fundamentals of defect detection/web inspectioncameramachinepapier.com/docs/fundamentals of defect...

4
ECS, Inc. (704) 844-1160 700 Sam Newell Road • Matthews NC Copyright 2009 Fundamentals of Defect Detection/Web Inspection Data Sheet Document 40.23 Capture • Process • Display Defect detection (also called web inspection in the application of web based manufacturing) is the process of identifying unwanted artifacts in a process. This could be sizing agents on calendar rolls, edge defects, slime spots, debris, holes, wrinkles or any other issue that creates an unwanted change in the grayscale value of the paper web. Fundamental elements of a system are cameras and lights that capture images and computers and software that process the images to determine what is a defect. The key requirement of any system is to capture the unwanted defect with absolute accuracy (no false triggers and no missed defect detection) in environments that may be less than ideal (dust, falling debris, water vapor, low frequency ambient lighting, space restrictions, etc). The final solution should only use the minimum components (camera, lights, processing computers) to satisfy the specific detection requirement so the total installed cost is as low as possible. Summary Overview A defect detection system can be broken down into four (4) primary parts: Key Factors: - Camera Type (line or area scan) - Camera Resolution (pixels) - Image Clarity - Camera Position, Shutter and FOV (Field of View) - Illumination Quality (LED single point or LED Beam) Key Factors: - 100% Software Driven - High Defect Recognition - Minimal False Defect Detection - Image Analysis is the main engine - Any GigE compliant camera can be used for image acquisition. This provides unlimited flexibility in frame rate and resolution. This implementation is seamless with ECS’s software driven image analysis engine. Image Analysis Dynamic Template Based Reference Image Analysis OPC Machine Data Input Start/Stop Logic for Image Analysis Process Data Parameters (examples include speed, roll ID, roll diameter, etc) Each Incoming Frame Dynamic Template ROI (Region of Interest) Comparison through multiple layer pass/fail filtering (defect threshold, pattern matching, direction analysis, wrinkle suppression, edge defect, etc) 1. Image Acquisition 2. Image Processing Comparison 3. Operator Interface 4. Data Export/ Alarm Output Key Factors: - Intuitive operator interface with imbedded screen movie OPL (One Point Lesson) tutorials - Defect pictures, defect parameters and roll map Detection of Defect One Camera Group Multiple Camera Groups Or Key Factors: - Excel output of defect data for mill wide distribution - ECS system control via any mill computer through multiple ECS user licenses - OPC or PLC output from defect data processing - Alarm output to stop process

Upload: dangcong

Post on 07-Feb-2018

235 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Fundamentals of Defect Detection/Web Inspectioncameramachinepapier.com/docs/Fundamentals of Defect Detection an… · Defect detection (also called web inspection in the application

ECS, Inc. (704) 844-1160 700 Sam Newell Road • Matthews NC

Copyright 2009

Fundamentals of Defect Detection/Web Inspection Data Sheet Document 40.23

Capture • Process • Display

Defect detection (also called web inspection in the application of web based manufacturing) is the process of identifying unwanted artifacts in a process.

This could be sizing agents on calendar rolls, edge defects, slime spots, debris, holes, wrinkles or any other issue that creates an unwanted change in the

grayscale value of the paper web. Fundamental elements of a system are cameras and lights that capture images and computers and software that

process the images to determine what is a defect. The key requirement of any system is to capture the unwanted defect with absolute accuracy (no false

triggers and no missed defect detection) in environments that may be less than ideal (dust, falling debris, water vapor, low frequency ambient lighting,

space restrictions, etc). The final solution should only use the minimum components (camera, lights, processing computers) to satisfy the specific detection

requirement so the total installed cost is as low as possible.

Summary

Overview A defect detection system can be broken down into four (4) primary parts:

Key Factors:

- Camera Type (line or area scan)

- Camera Resolution (pixels)

- Image Clarity

- Camera Position, Shutter and FOV (Field of View)

- Illumination Quality (LED single point or LED Beam)

Key Factors:

- 100% Software Driven

- High Defect Recognition

- Minimal False Defect Detection

- Image Analysis is the main engine

- Any GigE compliant camera can be

used for image acquisition. This

provides unlimited flexibility in

frame rate and resolution. This

implementation is seamless with

ECS’s software driven image

analysis engine.

Image Analysis

Dynamic Template Based Reference Image Analysis

OPC Machine Data Input

Start/Stop Logic for Image Analysis Process Data Parameters (examples include

speed, roll ID, roll diameter, etc)

Each Incoming Frame

Dynamic Template

ROI (Region of Interest) Comparison

through multiple layer pass/fail filtering (defect threshold, pattern matching, direction analysis,

wrinkle suppression, edge defect, etc)

1. Image Acquisition

2. Image Processing

Comparison

3. Operator Interface

4. Data Export/ Alarm Output

Key Factors:

- Intuitive operator interface with imbedded screen movie OPL (One Point

Lesson) tutorials

- Defect pictures, defect parameters and roll map

Detection of

Defect

One Camera Group Multiple Camera Groups Or

Key Factors:

- Excel output of defect data for mill wide distribution

- ECS system control via any mill computer through multiple ECS user licenses

- OPC or PLC output from defect data processing

- Alarm output to stop process

Page 2: Fundamentals of Defect Detection/Web Inspectioncameramachinepapier.com/docs/Fundamentals of Defect Detection an… · Defect detection (also called web inspection in the application

ECS, Inc. (704) 844-1160 700 Sam Newell Road • Matthews NC

Copyright 2009

Fundamentals of Defect Detection/Web Inspection Data Sheet Document 40.23

Capture • Process • Display

Image acquisition explains the method the system uses to capture images and defines the number and qual-

ity of pixels the software program receives to determine what is and is not a defect.

Our method of image capture is a GigE area scan camera (see data sheet 40.18) GigE specifies how the

signal is sent from the camera to the computer. As the name suggests— the camera scans a given area

with a rate of capture defined in frames per second. The amount of pixels the camera is able to scan is

called the camera resolution. The area the pixels represent (what the camera is looking at) is the field of

view (FOV). The other option is line scan. These cameras scan only one or a collection of lines. The image

must be moving under the cameras and the resulting image is created by ‘pasting’ the lines together. The

resolution of the camera is defined by the number of pixels it scans in the line (width) and rate of scanning

is expressed in Hz. The advantage of an area scan camera is it can be set to behave like a line scan cam-

era in variable amounts at different frame rates and the picture quality is superior. For example - it can

scan at 200 lines high by 750 pixels wide at 100 frames per second in one setting or 100 lines high by

750 pixels wide at 200 frames per second. This allows easy customization to fit various sheet widths, MD sheet exposure and machine speeds.

The amount of information (data rate) an area scan captures is the number of pixels it scans and how many times that scan is completed in one second.

The quality of those scanned pixels is defined by (1) focus and contrast (2) shutter speed (3) depth of field and (4) lighting. Lighting is the condition that

most effects all other parameters for the quality of capture. The light should be even across the camera field of view and have enough lumens to support

high speed shutter and depth of field (if the camera is set off to the side). LED in spot and wide area beam array is the preferred lighting technology

(see data sheet 40.09)

Image Acquisition

True Definition of ‘Camera Resolution’

Camera resolution (as defined by number

or pixels) is only a small factor in the true

‘resolution’ the camera is providing. More

important than total number of pixels is the

quality of those captured pixels. If the

pixels are out of focus, the stop action

clarity is too low (motion blur) and the

contrast is poor then a million pixels have

no value as compared to a far less number

of quality pixels.

Example:

Inspection Parameters

Sheet is 240 inches wide in the cross direction (CD)

Only 10” is exposed in the cross direction and the sheet travels at 3,600 feet per minute

Various defects of width .375 inches need to be captured

Solution

The camera must scan (number of pixels) a wide rectangular image and each scan must be done (frame rate) to capture the defect while in the camera

field of view. GigE area scan cameras have a sliding scale of pixel capture vs. frames per second. At high resolution the frames per second will be

lower and at lower resolutions the frames per second will be higher. Typically it is width (CD of the paper) that determines the number of cameras

needed. The machine speed typically determines the number of frames per second the camera setting must support.

A MD field of view of 10’ at 3,600 fpm requires at least 72 frames per second (at 3,600 fpm a defect travels 10 inches in the MD from one dis-

played frame to the next). At 90 frames per second one available model GigE camera (see www. Prosilica.com) can scan 1280 wide by 320 high.

Using a factor of 4 pixels per defect ratio one camera can cover 120” in the CD (.375 x 1280/4). There are more lines in camera pixel height (MD

field of view) than required which gives room for higher scan rates to accommodate increased machine speed.

‘How may pixels does it take?’ - the critical question.

This is one of the most important factors in web inspection

because it determines the number of cameras (and poten-

tially computers) required = COST. If the defect contrasts

significantly to the sheet and no ambient noise is present -

in theory only one pixel in the perfect case is required to

see a defect. In real applications this is not the case and

the defect to pixel ratio is typically from 3 to 6.

Note on Ambient Lighting

ECS has developed a dynamic

template image processing solution

that can use existing low frequency

lights for web inspection . This can

be extremely beneficial for doing

low cost web inspection. See data

sheet 40.12.

Height - number

of scanned lines

(measured in

pixels)

Pixel

Line Scan Width

(measured in pixels)

Figure 1

Area Scan (multiple scanned lines of

pixels)

Camera

FOV (Field

of View Defect

Page 3: Fundamentals of Defect Detection/Web Inspectioncameramachinepapier.com/docs/Fundamentals of Defect Detection an… · Defect detection (also called web inspection in the application

ECS, Inc. (704) 844-1160 700 Sam Newell Road • Matthews NC

Copyright 2009

Fundamentals of Defect Detection/Web Inspection Data Sheet Document 40.23

Capture • Process • Display

Operator Interface

Roll Map Table

(For selected group). A scroll window that shows

each roll map, roll ID number (OPC input), number

of defects, and length. Display is for selected day

(indicated above table).

Web Inspection Group

The selection of a group creates the entire web inspection

page experience

Roll Map Table Roll Map Defect Table (for selected roll) Defect Picture (for selected defect)

Defect Picture

The selected defect from the Roll Map and Defect table

is displayed. Comments and bookmarks can be added.

The mouse scroll wheel can be used to quickly advance to

new defects

Defect Table

Displays defect information including size, CM and MD

direction, time and comments

Image Processing

Figure 3

Image Processing Setup Page

Figure 2

Web Inspection Interface

The operator interface is fully integrated into the complete ECS software package. Many of the

web inspection installations have multiple groups of cameras along a process. In this case each

group of cameras is identified as a Web Inspection Group. Selecting one of these groups will dis-

play those defects captured by that collection of camera or cameras. See data sheet 40.12 for

more information on multiple group inspection systems.

Roll Map

Displays the roll map or current selected Inspec-

tion Group and Roll in cross direction (x-axis)

and run length (y-axis). Shows defect type,

trending, repeating defects and camera lanes.

Image processing is the main engine of defect detection. A dynamic template

is compared to incoming frames within an region of interest. Pixels that differ

in grayscale values are further analyzed to determine if the grayscale change

is a defect. The comparison follows a set of instructions that determines:

How the reference template image is established.

This is critical as it determines how well the system responds to

noise in the environment. Many web inspection applications are

done using ambient light with cameras at an acute angle to the

web. These are not ideal conditions for web inspection and ECS

has developed special processing filters to deal with such condi-

tions.

How the reference template is compared to the incoming frames within

the region of interest (ROI).

The comparison is a grayscale calculation done on a pixel by

pixel basis. The comparison is done in a multi layered pass/fail

filtering process. A possible defect image must pass several

filters before it is considered a defect. This multi-stage process-

ing eliminates false defect detection. Examples of process

stages include pattern matching, wrinkle suppression, ambient

Page 4: Fundamentals of Defect Detection/Web Inspectioncameramachinepapier.com/docs/Fundamentals of Defect Detection an… · Defect detection (also called web inspection in the application

ECS, Inc. (704) 844-1160 700 Sam Newell Road • Matthews NC

Copyright 2009

Fundamentals of Defect Detection/Web Inspection Data Sheet Document 40.23

Capture • Process • Display

Data Export / Alarm Output

Data Export is typically done with a one button Excel Export command. Re-

lated images to each entry can be exported in file or the user has the option

to export only the defect data. A PLC signal or OPC tag can be created

based on several factors including

Severity of defect (primarily size)

Defect position

Defect repetition

Customized Excel Macros can be created to output any number of specific

formats.

In many cases where defect detection is being accomplished in multiple positions along a process

each web inspection group may need a dedicated set of image processing instructions. This infor-

mation determines when image processing is suspended (example –splice) or when to become

active (example - above certain speeds). For establishing a roll map defect database the ECS

system must know when the roll starts, stops, is replaced by another roll and the corresponding

roll ID number. OPC is the preferred method to read this data from the mill wide process net-

work. ECS uses the industry standard Cogent Datahub software package

(www.opcdatahub.com). In this configuration the ECS computer is a full OPC server and can com-

municate directly to OPC clients and hosts. It also allows tags to be sent out from the ECS system.

PLC data can also be read for simpler applications where dry contacts and analog signals can

represent image processing suspend commands and speed signals

About ECS

ECS, Inc, (formerly Carotek ECS) was the first company to bring to market a digital event capture system for industrial manufacturing. From 1993 to now

we continue to develop, install and service the leading WMIS (web monitoring and inspection system) for pulp and paper, converting, packaging and

related processes. With corporate offices located in North Carolina, ECS is best equipped to handle your web monitoring needs for new systems, up-

grades, spare parts and service.

If you have any questions please feel free to contact: Brian Mock, Director 919.349.9001 (cell) John Larkin, Director 704.906.6210 (cell)

Image Processing continued - OPC Processing Commands

Figure 4

OPC Tag Chart

Figure 5

Excel Defect Export Worksheet