Region of Interest (RoI) Detection in Ground Penetrating Radar - - PowerPoint PPT Presentation

Region of Interest (RoI) Detection in Ground Penetrating Radar (GPR) Data

2D ENTROPY ANALYSIS

Presenter: Yu Zhang

Autobiography

RESEARCH INTERESTS

2

Autobiography

 University of Vermont

Doctor of Philosophy (Ph.D.) in Electrical Engineering 2012 – Present Advisor: Dr. Tian Xia

 Huazhong University of Science and Technology

Bachelor of Science (B.S.) in Electrical Engineering 2008 – 2012

3

Research Interests

Compressive OFDM GPR

Tx Rx Rebar Concrete 30% Compression 5% Compression Full OFDM Spectrum

4

Research Interests

Synthetic Aperture Radar (SAR) based GPR imaging

Cylinder Triangle Rectangle Test Scenario Regular GPR B-Scan Image GPR-SAR B-Scan Image

5

Research Interests

Low-Rank and Sparse Representation in GPR and Through-the-Wall Radar Imaging

Test Scenario GPR B-Scan Regular Clutter Removal L + S Representation

6

Research Interests

GPR Signal Processing related Problems

 Image De-noising  Clutter Removal  Image Migration  Region of Interest (RoI) Detection

7

Introduction

WHY GPR, GPR OPERATING MECHANISM & OUR GPR SYSTEM

8

Why GPR?

 Non-destructive evaluation (NDE) of transportation

infrastructure.

 GPR as a highly efficient NDE method:

 Concrete bridge deck inspection;  Asphalt pavement monitoring;  Highway rebar detection;  Railroad ballast condition assessment,  Soil moisture estimation.

9

GPR Operating Mechanism

 Subsurface medias of different dielectric constants  Each position: A-Scan trace  Assemble all A-Scan traces: B-Scan image

10

Home made GPR System

 Ultra-wide band (UWB) pulse generator  UWB antenna  High speed digitizer configuration  Wheel encoder  FPGA controller  PC & LabVIEW user interface

11

Prototype

12

Xia & Huston sitting on GPR 13

A ROBUST system package that can hold two adults’ weight!!!

Air-coupled Impulse GPR System Specifications

Data acquisition unit 8 Gsps, 10-bit resolution Sampling window width 40 ns Pulse repetition frequency 0 to 30 kHz tunable Horizontal resolution 1 cm at 60 miles/h survey speed Signal bandwidth 600 MHz to 2 GHz tunable Penetrating capability Up to 1 meter

14

Today’s Topic

ROI DETECTION IN GPR DATA USING 2-D ENTROPY

15

Why 2-D Entropy

 Large volume (overall 20 miles) railroad GPR data set

collected during the field test at Metro St. Louis and Massachusetts Bay Transit Authority.

Data collection in St. Louis Forest Park Station to Sunnen Station

 High resolution GPR system brings us ~300 GB data.  How to process such large volume data???

16

Why 2-D Entropy

 Large data volume

 Sophisticated data processing is too computationally

complex and even infeasible

 Properties of subsurface scatters or material are too

complex

 Obtain of prior knowledge or training data is unrealistic

 It is desirable to

 Develop an unsupervised and automatic GPR data

processing method that can effectively and rapidly identify suspicious features from big radargram

17

First Glance at 2-D Entropy Analysis

 Entropy is a measure of the uncertainty associated

with a random variable.

 Entropy characterization is explored to identify singular

regions within a large GPR data set

 High entropy value indicates high similarity  Low entropy value specifies high singularity

18

What do we have in a GPR railroad B-Scan image

19

Noises Clutter Cross-ties

• r Sleepers

Useless Background Multiple Peaks from

• ne Scattering

Unsupervised GPR RoI Detection based on Entropy

Raw Data Stacking Every 50 Traces Low Pass Filter Hilbert Transform A-Scan Decomposition Region of Interest 2-D Entropy Analysis Stacking Every 10 Trace Clutter Removal Subsurface Identification Pre-processing Power Information Characterization Subsurface Identification B-Scan Image Enhancement RoI Detection

20

Pre-Processing

 Step 1: Stack every 50 A-scan traces to calculate the

average to boost the signal-to-noise ratio (SNR).

• The selection of 50 traces for calculation considers the

balance between the obtainable image resolution and noise reduction performance

 Step 2: Apply Low Pass Filtering (LPF) with a 2 GHz

cutoff frequency.

• GPR transmission signal’s frequency: 600 MHz – 2 GHz

21

Power Information Characterization

 Hilbert Transform is implemented to extract the pulse

envelope.

 Hilbert Transform of signal 𝑡 𝑢

𝑡 𝑢 = ℋ 𝑡 = ℎ 𝑢 ∗ 𝑡 𝑢 = 1 𝜌𝑢 ∗ 𝑡(𝑢)

 Analytic signal

𝑡𝑏 𝑢 = 𝑡 𝑢 + 𝑗 𝑡(𝑢)

 Signal’s envelope

𝑡𝑏 𝑢 = 𝑡 𝑢 2 + 𝑡(𝑢)2

22

Power Information Characterization (con.)

 Ricker wavelet source – 3 peaks  1st pulse - antennas’ direct coupling  2nd pulse - reflection signal from 1st scatter  3rd pulse – reflection signal from 2nd scatter

GPR A-Scan trace GPR A-Scan envelope

23

Subsurface Identification

 A-Scan decomposition is performed to isolate

subsurface layer and narrow down the scope of data

 Transmitter and receiver antennas’ direct coupling

pulse as the reference signal

 By performing iterative cross correlations, an A-Scan

waveform is decomposed into the combinations of multiple pulses of varying amplitude and phases characterizing the reflection signals from different scatters

24

Subsurface Identification (con.)

Direct coupling signal 1st backscattering pulse 2nd backscattering pulse

25

B-Scan Image Enhancement

 Step 1: Remove the background signal using a 2-D

High Pass Filter (HPF)

• In horizontal direction, the frequency bandwidth of clutter

is much narrower than that of subsurface scattering signals.

 Step 2: Stack every 10 A-scan traces

• Further improve signal SNR as well as reduce data

volume and redundancy

26

Renyi’s Entropy

 GPR backscattering signal 𝑍 𝑢 can be modeled as

𝑍 𝑢 = 𝐸 𝑢 + 𝑇(𝑢) 𝐸 𝑢 - reflection signal from objects of interest, 𝑇(𝑢) - background signals.

 Power normalization is first performed

𝑧𝑗 𝑢 = 𝑍

𝑗(𝑢) 2

𝑗=1

𝑁

𝑍

𝑗(𝑢) 2

𝑧𝑗 𝑢 - normalized signal, 𝑗 - trace index, 𝑁 – total number of traces included, 𝑢 - time index of data points on each trace.

 Renyi’s entropy: 𝐹𝛽 𝑢 =

1 1−𝛽 loge 𝑗=1 𝑁

𝑧𝑗 𝑢

𝛽

𝛽 – the entropy order. (𝛽 = 3 here)

27

2-D Entropy Based RoI Detection

 Step 1: 2-D Renyi entropy calculation  Step 2: Entropy curve smoothing using moving average

method: 𝐹𝑡𝑛𝑝𝑝𝑢ℎ 𝑜 = 1 𝑛

𝑗=𝑜−𝑛+1 𝑜

𝐹(𝑗) 𝐹(𝑗) – entropy value of 𝑗th data in entropy sequence, 𝑛 is selected as 5% of the total number of entropy data points

• Optimal smoothing performance as well as data resolution

 Step 3: Adaptive entropy threshold determination using OTSU

method.

• Minimize inner-group variance & maximize inter-group variance

28

Lab Experiment

Railroad ballast test platform Side view for the subsurface structure  The ballast layer of 0.3 meters thickness is laid above the

soil

 0.75 meters apart from the left end of the platform, an

area of 0.45 meters length and 0.2 meters depth is filled with the fouled ballast, which is a mixture of sand, ballast, and water

29

RoI Detection (1)

Raw B-Scan image Pre-processed B-Scan image

30

RoI Detection (2)

Signal Envelope Signal Decomposition

31

RoI Detection (3)

Ballast layer image Enhanced B-Scan ballast layer image

32

RoI Detection (4)

Entropy along Travel Time (y-axis) Entropy along Scan Axis (x-axis)

33

RoI Detection (5)

RoI Detection Result based on 2-D Entropy analysis

34

Railroad Ballast Field Test

Home made GPR system mounted on vehicle during field tests

35

Test Sites

Metro St. Louis MetroLink blue line (1:00~4:00 am) Railroad ballast

36

RoI Detection (1)

Raw B-Scan image Pre-processed B-Scan image

37

RoI Detection (2)

Signal Envelope Cross-tie marked by signal decomposition

38

RoI Detection (3)

Ballast layer image Enhanced B-Scan ballast layer image

39

RoI Detection (4)

Entropy along Travel Time (y-axis) Entropy along Scan Axis (x-axis)

40

RoI Detection (5)

Suspicious fouled ballast regions marked by white rectangle

41

Conclusions

 The unsupervised automatic ROI detection method

developed in this study can effectively identify regions

• f interest in subsurface for further in-depth analysis.

 The proposed unsupervised automatic GPR data

processing algorithm has been effectively applied to laboratory and field test data.

 The analysis results prove that the proposed algorithm

can correctly identify the region of interest and can accurately measure the region’s location.

42

Future Work

 Local entropy & multi-scale entropy analysis

43

Q & A?

Q&A?

44