardware counter, this dissertation designed software counter program. In utilizing in practice, the hardware counter has the virtues such as simplicity and steady function. A difficult problem is in the face of us when measure the speed of vehicle by classical measurement method. Because an acoustics signals contains a great number of datum, it is impossible and infeasible that find the time delay between two points. This dissertation employed MSE (Mean Square Error) to compute time delay and gained an ideal result. For the shortcoming that detection area is too long, this dissertation exist in this method, this dissertation bring forward determinant the speed of vehicle by the feature of acoustics, also carry through pilot study on feasibility. This dissertation studied the technology of automatic vehicle classification by passive acoustics signals. Use wave-let coefficients and AR model parameters as recognition characteristics, designed the BP-NN, applied ameliorated neural net technology, increase the speed of training. On the condition of much featuredimension, select the feature by hypothesis testing and rough sets theory. The results indicate that rough sets theory is superior to hypothesis testing when the feature dimension is enormous. This dissertation brings forward a fire-new technology of traffic accident directly automatic detection, which is differ from the principle of judge whether happen traffic accident by arithmetic ago. It is based on microcosmic mechanism of traffic accident, and based on the objective fact that there is a difference of the yawp and the vibration of bodywork between accident and normal, adopt PR method, abstract the feature of collected abnormal signals, contrast to normal signals, achieve traffic accident automatic detection. It enriches the technology of expressway traffic incident. It can be applied to traffic accident automatic alarming in city road cross and little flux road. The detector can be fixed over the road in city road cross. If be built in a vehicle, it can automatic alarm when traffic accident happen. Combining long-distance communication technology and GPS technology, it can carry out automatic alarm; provide necessary information for emergent succor. On the condition of that it is difficult and cost much to collect some data, this dissertation adopted the method of combining physic simulation experiment and real test experiment.It is proved by practice that this scientific research thinking is feasible.