Real-Time Characterization of Slurry at the Wet End of a Paper ...

builderanthologyAI and Robotics

Oct 19, 2013 (3 years and 7 months ago)

61 views

Real
-
Time
Characterization of
Slurry at the Wet End of
a Paper Machine

Applied Machine Vision ‘99

September 21, 1999


Hamed Sari
-
Sarraf, Ph.D.

James S. Goddard, Ph.D.

Oak Ridge National Laboratory

Oak Ridge, Tennessee

September 21, 1999

Applied Machine Vision '99

2

Overview


Project Scope


System Description:


Stroboscopic Subsystem


Depth Profiling Subsystem


Facet Model
-
Based Algorithm
Design and Testing


Real
-
Time Algorithm
Implementation


Concluding Remarks

September 21, 1999

Applied Machine Vision '99

3

Papermaking Process

September 21, 1999

Applied Machine Vision '99

4

Scope of Work


Development of an intelligent, on
-
line vision
system for measuring and interpreting
pertinent wet
-
end parameters. System
components include:


Stroboscopic subsystem for capturing intensity
images of slurry


Depth profiling subsystem for recording
topographic structure of slurry


Image analysis algorithms for automatic
information extraction

September 21, 1999

Applied Machine Vision '99

5

Project Scope (continued)


System offers a utility for researchers and
mill operators as a tool to monitor and/or
predict formation, table activity, and
headbox dynamics


AF&PA Agenda 2020 Sensors and
Controls

Department of Energy Office of
Science

September 21, 1999

Applied Machine Vision '99

6

Project Accomplishments


Developed and deployed
stroboscopic subsystem


Captured high
-
quality,
high
-
resolution images of
slurry in paper mill


Developed and deployed
depth profiling subsystem


Obtained data from on
-
line range measurements

September 21, 1999

Applied Machine Vision '99

7

Project Accomplishments


Two image analysis
algorithms developed for
real
-
time implementation:


Algorithm based on
wavelet transform and
correlation dimension for
global homogeneity


Algorithm based on facet
model to quantify
topographic structures of
slurry

September 21, 1999

Applied Machine Vision '99

8

Stroboscopic Subsystem

Laboratory Setup

Support Frame

September 21, 1999

Applied Machine Vision '99

9

Stroboscopic Subsystem
Hardware

Slurry

Datacube
MaxPCI

PC

MD

CCD Camera (Pulnix TM9701)

Lens/
Polarizer

Sync

Captures and
analyzes up to
30 frames/sec.

September 21, 1999

Applied Machine Vision '99

10

Deployed Stroboscopic
Subsystem

Images from
Paper Mill

Headbox

Dryline

Strobe and Camera Setup

September 21, 1999

Applied Machine Vision '99

11

Wet End Intensity Images


Frame rate: 30 fps


Field of view: 1m x 1.2m


Web speed: 450 m/min


Flash duration: 20
m
獥s


Resolution: 1.6 mm/pixel

September 21, 1999

Applied Machine Vision '99

12

Paper Mill Field Test Images

Typical Image from 1st Field Test

Typical Image from 2nd Field Test

September 21, 1999

Applied Machine Vision '99

13

Depth Profiling Subsystem


An off
-
the
-
shelf smart
sensor was utilized to
provide real
-
time range
information

IVP Ranger
RS2200


Characteristics include:


256x256 CMOS detector


On
-
chip processing for
range calculation


Data processed at 100s of
frames per second

100 mw, 830 nm Laser

September 21, 1999

Applied Machine Vision '99

14

Depth Profiling Subsystem
Hardware

Slurry

PC/IVP
Interface

PC/
Datacube
MaxPCI

MD

IVP Ranger
RS2200
Camera

Lens/
Polarizer/
IR Pass

Captures and
analyzes 100
-
200 frames/sec.

Diode laser with
line output

830 nm, 100 mw

September 21, 1999

Applied Machine Vision '99

15

Depth Profiling Subsystem

Intensity Image of Laser

Line on Slurry

MD

Slurry Profile

September 21, 1999

Applied Machine Vision '99

16

Depth Profiling Subsystem

MD

MD

Slurry Profile with 30 mm Lens

Slurry Profile with 75 mm Lens

September 21, 1999

Applied Machine Vision '99

17

Depth Profiling Subsystem


Extracts topographic structures of slurry in 3D


Increase sensitivity of system while expanding its
field of view through two approaches:


Moving closer to the dryline


Using an anamorphic lens


Increase system resolution in machine direction by
investigating a five
-
line laser projection configuration:


Algorithm for extracting range data customized for
this new configuration


Laboratory setup substantiated functionality

September 21, 1999

Applied Machine Vision '99

18

Labeling Topographic
Structures Using Facet Model

Input Image

3D Representation

Labeled Structures

September 21, 1999

Applied Machine Vision '99

19

Topographic Structures
at the Wet End

Slurry

Water

September 21, 1999

Applied Machine Vision '99

20

Structure Detection Flow
Diagram

Median
Filter

Background
Removal

Facet
Model

Morphological
Processing

Enhance
Image

Speckle
Removal

Detect
Features

Combine
and Filter
Features

September 21, 1999

Applied Machine Vision '99

21

Examples of Raw Intensity
Images

Frame (x)

Frame (x+1)

September 21, 1999

Applied Machine Vision '99

22

Filtered Intensity Images

Frame (x)

Frame (x+1)

September 21, 1999

Applied Machine Vision '99

23

Image Analysis: Topography

Filtered Input Image

Preprocessed Image

September 21, 1999

Applied Machine Vision '99

24

Image Analysis: Topography

Topographic Labels

Extracted Hillsides

September 21, 1999

Applied Machine Vision '99

25

Image Analysis: Topography

Detected Structures

Overlaid Structures

September 21, 1999

Applied Machine Vision '99

26

Image Analysis: Topography

Preprocessed Image

Overlaid Structures

September 21, 1999

Applied Machine Vision '99

27

Real
-
Time Facet Model Implementation
On Datacube MaxPCI

Acquire

Image

Image

Store

Camera

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

Convolution Masks

k
1

Store

k
2
Store

k
3

Store

k
4

Store

k
5

Store

k
6

Store

k
7

Store

k
8

Store

k
9

Store

k
10

Store

Polynomial Coefficients for Cubic
Polynomial Approximation:

M2

September 21, 1999

Applied Machine Vision '99

28


Gradient


Grad. Mag


Eigenvalues
of Hessian


Eigenvectors
of Hessian

Coefficients
(k
1

… k
10
)

Peak

Pit

Ridge

Ravine

Saddle

Flat

Hillside

Real
-
Time Facet Model Implementation
On Datacube MaxPCI (continued)






Conditional

Processing







September 21, 1999

Applied Machine Vision '99

29

Concluding Remarks


System offers immediate utility for
researchers as well as mill
operators.


System provides ability to monitor
structure and topography of slurry
as an aid in predicting final quality
of finished paper.