The SIMCA algorithm for processing ground penetrating radar data and its practical applications

Sengodan, Anand (2013) The SIMCA algorithm for processing ground penetrating radar data and its practical applications. PhD thesis, University of Glasgow.

Full text available as:
[thumbnail of 2012sengodan2phd.pdf] PDF
Download (42MB)
Printed Thesis Information:


The main objective of this thesis is to present a new image processing technique to improve the detectability of buried objects such as landmines using Ground Penetrating Radar (GPR). The main challenge of GPR based landmine detection is to have an accurate image analysis method that is
capable of reducing false alarms. However an accurate image relies on having sufficient spatial resolution in the received signal. An Antipersonnel mine (APM) can have a diameter as little as 2cm, whereas many soils have very high attenuation at frequencies above 450 MHz.
In order to solve the detection problem, a system level analysis of the issues involved with the recognition of landmines using image reconstruction is required. The thesis illustrates the development of a novel technique called the SIMCA (“SIMulated Correlation Algorithm”) based on area
or volume correlation between the trace that would be returned by an ideal point reflector in the soil conditions at the site (obtained using the realistic simulation of Maxwell’s equations) and the actual trace. During an initialization phase, SIMCA carries out radar simulation using the system parameters of the radar and the soil properties.
Then SIMCA takes the raw data as the radar is scanned over the ground and uses a clutter removal technique to remove various unwanted signals of clutter such as cross talk, initial ground reflection and antenna ringing. The trace which would be returned by a target under these conditions is then used to form a correlation kernel using a GPR simulator. The 2D GPR scan (B scan), formed by abutting successive time-amplitude plots taken from different spatial positions as column vectors,is then correlated with the kernel using the Pearson correlation coefficient resulting in a correlated image which is brightest at points most similar to the canonical target. This image is then raised to an odd power >2 to enhance the target/background separation.
The first part of the thesis presents a 2-dimensional technique using the B scans which have been produced as a result of correlating the clutter removed radargram (’B scan’) with the kernel produced from the simulation. In order to validate the SIMCA 2D algorithm, qualitative evidence was used where comparison was made between the B scans produced by the SIMCA algorithm with B scans from some other techniques which are the best alternative systems reported in the open literature. It was found from this that the SIMCA algorithm clearly produces clearer B scans in
comparison to the other techniques.
Next quantitative evidence was used to validate the SIMCA algorithm and demonstrate that it produced clear images. Two methods are used to obtain this quantitative evidence. In the first method an expert GPR user and 4 other general users are used to predict the location of landmines
from the correlated B scans and validate the SIMCA 2D algorithm. Here human users are asked to indicate the location of targets from a printed sheet of paper which shows the correlated B scans produced by the SIMCA algorithm after some training, bearing in mind that it is a blind test. For the second quantitative evidence method, the AMIRA software is used to obtain values of the burial depth and position of the target in the x direction and hence validate the SIMCA 2D algorithm.
Then the absolute error values for the burial depth along with the absolute error values for the position in the x direction obtained from the SIMCA algorithm and the Scheers et al’s algorithm when compared to the corresponding ground truth values were calculated.
Two-dimensional techniques that use B scans do not give accurate information on the shape and dimensions of the buried target, in comparison to 3D techniques that use 3D data (’C scans’). As a result the next part of the thesis presents a 3-dimensional technique. The equivalent 3D kernel is formed by rotating the 2D kernel produced by the simulation along the polar co-ordinates, whilst
the 3D data is the clutter removed C scan. Then volume correlation is performed between the intersecting parts of the kernel and the data. This data is used to create iso-surfaces of the slices raised to an odd power > 2.
To validate the algorithm an objective validation process which compares the actual target volume to that produced by the re-construction process is used. The SIMCA 3D technique and the Scheers et al’s (the best alternative system reported in the open literature) technique are used to
image a variety of landmines using GPR scans. The types of mines included plastic, wooden and glass ones. In all cases clear images were obtained with SIMCA. In contrast Scheers’ algorithm, the present state-of-the-art, failed to provide clear images of non metallic landmines.
For this thesis, the above algorithms have been tested for landmine data and for locating foundations in demolished buildings and to validate and demonstrate that the SIMCA algorithms are better than existing technologies such as the Scheers et al’s method and the REFLEXW commercial software.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Keywords: Ground Penetrating Radar (GPR), area correlation, volume correlation, iso-surfaces, Finite-difference-time-domain (FDTD) simulation
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
U Military Science > U Military Science (General)
Colleges/Schools: College of Science and Engineering > School of Computing Science
Supervisor's Name: Cockshott, Dr. W. Paul
Date of Award: 2013
Depositing User: Mr. Anand Sengodan
Unique ID: glathesis:2013-4177
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 12 Apr 2013 08:31
Last Modified: 23 Apr 2013 15:15

Actions (login required)

View Item View Item


Downloads per month over past year