110 CoD process whether the gluing object being accepted as perfect gluing object or
recognize it as rejected object.
4.4.1 Constraint in Template Creation
In all the experiments extra care must be taken in order to achieve good acceptance rate. During the template creation step, the size of the required features must be
considerably large so that the created outline does not overlap as this can causes inaccuracy. Not only that, the extraction between background and its image must be done
perfectly in capturing a good template images. For experiment 1 2D Matching and 2
Classification’s of Defect, simple shapes are used for template creation. In order to achieve good template matching results, the
pixel value difference between the object and the background must be sufficient for the system to work effectively. During the training step, the resultants template must have an
obvious shape. The training template for experiment 1 are shown as example, the pyramid shape is shown in Figure 4.15 and the defects template gap and bumper defect are shown
in Figure 4.16 and Figure 4.17 where as Figure 4.18 shows when a poor pixel difference that was used during template creation therefore the resultant template is undesirable. All
the Figures on the left side show the capture image of the objects and Figures on the right side shows the reduced image showing the trained template in red lines.
111 Figure 4.15: Pyramid Shape and its Corresponding Template
Figure 4.16: Defect Gap and its Corresponding Template
Figure 4.17: Bumper Defect and its Corresponding Template
112 Figure 4.18: Poor Template for Image Pyramid
In developing good template image, each characteristics of the object must be obtained perfectly as shown in Figure 4.15, Figure 4.16 and Figure 4.17. Better template
image created will result to better recognition rate. As shown in table below where it shown the difference between good and poor template that cause greatly in recognition
result. The results is comparing using 50 tested images same as experiment 1. Table 4.5: Comparison between Good and Poor Template Image
Condition Template Image Recognition Rate
Recognition Rate
Good Template Image 0.9696196
96.96196 Poor Template Image
0.824591 82.4591
113 The percentage error is too large to be implemented in industry where it can reflect
to their quality product. Without proper step in defining good or poor template will be crucial to machine vision system where it cause system to be less effective and productive.
4.4.2 Constraint in Camera Placement