选择正确的检查工具
在引入机器视觉之前要检查的主要物品
-
- 选择检查所需的设备
- Choose the correct devices that meet the inspection requirements.
- 相机/控制器/照明/镜头/显示器
-
- Sensing and judgment
- 对实际目标进行测试。
-
OK和NG产品的参考零件
检查周期时间
各种检验物品
-
- 选择安装位置和过程
- 查看特定的安装位置。
-
摩托/固定的目标
环境条件，包括环境光和振动
-
- Controls for automation
- 查看I / O控件。
- Image capture timing / Judgment output / PLC control / Data output
-
- On-site testing
- 测试实际生产线。
-
精细设置调整
Statistics
I / O控制检查
-
- Understanding basic operations
- 基本设置程序维持稳定检查。
- 设置公差/灵敏度调整/更改检验设置/项目注册
检测判断：确定检查是否实际上是可能的。
Preparation required for judgment
准备几个有缺陷和非缺陷工件的样本。
To confirm the detectability with machine vision, it is effective to test it by using limit samples of defective and non-defective workpieces.
准备几种类型的限制样本将使结果更接近实际线路的结果。
Differentiation of MUST and WANT
对于一些检查,您可以设置明确的排名for the detectability by differentiating MUST items that must be detected and WANT items that are preferably detected.
![[无缺陷工件] [有缺陷的工件（必须）：切削/毛刺/变形/白色污渍] [有缺陷的工件（想要）：切削（小）/毛刺（小）/变形（小）/白色染色（小）]](http://www.lambarine.com/Images/ss_visionbasics_primer3_img03_1694745.jpg)
When a specific quantitative difference is set between MUST and WANT, it is easier to judge the stability of detectability. For example, when you inspect a target with a size of 20 mm in the field view of 25 mm with a megapixel camera, one pixel measures 0.025 mm. On the assumption of one pixel as the minimum unit, the theoretical detection limit under this condition is 0.025 mm. In reality, however, there are various extraneous conditions so you need to allow some margin for the detection limit.
在上面的示例中，当必须级别的深度为0.5mm时，可以将其识别为20像素的变化，并且可以尽可能判断检测。当要级碎裂的深度为0.05mm时，它是2个像素的变化，并且可以假设切屑接近检测限。
MUST-level chipping: 0.5 mm → reliable detection
想要水平碎屑：0.05 mm→检测限
Checking the judgment tolerance and margin
判断检测稳定性的可靠方法是检查以多种缺陷和非缺陷工件测量的值的统计数据。下图是使用CV系列机器视觉的统计分析功能测量256检测目标的结果。超过上限线的目标被检测为NG。该图显示了基于上限设置检测到工作原理的水平。
When tolerance (upper limit) is set with sufficient margin
The average of the measured values is about 6.3. The red circles represent MUST-level defective workpieces and all of them well exceed the upper level.
The blue and green circles represent WANT-level defective workpieces. When the upper limit is set to 17.0, the workpieces in blue circles can be detected as NG.
虽然无法使用此设置检测到绿色圆圈中的想要级别有缺陷的工件，但没有机会检测不正确的工件。
When the upper limit value is set more severely, from 17 down to 11, the number of non-defective workpieces reduces from 250 to 244, showing a reduction in yield rate.
当严重的耐受设置以尽可能地检测到想要级别的缺陷工件
The attempt to lower the upper limit further to detect green circles representing WANT-level defective workpieces may also detect non-defective workpieces as NG because the limit coincides with the maximum value of the fluctuation of non-defective workpieces.
这个例子表明我有缺陷的工件n green circles are at the boundary with non-defective workpieces.
Takt time and image processing time 1
Every time you use machine vision for inspection, you have to think about the processing speed of the system. The latest machine vision technology is capable of ultra high-speed processing and can inspect up to 100 targets per second depending on the inspection details. Note, however, that the image processing time varies greatly depending on the number of pixels of the camera, processing details, the number of processing items, etc. It is important to check the processing speed of the machine vision system and the takt time of the inspection line.
机器视觉处理流程
The following shows the flow of the inspection using machine vision.
![[触发输入>图像捕获/传输>图像处理>判断输出>控制] [机器视觉处理速度图像捕获/传输：几十多个MS /图像处理：几到几百个MS [当转换为检查周期时时间几百到几千个目标/分钟]](http://www.lambarine.com/Images/ss_visionbasics_primer3_img06_1694748.jpg)
Takt time and image processing time 2
The minimum detectable resolution varies depending on the number of pixels of the camera used. As the number of pixels of the camera increases, the resolution also becomes higher, but the processing time increases as well. The following is an example of detection of black spots on a container when 0.31, 2, and 5 megapixel cameras are used. When the number of pixels in the binary images captured with these cameras are compared over the same field of view, there are great differences in the number of detecting pixels. This means that cameras with a higher number of pixels can provide more detailed detection. On the other hand, cameras with a higher number of pixels require longer processing time.
*下面所示的处理时间和像素数是典型的示例。处理时间是最短的触发间隔。
