DEVELOPMENT OF DETECTION ENHANCEMENT ALGORITHM FOR GROUND PENETRATING RADAR

Abstract

Detection of buried object is quite interesting and still challenging task. Although, several sensors and techniques have been developed by various researchers but identification of the buried targets with precise information is still need more attention. In this field, some popular technologies, like Metal detector and Ground Penetrating Radar (GPR) are used to detect the buried objects. In these technologies, GPR has more importance than others, because it has greater penetration capabilities than other, during the detection of buried object. The advantage of the GPR is that it can detect buried objects in presence of unpredictable clutter as well as in different terrain. Nowadays, there is an increasing need to explore non-destructive and non-invasive geophysical techniques to investigate about the subsoil characteristics as well as the utilities like drainage pipe, sewage pipe etc. A large number of NDT techniques are available for pipe detection such as pulse induction, resistivity methods, acoustic techniques, as well as magnetic or electromagnetic methods [Prego et al. 2017] but acoustics noise and electrical noise limits the performance of these methods because of the material characteristics of the buried objects whereas GPR has advantage over other techniques because of its data acquisition, accuracy and resolution. Currently, available GPR can be classified by the domain of work and the modulation. Impulse GPR is a time domain GPR, which is the most popular due to its simplicity and low cost, whereas, Frequency Modulated Continuous Wave (FMCW) GPR and Stepped frequency continuous wave (SFCW) GPR works in the frequency domain. In present research work, indigenous developed SFCW GPR has been used for collecting the data at various test sites. Current state of art of data processing of SFCW-GPR uses the classical approach, where collected raw data is converted from frequency domain to time domain by Inverse Discrete Fourier Transform (IDFT) and fixed noise removal by the existing de-noising techniques. Clutter reduction algorithms have been proposed by the researchers time to time for detection of buried objects. More attention needs to be paid on data processing of GPR because dielectric constant of buried objects, soil moisture and site topography limits the detection capabilities of GPR. While processing the GPR data, several challenges were observed, like reflection from earth surface, lossy medium, resolution etc. Since, first reflection is always from ground in GPR, so first challenge in GPR imaging is how to select the appropriate background subtraction techniques to eliminate the ground reflection fully or partially. Most of the researchers have vi used time gating for removing the ground reflection, but useful information may also be eliminated by the time-gating. Therefore, still more research is required to analyze the background on real GPR data. Apart from fixed reflection (i.e. earth surface, cable delay etc.), there may be reflections from some natural objects like rocks, roots etc., due to which false detection occurs. These reflections are known as “Clutter”. Researchers proposed so many techniques such as statistical technique (ANOVA and HANOVA), Adaptive clutter reduction techniques, Singular Value Decomposition (SVD), Principal Component Analysis (PCA), Independent Component Analysis (ICA) etc., but results are satisfactory only for particular conditions. There is much more scope to improve the clutter reduction challenges by developing the suitable algorithm/techniques. Another major challenge is enhancement of detection because low dielectric buried objects like PVC pipe, water bottle etc. have weaker reflection than high dielectric buried objects like metal plate, metal pipe etc. Many researchers had put their effort to enhance the detection capability for GPR imaging. Current popular algorithms are 2D Least Mean Square (LMS), OTSU, Block-Median Pyramidal Transform etc. But these image de-noising methods are applicable only when noise is i.i.d (independent and identically distributed) in nature. In case of GPR, noise distributions are not identical. Therefore, this challenge also requires further research to enhance the detection, especially for low dielectric buried objects. GPR data contaminated by the clutter, noises and lossy soil surfaces, which limits the performance of detection in GPR imaging. Also, low dielectric constant buried objects degrade the detection in the GPR imaging. Therefore, the broad objective of this thesis is “Development of Detection Enhancement Algorithm for GPR”. Further to comply the objective, four important tasks have been attempted in present research, which are as follow:  Critical Review of Background Subtraction Techniques  Development of Adaptive Threshold and Data Smoothening Algorithm for Target Detection for GPR Data  An Efficient Application of ANN for Non-metallic Pipe Detection  Novel Adaptive Buried Non-Metallic Pipe Crack Detection Algorithm For Ground Penetrating Radar Thesis has been divided into seven chapters, which are briefly described as follows: The introduction of the thesis has been depicted in Chapter 1, which includes motivation, major research gaps, and frame work of research, SFCW GPR design and GPR imaging parameters, which have been used in the research work. vii Chapter 2 discusses the brief literature review of related works which gives the idea of current state of the art and challenges. In Chapter 3, study and critical review of background subtraction techniques on real GPR data has been described. Ground penetrating radar (GPR) is used to detect the underground buried objects for civil as well as defence applications under varying conditions of soil moisture content. The capability of detection depends upon soil moisture, target characteristics and subsurface characteristics, which are mainly responsible for contaminating the GPR images with clutter. Researchers earlier have used averaging, mean, median, Eigen values, etc. for subtracting the background from GPR data [Solimene et al. 2013]. To analyze the background subtraction or clutter reduction problems, in this chapter, we have experimentally reviewed background subtraction techniques with or without target conditions to enhance the target detection under variable soil moisture contents. Indigenously developed GPR has been used to collect the data for different soil conditions and several background subtraction signal processing techniques were critically reviewed like, Mean, Median, Singular Value Decomposition (SVD), Principal Component Analysis (PCA), Independent Component Analysis (ICA) and Training methods. The signal to clutter ratio (SCR) measurement has been used for performance evaluation of each technique. The relative merits and demerits of each technique have also been analyzed. The background subtraction techniques have been applied to experimental GPR data and it is observed that in comparison of mean, SVD, median, ICA, PCA, the training method shows the highest SCR with buried target. Finally, this review helps to select the comparatively better background subtraction technique to enhance the detection capability in GPR. Chapter 4 explores the development of adaptive threshold and data smoothening algorithm for target detection for GPR data, which can help to separate out clutter and target and reduce the false detection of target. Clutter reduction is another challenge in the signal processing of GPR. There are many approaches available to separate the background and foreground in image processing applications. Currently, researchers are focusing on wavelet de-noising, curvelet threshold, Edge Histogram Descriptor (EHD) threshold, Otsu thresholding ,recursive thresholding and adaptive progressive thresholding [Valaparla and Asari 2001]. In fixed and predictable background conditions, above techniques separate background and foreground efficiently. In a common scenario, background reference is blind due to soil surface moisture content and its non-linearity. There are many methodologies proposed time to time by researchers to solve this blind reference background separation. But challenges still remains, viii because there are two major problems in ground penetrating radar imaging such as targets like ground enhances the false detection and non-metallic target detection, where the threshold decision is a critical task. In this chapter, a novel real time adaptive threshold algorithm is proposed for ground penetrating radar image processing. The blind threshold was decided to use normal random variable variance and image data variance. Further, the image was smoothened by random variance ratio to image data variance. Experimental results showed satisfactory results for the background separation and smoothening the targeted image data with the proposed algorithm. Chapter 5 provides the efficient application of a neural network based approach for enhancing the detection capability of SF-GPR, especially for detecting the Non-metallic object like PVC pipe, water bottle and Landmines. Detection enhancement in the image processing of GPR is enhanced capability to discriminate the target and clutter. “Buried target may be metallic or non-metallic”, when buried target is non-metallic, then detection stage involves lot of challenges due to low dielectric constant of buried target. Mean subtraction, median removal, singular value decomposition (SVD), Principal component analysis (PCA) and Independent component analysis (ICA) are very popular approaches to extract the buried target information in presence of clutter and background noise for GPR applications. Clutter and background reduction and detection of low dielectric constant buried object with variable soil conditions are the challenging tasks in GPR. But available techniques are not able to extract the non- metallic target information, due to low dielectric constant. Therefore, in this chapter, we have proposed a neural network and statistical mean to standard deviation threshold based approach for subtracting background and for enhancing the detection of low dielectric constant buried object. ANN approach is based on the collection of large amount background data with soil moisture variation. These background data is statistically analysed to compute the mean to standard deviation thresholding. After that, motion filter estimate the actual pixel intensity of PVC pipe in linear manner. The enhanced target detection and background subtraction are achieved directly from proposed trained neural network and quite satisfactory results have been obtained. Chapter 6 deals the capability of image processing of GPR to detect the crack in the buried PVC pipe. For this purpose, a novel adaptive crack detection algorithm has been developed, which is mainly based on the covariance analysis of real GPR data and covariance analysis of the normally distributed synthetic data. As Ground Penetrating Radar (GPR) may be used to detect the cracks in a buried pipe, there are only a few techniques, like statistical ix approach Robust Principal Component Analysis (RPCA) [Kalika et al. 2015] etc. to detect the crack in the buried objects. Buried non-metallic pipe crack detection is an important application of GPR to analyze the structural health of underground pipelines. The strength of the reflected signal may be feeble from a cracked location as compared to position with respect to that from other position of the pipe. Currently, the crack detection is a challenging task, especially when the buried pipe is non-metallic and the soil moisture varies. In order to efficiently detect the crack in a buried PVC pipe, a novel adaptive crack detection algorithm has been developed with the help of covariance of the real GPR data and covariance of normal distributed synthetic Gaussian data. Results are evaluated and validated to show the effectiveness of crack detection algorithm and found that proposed algorithm is quite suitable to detect the crack in buried PVC pipe. Finally, the thesis has been concluded in Chapter 7, summarizing the obtained results and depicting the considerable contributions made in the thesis. The view of future exploration utilizing present results has been conferred.