Ultra-thin asphalt pavement layer interface detection and time delay estimation based on MUSIC Huilin Zhou 1 , Qiang Zhang 1 ,Jian Zhu 2 1 Department of Electronic Information Engineering, Nanchang University, China, 330031 E-mail: zhouhuilin@ncu.edu.cn, zhangqiang581016@126.com 2 Quality Supervision Station,Highway Administration bureau,Jiangxi province,China Abstract—Ground penetrating radar (GPR) is a non- destructive and non-intrusive subsurface investigation equipment which has been widely used in the field of civil engineering. Some challenges which GPR encountered for pavement subsurface inspection are: 1)GPR signal are easily corrupted by noise; 2) commercial GPR whose resolution is depended on system bandwidth, is not suitable for ultra-thin thickness estimation. This paper focus on using super-resolution MUSIC algorithm for ultra-thin pavement layer-interface detection and time-delay estimation compared to threshold detector and match filter.The results based on the simulation data under different conditions and measured GPR data are also presented. keywords: GPR, MUSIC,layer-interface detection,time delay estimation I. I NTRODUCTION Ground Penetrating Radar [1] (GPR) is a new and effective shallow underground detection equipment. Its non-destructive, fast, high efficiency, non-contact with the road surface and can directly obtain the road profile, etc., therefore, it is widely used in the field of highway inspection. How to estimate the thickness of the expressway quickly and efficiently, especially how to estimate of the thickness of the ultra-thin asphalt layer accurately is the most basic and important daily work of highway acceptance and routine maintenance. However, GPR can not reflect the characteristics of the targets directly as the optical imaging decive. Thus how to automatically interpret the collected GPR data is a critical problem. The automatic interpretation of GPR data is mainly using effective pavement layer-interface detection algorithm to estimate the arrival time and amplitude of each reflected signal. Then using the empirical formulus, the dielectric constant, thickness and water content of each layer of pavement can be automatically estimated. Several algorithm for pavement layer-interface au- tomating detection have been proposed. Lahouar, S. [2]explored using the threshold detector and matched detector to estimate the time delay and amplitude of interface reflected signals respectively.And the experimental results shows (1) In the low SNR conditions, the matched detector outperforms the threshold detector;(2) When the SNR increases the performance of the two detectors gradually converge. Threshold and match filter are the simplest methods for pavement layer-interface detection. however, they are only able to resolve two succesive echoes for a BΔτ ≥ 1, where B is GPR operating bandwidth,Δτ is the minimal time shift between two echoes that the processing is able to distin- guish. Assumed the GPR operating bandwidth and medium dielectric constant are 1GHz and 4 respectively, the above mentioned methods can detect layer thickness roughly greater than 7.5cm. That is not suited for the ultra-thin pavement layer- interface detection whose thickness is approximately 2cm. Recently,several high and super resolution algorithms for thin and ultra-thin pavement layer-interface detection and time delay estimation have been proposed. Li and Wu [3] propose an iterative algorithm for time delay estimation, which can offer supperresolution capability up to BΔτ =0.45. Chahine,K. [4]proposes a independent compo- nent anaylsis algorithm for thin pavement thick estimation, and the experimental results shows that under the condition of BΔτ =0.45, SNR=0.35dB, the relative error is 0.9%. Bastard, C.L. proposes three superresolution methods(root- MUSIC, root-Min-Norm, ESPRIT) for thin pavement thick- ness estimation, experimental results show that these algorithm can resolve overlapping echos whose BΔτ can up to 0.1. However, these high and super resolution algorithm are based on the simulation data or GPR measured data under ideal condition. This paper aims at using MUSIC methods for ultra-thin pavement layer-interface detection and time-delay estimation based on simulation and time-domain impulse vehicle vehicle- launched GPR raw data without any preprocessing. II. METHODOLOGY A. GPR receiving signal model [2] According to Labour. S, the layer-interface reflected wave- form dont significantly changed as the incident pulse shapes. So, the received signal are simply the time-shifted copies of the transmitted signal e(t), the GPR received signal can be modeled as follows r(t)= K  k=1 s k e(t - T k )+ ω(t) (1) where e(t) is the GPR transmitting pulse, T k is the travel time of the kth interface reflected signal to recceiving antenna, rep- resents the kth layer-interface Reflection coefficient depending on the dielectric constant contrast on successive layer,ω(t) is an additive Gaussian white noise. The focus of this paper is to detect the layer-interface reflecting signal using MUSIC method and estimate the time delay of sucessive layer through τ k = T k+1 - T k . 978-1-4244-3709-2/10/$25.00 ©2010 IEEE