Wavelet Analysis And Reinforcing Bar Detecting Of Ground Penetrating Radar Signal

Ground penetrating radar (GPR) is a new non-destructive method for detecting reinforcedconcrete structures. It has lots of advantages of high detecting efficiency, high accuracy, widefrequency channel, strong ability of anti-interference and so on. So GPR is more and moreapplied in engineering structures, urban construction, hydrological reconnaissance, environmentproject, archaeology and so on.But compare to the fast growing of GPR' s application, the development on the signalprocessing is slower. The aim of signal processing is to eliminate noises, pick up usefulparameters and enhance accuracy of object recognition.In the concrete structure, reinforcing bar's probe is a point of radar probe, and the detectionof reinforcing bar diameter can not be solved in a long term. As reflected signal from reinforcingbar is an instantaneous signal, and the wavelet transform is an effective method of instantaneoussignal processing. This paper analyzes the advantages and disadvantages of traditionalcontinuous wavelet transform and discrete wavelet transform; try to make use of stationarywavelet transform to extracting the implicative information of radar signal.After the radar signal is processed by wavelet transform, we can acquire the detailcoefficient which contains information that can reflect the reinforcing bar diameter. We use twomethods to pick up information that can distinguish reinforcing bar diameter. One is usingwavelet coefficients imaging method to find the particular area which expresses the reinforcingbar, getting reinforcing bar diameter using assimilating method, this method obtained betterresults in the certain scope. The other is confirming the impacted area of reinforcing bar bywavelet coefficients, by analyzing the symlets wavelet coefficient matrix, in different time point,some numerical value will occur break by the reflection of reinforcing bar. By confirming thearea of break, we could acquire the reinforcing bar diameter. The result shows that compare tothe method one, method two make a great progress, the discriminator in the reinforcing bardiameter is more accurate, and recognizes scope also more wide.The parameter picked up by this paper offer more abundant and valid parameter to the neuralnetworks, which is our next work. This paper also provided the direction to the next research——artificial neural network identify characteristic parameter automatically.