How is machine vision inspection implemented?
Release date:
2015-11-13 00:00
Source:
Visual inspection essentially involves first capturing an image of a合格 (qualified) workpiece. This image is then used as a reference to compare against images of other workpieces being inspected. An (adjustable) threshold is set: if the degree of similarity between a workpiece and the reference exceeds this threshold, the workpiece is deemed合格; otherwise, it’s considered不合格. — Is this understanding correct?
If the workpiece being tested is not precisely positioned within the area being captured—for example, if it’s rotated at an angle (or slightly misaligned)—will this affect the inspection results? Could you provide a simple explanation on the topic?
Simply put, it works like this: capture a passable image, compare it to a standard-quality reference image, and set a specific threshold. If the comparison meets or exceeds the threshold, the result is deemed "OK"; otherwise, it’s classified as "NG." Of course, this process relies heavily on the image-processing algorithms used.
Additionally, in our GOI software, you can set angular deviations and adjust both horizontal and vertical scaling. However, the angles and offsets shouldn’t be too extreme—exceeding the specified range would also render them invalid. The same functionality exists in the AOI software as well. Therefore, within certain limits of offset, the angle and position won’t affect the test results. Moreover, our MARK points allow for fine-tuning and correction if needed.
AOI Inspection Principle: This method employs a combination of advanced algorithms, including "parameter calculation," feature extraction, and image comparison, ensuring high detection accuracy. Integrated with artificial intelligence technology, it simplifies device programming while delivering consistent, reliable performance. First, MARK points are set to precisely locate each component. Next, images are captured to establish a qualified reference template. Parameters are then defined, and the component is tested against this template. After testing, the system compares the captured test image with the reference image to determine whether the component meets the specified criteria—passing or failing.
Visual inspection is conducted in two steps: Inspection—this involves the process of calculating using our company’s proprietary algorithm; and Measurement—this refers to the specific results obtained through the algorithm. The software includes a bias correction feature, ensuring that deviations within a certain range won’t affect the test outcomes.
