Documente Academic
Documente Profesional
Documente Cultură
Year
Publication/Typ
e
Elsevier/
Carnegie Mellon
University
ASCE
International
Journal of
Emerging
Technology and
Advanced
Engineering
Southern African
Transport
Conference (SATC
2015)
BEST: IJMITE
KEYWORDS:
pothole
pothole detection
visual
techniques image processing
Camera model
Processor
Filtering techniques /
Pre-processing
techniques
a HP Elite Autofocus
Webcam
Smartphone camera
(Samsung galaxy)
Intel i7 CPU, 8 GB
Ram
1080p ; 10 Hz
60 seconds taken to
process image
Hewlett-Packard Elite
Autofocus webcam
Not Specified
Not Specified
Gaussian Filter
Not mentioned
Intensity normalization
(Contrast stretching)
Pixel saturation
crack detection
Segmentation
techniques
No of
Sample
Shape extraction/
s
Comparsion
(Pothol
techniques
es)
Accuracy / Efficiency
Morphological erosion
folllowed by
connecting branching
points
70
Accuracy - 86% ;
Precision - 82%
Recall - 86%
Accuracy - 94 % ;
Precision - 75%
50
TP- 42 ; TN - 8 ; FP-14
1. Pixel Information
Calculation
15
93%
100%
3. Second order
moment (Variance)
15
100%
4. Cross Co-relation
calculation
15
73%
Precision - 81%
Ordered histogram
intersection
Otsu's algorithm
Spectral Clustering
( k-means algorithm) 50
Accuracy - 81%
Spectral Clustering
Dividing image into blocks ( k-means algorithm) 50
Accuracy - 76%
102
Accuracy - 73.5% ,
Precision - 80%
TP 44; FP-11 ;
Accuracy - 83%
Problems / Limitations
Remarks
Introduces pothole
tracking
Reference/LI
NK
Not mentioned
Cost effective
Year
KEYWORDS:
2004
Publication/Type
IJARET
NCHRP-IDEA
Camera model
Processor
Algorithm
2 Point GreyFlea 3
Disparity calcuation
(matching cost
computation, controlled
search range)
Webcams placed at
distance of 6 cm
Matlab 2016
Points matching
D = Ave( Sum ( (left)^2
- (Right)^2))
Camera calibration
No of
sample
s
Conversion between
disparity and
eucledian, surface
fitting
Accuracy /
Efficiency
Not mentined
Soble edge
detection followed
by erosion and
dilation
24
83.30%
Calculation of depth
Speeded up Robust Feusing disparity
3D surface
reconstruction
97%
Harris Corner
detection,
interpolation
3D reconstruction
Not yet
implemented
Problems / Limitations
*Detailed information
given in paper
Remarks
Reference/LINK
http://mioh-utc.udmercy.edu/research/ts43/pdf/MIOH_UTC_TS43_2012Final_Rpt_Pavement_Distress_Evaluation_U
sing_3D_Depth_etc.pdf
http://onlinepubs.trb.org/onlinepubs/archive/studies/idea/finalreports/highway/NCHRP88_Final_Report.
KEYWORDS:
Year
2009
Publication/Typ
e
Camera model
Processor
Filtering techniques /
Pre-processing
techniques
Segmentation
techniques
No of
Sample
Shape extraction/
s
Comparsion
(Pothol
techniques
es)
http://www.dot.nd.go
v/divisions/materials/r
esearch_project/ndsu
0801final.pdf
Accuracy /
Efficiency
Problems / Limitations
Remarks
Reference/LINK
Year
Publication/Typ
e
Camera model
Elsevier/
a HP Elite Autofocus
Webcam
Carnegie Mellon
University/
Smartphone camera
(Samsung galaxy)
1080p ; 10 Hz
ASCE
Hewlett-Packard Elite
Autofocus webcam
Resolution: 640 480
pixels with a frame rate
of 30 frames/s.
International
Journal of
Emerging
Technology and
Advanced
Engineering
Not Specified
Southern African
Transport
Conference (SATC GoPro (model not
2015)
specified)
Resolution - 1270*720 ;
60 fps
BEST: IJMITE
Processor
Filtering techniques /
Pre-processing
techniques
Segmentation
techniques
Shape extraction/
Comparsion
techniques
HSB based on
traingle algorithm
Morphological erosion
folllowed by
connecting branching
points
Intel i7 CPU, 8 GB
Ram
Not Specified
60 seconds taken to
process image
1. Pixel Information
Calculation
2. First order moment
(Mean)
3. Second order
moment (Variance)
4. Cross Co-relation
calculation
Gaussian Filter
Dilation after
application of Canny
Edge Detection
Convex hull algorithm
Sobel Edge
detection
Ordered histogram
intersection
Not mentioned
Otsu's algorithm
Spectral Clustering
( k-means algorithm)
Intensity normalization
(Contrast stretching)
Spectral Clustering
( k-means algorithm)
Pixel saturation
No of
Sample
s
(Pothol
es)
70
Accuracy /
Efficiency
Problems /
Limitations
Remarks
Accuracy - 86%
; Precision 82%
Recall - 86%
TP - 31; FP-7;
TN-72; FN -5
50
Accuracy - 94
% ; Precision 75%
the proposed
method
assumes that
only one
pothole enters
the viewport at
a time
Introduces pothole tracking
TP- 42 ; TN - 8 ;
FP-14
15
93%
15
100%
15
100%
15
73%
Vehicle cannot
travel faster
than 40 Kmph
112
1. Canny Edge
Detection is
Precision - 81% very strict.
TP-72; FN-24;
FP-16
2. Cannot
detect potholes
on outer edge
of roads
102
Accuracy 73.5% ,
Precision - 80% Not mentioned
TP 44; FP-11 ;
50
Accuracy - 81%
50
Accuracy - 76%
24
Accuracy - 83%
Cost effective