外观检查（异物，缺陷，缺陷）
Appearance inspection (foreign particles, flaws, defects)
Appearance inspection checks for foreign particles, flaws and defects on the surface of parts or products. Appearance inspection typically includes:
- Checking for foreign particles on food packages
- Checking for stains on cloth
- Checking for flaws on metal/resin components
- Inspection for defects, such as chipping or burrs, generated during resin/rubber molding processes
- Inspection for defects in unlit LEDs
Appearance inspection used to rely on visual inspection. Due to increased factory automation (FA), image processing systems have seen increasing use. This page describes applications of image processing that improve inspection for foreign particles, flaws and defects, and introduces the basic principle and latest trends of appearance inspection.
Basic principle of appearance inspection for detection of foreign particles, flaws and defects in images
Advantages of introducing image processing
Appearance inspection is used to find foreign particles, stain, flaws and chipping and to prevent outflow of defective workpieces. Unfortunately, visual inspection has limitations. 100% inspection consumes a large amount of labor and incurs high costs. There are also problems with variation in accuracy due to individual differences among workers and human errors.
It is difficult to find minute flaws or stains, so images must be magnified with microscopes to ensure quality. If the number of targets is limited, offline inspection with a microscope may be possible. If, however, the number of targets is in the thousands or tens of thousands, inspection requires tremendous labor, resulting in significant reductions of production efficiency. Image processing is indispensable for achieving both quality and production efficiency under these conditions.
Capable of differentiating minute foreign particles, flaws and defects
Appearance inspection used to rely on human eyes. Machine vision and image processing technologies have now developed to the point that they can detect minute foreign particles, flaws and defects. KEYENCE provides a lineup of machine vision that range from standard 0.31 megapixel models to high-resolution 21 megapixel models. We can suggest cost-effective image processing systems that meet the specific needs of customers.
Sometimes customers who are considering introducing image processing systems ask“what is the minimum size of foreign particles or black spots that can be found via inspection?”The approximate size can be calculated with the formula below.
Minimum detectable size = B / A x C
- A: Number of pixels of the image pickup device in the Y direction
- B: Field of view (Y direction) [mm]
- C: Minimum number of pixels on the image pickup device that can detect a target [pixels]
像素的数量因所使用的相机而异。例如，0.31百万像素摄像头具有480个垂直像素，而21百万像素摄像头具有4092个垂直像素。该值指的是[a：y方向上图像拾取设备的像素的数量]。
项目B，视场（y方向）可以通过更改镜头更改为10毫米，100毫米或任何其他值。项目C，图像拾取设备上可以检测目标的最小像素数，应在正常条件下设置为3像素，或在特定困难条件下的5个像素。
Under the condition above, assume that A = 480 pixels, B = 50 mm and C = 3 pixels, and the calculation is:最小检测尺寸= 50 /480 x 3 = 0.312 mmThis means that foreign particles or flaws as small as 0.312 mm can be found in the inspection.
如果您使用21百万像素摄像头，则计算是：最小检测尺寸= 50 /4092 x 3 = 0.037毫米You can find foreign particles or flaws as small as 0.037 mm, which are difficult to find visually.
When higher inspection accuracy is desired, you need to make the minimum detectable size smaller by using a high-resolution camera such as 21 megapixel type, or by reducing the viewing angle.
Inline 100% inspection is possible
Visual inspection requires offline appearance check, depending on the inspection item. Image processing, however, enables accurate inline inspection for foreign particles, flaws and defects.
Relationship between target speed and image processing speed
Intermittent feeding
These calculation formulas give the maximum number of inspections per minute and the machine vision system processing speed required by using an example of intermittent feeding, in which inspection targets are fed intermittently and the line is stopped for a specific period when each target arrives in front of the camera.
每分钟的最大检查数量= 60秒 /机器视觉系统的处理时间秒
Example: When the processing speed of the machine vision system is 20 ms
60秒/ 0.02秒(50 = 3000次/分钟times/second)
General appearance inspections can be processed within 20 to 100 ms. If you know the desired inspection speed, you can obtain the necessary processing speed with this formula:
Processing speed required of the machine vision system (ms) = 1 second/Desired number of inspections per second x 1000
Example: When the desired number of inspections is 50 times/second
1 / 50 x 1000 = 20 ms
You can use the formula above to select machine vision that can satisfy your requirements. Note that this is for the case of intermittent feeding. For continuous feeding, where products are fed constantly without stopping the line, you need to consider the shutter speed.
Continuous feeding
对于连续喂养，您需要确保对线速速度的足够快门速度。否则，捕获的图像将变得模糊，准确的检查将是不可能的。通常，设置速度，使移动的距离约为最小可检测目标大小的五分之一。
Example: the desired minimum detectable size is 1 mm and the line speed is 1 m/second
Reference shutter speed = 1 (mm) / 5 / 1000 (mm/s) = 1/5000 second
Basics of appearance inspection: Image enhance filters
In appearance inspection, you need to recognize and differentiate minute flaws and chipping. Image enhance filters play an important role in producing stable inspection results.
Shading correction
很难使用binary processing，described in the presence inspection section to find foreign particles or flaws. Shading correction cancels gloss or shadows on the target surface to allow accurate extraction of stains or flaws only.
Bidirectional smoothing filter
This filter removes hairlines or other patterns in the background, and noise. The smoothing effect can be set in the X and Y directions separately. This filter allows extraction of foreign particles only.
Blob filter
This filter appliesblob analysisdescribed in the presence inspection section to image enhancement. It allows extraction of specific factors only, while removing gloss, shadows, backgrounds, and surface unevenness.
Contrast conversion
This filter creates an image with the best contrast for each area to enhance edges or reduce background noise. Expanding the range of shade difference makes the detection of stains or other defects easier.
Practical applications
图像处理已经被有效地用于变化ous appearance inspections. Here we'll see some examples of practical applications.
Inspection for swarf remaining on pistons
Swarf remaining on pistons for car engines is hard to recognize visually and often overlooked during inspection. Image processing systems allow accurate recognition and differentiation of microscopic swarf.
Appearance inspection of chip capacitors for various defects
Appearance inspection of chip capacitors for various defects, such as stains, flaws and chips, can be completed simultaneously through the introduction of image processing. Reliable 100% inspection can be achieved and accumulated inspection data is helpful for process improvement.
在托盘上检查异物
随着食品安全引起越来越多的关注，图像处理系统的引入在食品行业的上升。先前通过样品检查检查了托盘上的异物。图像处理允许100％检查而无需额外的时间或精力。检查还可以检查和识别细微的污渍，从而有助于稳定的质量。
