Visual inspection is a type of sensory testing (an inspection that uses human senses for evaluation) and is an effective quality control method that utilizes human flexibility.
However, as noted in "Challenges of Visual Appearance Inspection," there are various challenges that arise "precisely because it involves humans."
Here, we introduce two alternative approaches as solutions to these challenges: "Functional Inspection" and "Inspection Utilizing Image Processing Technology.
1. Determination through Performance Testing
Performance Inspection for evaluating whether the performance of a product or system meets specifications or requirements. For instance, if the purpose of visual inspection is to confirm the presence or absence of cracks in a product, it may be possible to replace visual inspection with strength measurements for crack confirmation or color measurements for color confirmation.
Thus, in performance testing, using measuring instruments allows not only for pass/fail determination but also for recording quantitative data, making it a useful method of determination if replacement is feasible.
On the other hand, performance testing sometimes involves selecting representative products from mass-produced items and conducting destructive tests, depending on the content. Therefore, it can be challenging to apply it as an alternative method to visual determination inspection, where 100% inspection is required.
Additionally, when considering performance testing as an alternative to visual determination, it is necessary to consider whether performance testing can fully satisfy all aspects of visual determination.
2. Determination Using Image Processing Technology
Inspection utilizing image processing technology involves capturing images of products using cameras or sensors instead of human eyes traditionally used in visual inspection, and analyzing those images using computers in place of the human brain to detect defects.
This technology was initially employed in the medical field, primarily using imaging techniques such as X-rays and ultrasound starting from the 1980s.
Subsequently, the utilization of image processing technology began in the inspection of semiconductor components, where images captured by cameras were analyzed by computers.
Furthermore, with the advent of "Deep Learning" technology in the 2000s, recognized as the second AI boom, image processing technology became applicable to a broader range of fields.
For instance, advancements have enabled "person and object recognition for autonomous driving," and "detection of people’s movements and attributes (age group, gender) within camera views." Additionally, it has become possible to perform "product differentiation," which was previously challenging.
However, despite the remarkable advancements in image processing technology, it does not guarantee the detection of all types of defects. If the desired area cannot be captured as an image in the first place, it cannot be identified.
Firstly, if you cannot capture the area you want to identify as an image, it is not possible to perform determination. Moreover, to achieve the intended determination, it is necessary to appropriately set the conditions for image processing.
It is crucial to understand the challenges of both visual determination and determination using image processing technology, and to carefully evaluate which method is most suitable while continuously evaluating the options.
3. Summary
As alternative methods to visual determination, there is potential to utilize various technologies, such as "Determination through Performance Testing" and "Determination Using Image Processing Technology."
On the other hand, there is no universal solution that "can completely replace visual determination with this technology." Applying appropriate alternatives according to the content of visual determination and the challenges present is crucial.