There are few methods of Surface Inspection using pixel based method, namely :
1. Traditional Object Recognition: this methods is identifies object by generalizing object's image pattern. The first step is extraction the features of the object and then train the system according to the features by using classifier algorithm, such as Neural Network or SVM. This method will face different challenges, such as : illumination variance and distribution, object surface characteristics, orientation and occlusion
2. Blob/Particle Analysis : This methods will extract any object in the surface by separate it from the background
then group it accordingly to form a blob.
The geometry of the blobs is then used for surface detection. This is very fast and practical method for simple surface
inspection. The disadvantage of this method is that separation of the object from the background is not an easy task
due to noise as a result of change of color, scratch, marking etc.
3. Template Matching with Normalized Grey Scale correlation :
In this method the template is stored and then matched with the object inspected with some correlation method.
The advantage of this method is that it is much more accurate and more robust than the previous two methods. The added advantages are that it is relatively easy to train an object and that the object does not need to be separated from
the background. The disadvantages of such a method are that it cannot handle much variation in rotation and size and is highly affected by non-uniform shading. In the application of inprocess inspection in industry, changes in rotation and size as well as non-uniform shading is the standard rather than the exception.
Now in contrast, the vector based image processing will be described and compared. In the previous three methods described above, the image processing are applied on the basis of pixel by pixel, hence pixel based image processing.
This method is slow and challenged by problems in different illumination, pose and occlusions.
Vector based image processing is totally different that it converts all the pixel into geometric features by means of synthetics or mathematical model.The geometric feature can be line segments, arcs, angles and open or closed geometric shapes.
By using geometric features, the image analysis is not affected by color changes or non-linear changes in size such as those found with components due to manufacturing variations. This method is also robust against variance in shading an non-linear lighting.
Senin, 12 Mei 2008
The comparison of Pixel and Vector Based Image Processing
Label:
image processing,
pixel,
surface inspection,
vector
Langganan:
Posting Komentar (Atom)
Tidak ada komentar:
Posting Komentar