Applicative Framework for End-to-End AOI Implementation
OVERVIEW
Timely and accurate detection of defects helps industries to apply quality control strategies to maintain a competitive edge . The target is to achieve 100 % qualified products while reducing internal scraps . Among the techniques used to assess the product quality , optical inspection approach for defect detection is one of the most common procedures used in industry . Optical inspection techniques can be performed by human inspectors ( manual optical inspection ) or automatically by using sensors and image processors . This last technique takes the name of Automatic Optical Inspection ( AOI ).
In the competition between manual and automatic approaches , inspection speed and accuracy play a determinant role and the gap between the two is becoming much more difficult to fill . Indeed , the fast and increasing development of technology has pushed the success of AOI over manual inspection . The most advanced systems are capable of identifying a variety of surface defects such as nodules , missing components , scratches and stains as well as tiny defective deviations , ( with low intensity and contrast ) that are difficult to detect with the naked eye . Moreover , according to M . -J . -J . Wang and C . -L . Huang 1 , human vision inspection performance gets worse with repetitive routine jobs due to fatigue .
Three key elements drive Automatic Optical Inspection processes : the image capturing system , composed mainly by a camera and its controlling software , one or more light sources and the running application for image analysis .
The former may present many variants depending on the complexity of quality control to carry out : it may involve single or multiple cameras to provide a better 2D imaging or even a 3D capture .
Mainly , the vision systems allow two different acquisition types ( from now on , one single camera is taken as reference ):
• Streaming video : the camera takes a streaming video and extracts frames from it . The image processing is performed on the captured frame . This approach is not highly accurate in terms of image quality but guarantees very high speed .
• Still image capturing : the system is placed relatively close to the object to be inspected and takes a picture by reacting to an external or internal input .
1
M . -J . -J . Wang and C . -L . Huang , ‘‘ Evaluating the eye fatigue problem in wafer inspection ,’’ IEEE Trans . Semicond . Manuf .
4 March 2022