A Signature Based Approach for Quality Checking of

55 downloads 0 Views 5MB Size Report
Figure -02 The Design of the System (Solid Works Drawing). 8th International ... o The table/conveyor is required to be transparent to support for back lighting. .... Quality Checking of Biscuits” in FETSAC 2014 (Faculty of Engineering Student.
A Signature Based Approach for Quality Checking of Biscuits The 8th International Conference in Ubi-Media Computing 2015 August 24/25

P.D.S.H. Gunawardane, P.T.R. Dabare Dept. of Mechanical Engineering Faculty of Engineering Technology The Open University of Sri Lanka

Presented by, P.D.S. Hiroshan Gunawardane

Introduction o This system is designed to used in automation process in quality assuring (QA).

o The design is based on Machine Vision and Digital Image Processing Techniques. o Biscuits which gone through baking ovens will sends to packing through QA. o Traditionally this is achieved by human labor and removed defects increase the quality of the product. o For export quality products fine tuning of QA is a major factor. o The conventional QA is always under fatigue and boredom due to human labor.

o Therefore introducing proper Machine Vision techniques will increase the efficiency and the productivity of this process. 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

2

Introduction Contd. o The Machine Vision based design is focused on detecting cracked and damaged biscuits generally that are to be removed from the conveyor. o The design is concentrated on image acquisition under controlled and conditioned lighting. o Then the image is filtered for noises, enhanced, feature extracted, representation and description is performed to reject non standard and defective biscuits. o The introducing of Machine Vision system for this process will increase the efficiency, reliability, and quality of the production. o Mainly this can be benefitted to the company interns of profits. 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

3

Block Diagram of the System

Detecting of the Biscuits

Capturing an Image

Analyzing the Quality of the Biscuit

Displaying/Rejecting the Biscuit

Backlighting System

Figure -01 Block Diagram of the System

8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

4

Design of the System

Stand

Camera

Transparent Surface

Biscuit Conveyor

Light Source

Figure -02 The Design of the System (Solid Works Drawing) 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

5

Overview of the Design o The data acquisition system shown in Figure-02 can be synchronized with a rejecting system to reject the defects. o Data acquisition is designed in MATLAB Image Acquisition Tool Box. o The table/conveyor is required to be transparent to support for back lighting.

o Selected lighting technique is back lighting. o Therefore the object is placed between camera and the light source. o Because of the main image details is extracted edge, back lighting is used.

o Also during the image acquisition the intensity of the backlighting is controlled.

8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

6

Block Diagram for the Proposed System (Mechanism)

Figure -03 Block Diagram for the Proposed System (Mechanism) 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

7

Illumination Technique(Back Lighting System)

Camera Stand Back Lighting

Light Source

Figure -04 Back Lighting System (Solid Works Drawing) 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

8

Back Lighting System Cont. o Here back lighting technique has been used. The object is placed between camera and the light source. Because of the requirement is edge extraction, the most appropriate techniques is to use back lighting. o The object was placed on a transparent glass and then a thin sheet of paper was placed on the glass to control the light to camera and then back lighting has been used

Video – 01 Demonstration of effective application of back lighting

8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

9

Image Processing

o Once image is captured, MATLAB Image Processing Tool Box was used process the image. o Figure-05 shows how the program flow was designed to process the image.

Figure -05 Program Flow 8th

International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

10

Image Processing Contd.

o The signature, either can be plot to visualize or can be compared against a standard matrix to identify defects.

Note : The bound2im and the signature functions were not built in with MATLABR2010a image processing tool box there for those additional custom m-functions were taken from Digital Image Processing Using MATLAB (Rafael C. Gonzalez). 8th

Figure -06 Program Flow Contd.

International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

11

Image Processing Contd.

Figure -07 Boundary Generation and Signature Plot for Various Shapes of Biscuits. 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

12

Comparing & Decision Making o Comparing is used for the decision making. o We have to get the data for a particular cased and compare with a standard case. o First the standard image was taken and then saved it. Then we have to compare it with the current image. In some cases one matrix is larger than the other one and the comparing couldn’t perform. Therefore we have to create a matrix with value 1 for the largest matrix and then assign the smaller one to larger one before compare.

o Also we are here allowed a variation of value 25 and that can be either tune manually or intelligent control methods can be introduced to tune it more accurately. 13 8th

International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

Testing Methodology o Three shapes(round shape, square shape and rectangular saw trothed edged shape) of biscuits are used for testing. o Crack description and rotation angle was used as parameters. o The K parameter was generated by comparing with the template image.

[a] No Crack

[b] Perimeter Damages

[c] Internally Cracked but Not Separated

[d] Cracked and Separated

[e] A large Piece is missing

Figure -08 Round Shape Biscuit. 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

14

Testing Methodology

[a] No Crack

[b] Perimeter Damages

[c] Internally Cracked but Not Separated

[d] Cracked and Separated

[e] A large Piece is missing

Figure -09 Square Shape Biscuit.

[a] No Crack

[b] Perimeter Damages

[c] Internally Cracked but Not Separated

[d] Cracked and Separated

[e] A large Piece is missing

Figure -10 Rectangle Saw toothed Shape Biscuit. 15 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

Test Results Summary of Test Results Round Shape

Square Rectangul Shape ar Shape

No Crack

10

10

10

A damage in perimeter

10

10

10

0

0

0

10

10

10

10

10

0

Internal Crack /Crack does not Separate Cracked and separate (two pieces) A large piece is missing

Table -01 Test Results for Various Shapes of Biscuits.

Round Shape

Square Shape

Rectangular Shape

12 10 8 6 4 2 0 No Crack A damage Internal Cracked in Crack and perimeter /Crack separate does not (two Separate pieces)

A large piece is missing

Graph -01 Summary of the Test Results.

8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

16

Test Results Contd. Rotation Angle No Change Increased No Change A damage in perimeter Increased Internal Crack/ Crack does not No Change Separate Increased No Crack

Cracked & Separate (two pieces)

A large piece is missing

Round Shape Square Shape

Rectangular Shape

Min K Used K Min K Used K Min K Used K 3 10 6 10 4 10 3 10 60 80 115 130 25 20 25 15 80 55 25 20 85 50 115 100 6 10 10 15 5 10 6 10 60 70 5 10

No Change

> 100

100 > 100

100

> 100

100

Increased No Change Increased

> 100 100 100

100 > 100 75 100 75 70

100 80 50

> 100 > 100 > 100

100 OR OR

Table -02 Test Results of K Parameter Variation for Various Shapes of Biscuits. 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

17

Conclusion o The proposed methodology (signature based approach) for the quality checking shows 73.33% in accuracy. o About another 7.5% of accuracy could be gained by using a mixed lighting system. o This system shows a good accuracy for symmetrical shapes and for other types an intelligent based approach is required to inspect quality. o The decision making based on the K parameter could be made more robust by introducing a fuzzy logics based approach to the system. o The input fuzzy variables would be the K parameter and orientation of the biscuit while the output fuzzy variable would be the quality of the biscuit. 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

18

Enhancements & Further Enhancements Required o There is only one biscuit per time (Single Biscuit Queue). o The biscuits are guided to position correctly to the camera on the conveyor. o No angular displacement in between standard and current images. o The conveyor speed is relatively small. o The shape feed to the image processing sequence has only one boundary. o The conveyor is designed according to the lighting system that is mentioned in the section Lightning. o Here the images are not taken in the real industrial situation. o There for the noise is very limited. But in the industry there may be various types of errors due to burned pixels and other electromagnetic inductive interference. 8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

19

Enhancements & Further Enhancements Required o Those can be modeled respectively by salt and pepper noise and Gaussian noise. o Therefore we can introduce salt and pepper noise and Gaussian noise to our system and filter it. o The best option for reduce these noise is median filter. .

Figure -11 Gray Image

Figure -12 Salt Pepper Noise

Figure -14 Salt pepper Filtered.

Figure -13 Gaussian Noise

Figure -15 Gaussian Noise

8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

20

ACHIEVEMENTS o Best Presenter Award for “A Computer Based Vision System for Real – Time Quality Checking of Biscuits” in FETSAC 2014 (Faculty of Engineering Student Academic Conference in The Open University of Sri Lanka.

o Best Project Award for “ Quality Checking of Biscuits using Machine Vision and Fuzzy Logics” in 11th Annual Sessions, SLAAI (Sri Lankan Association for Artificial Intelligence) 2015.

8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

21

THANK YOU

8th International Conference in Ubi Media Computing – 2015 A Signature Based Approach for Quality Checking of Biscuits

22

Suggest Documents