A Comprehensive Analysis of Machine Vision Distortion: Understand It in One Article!

In the field of machine vision, distortion is a "big problem" that cannot be ignored! How much do you know about it? Today, let's have an in - depth discussion about machine vision distortion.

• First, let's talk about parallax distortion. The imaging principle of the conventional FA lens we usually use is like light rays diverging from a single point.

When photographing a three - dimensional part, if the part is not in the center of the frame but at the edge, the side of the product in the photo is likely to be distorted.

In high - precision inspection scenarios, this is simply an "accuracy killer", which can seriously affect the accuracy of the inspection results. So, how to solve this problem? The telecentric lens is the "savior"!                     

 It can shoot vertically at the product, perfectly avoiding the distortion caused by parallax. However, this lens also has a small "flaw". It needs to be as large as the object being photographed, so it is often expensive.

• Next, let's discuss installation distortion. No matter what equipment is installed, there will inevitably be some errors. In scenarios where the inspection accuracy requirement is as high as 2 microns, the installation must be kept as vertical as possible. This requires careful correction of the levelness and perpendicularity of the installation surface. Only when the imaging is accurate can the subsequent algorithms function properly. If the imaging is deviated, no matter how powerful the algorithm is, it won't work.

• There is also lens distortion. During the production process of lenses, there will be more or less some errors, and the magnitude of these errors is related to the price of the lenses. The key is that even extremely expensive lenses cannot completely eliminate these errors. Don't worry, we have visual algorithms to "save the day".

Take this calibration plate as an example. The densely packed small dots on it have extremely high dimensional accuracy. Don't underestimate this glass plate; it's worth several thousand yuan! When a camera takes a picture of it, the system can automatically calculate the distortion and then perform correction. Isn't it amazing?

These are the common types of distortion in machine vision! I hope this knowledge can give you a deeper understanding of machine vision.