104
4.3 Experimental Results
The integration between software and hardware is important aspect in determining the approval and recognition of the method being used. So, this system is designed and
developed based on software and hardware capabilities. For this chapter, its consists of three results recorded in determining that the software and hardware results is tally based
on the shape matching DSM application, classificat ion of defect’s type DIM application
and correction of defect CoD application.
4.3.1 2D Matching
The development of this complete system comes with 2D matching application also known as Defect Shape Matching DSM. For the simulation, algorithm is tested with four
parameters that would affect the recognition rate such as scales of image captured, rotation of image and lighting from environment. Those parameters are important aspects in
considering the algorithm perfectly fit with environment or not. For this section, the Figure 4.12: Correction of Defect Experimental Setup
105 experimental being done in 10 experiments each of them have 5 pyramids design using
plain cardboard with same dimension. The idea based on the real-time experimental by integration between software and hardware consists camera and KUKA arm robot. Table
4.2 shows the results occur from 50 tested images. Table 4.2: Defect Shape Matching DSM for 50 Tested Images
No. of Experiment
Total Pyramid
Average Recognition Rate
Total Pyramid Classified
1 5
0.974635 5
2 5
0.984562 5
3 5
0.975542 5
4 5
0.966687 5
5 5
0.958846 5
6 5
0.977913 5
7 5
0.979128 5
8 5
0.934389 4
9 5
0.983135 5
10 5
0.961359 5
Total 50
0.9696196 49
Success Rate 96.96196
98.00
4.3.2 Classification of Types of Defect
Based on the experiment 3 in classifying the defect’s types, three types of defect used as model template and each of them being trained in training phased of Defect
Inspection Monitoring DIM algorithm. Three of them are; 1 gap defect; 2 bumper defect; and 3 bubble defect. There are 50 images being tested using this system in findings
106 the relevant information regarding of successful recognition rate in classified defects occur
in every images. The results are recorded as shown in table below. Table 4.3: Result for Defect Inspection Monitoring DIM for 50 Tested Images
Total Defect in Image
Total Image Total Defect
Defect Detected Defect
Undetected
1 2 2
2 2
7 14 14
3 6 18
18 4
10 40 39
1 5
7 35 32
3 6
5 30 28
2 7
7 49 47
2 8
6 48 46
2 Total
50 236 226
10 Percentage
95.76 4.24
According to data collected, there are about 226 from 236 defect are classified as defect through the system which means the system achieved about 95.76 percent of
successful recognition rate. There is still having its error in determining the defects from tested images which compile about 4.24 percentage of failure. Results of each image
process with its defect classification shown in Figure 4.13 below.
107 Figure 4.13: Result for Defect Inspection Monitoring DIM
4.3.3 Correction of Defects