When selecting a machine vision system, many people immediately ask: “I need an accuracy of 0.1 mm—what megapixel camera should I use?”

It sounds straightforward, but actually, one crucial condition is missing: how large a scope you’re looking at.

Although both have 5 million pixels, if you’re looking at just a small component, the image can be incredibly detailed. But if you’re examining a large, whole workpiece, the actual size represented by each pixel will become much larger. The camera’s pixel count hasn’t changed, yet the level of detail you can see is completely different.

So,Resolution, accuracy, field of viewThe three elements cannot be considered separately; doing so easily leads to calculation errors.

First, state the question completely.

The most common question asked on-site is: “I’d like to achieve 0.05 mm—what camera should I use?”

This question can't be answered directly, because you haven't specified the scope you're referring to.

If you look only at the 50mm-wide area, a precision of 0.05mm might not be difficult to achieve.

If you need to inspect a large workpiece that’s 300 mm wide at once, the same precision requirements will significantly increase the demands on the camera, lens, lighting, and even the installation space.

With the same precision requirement, the difficulty is completely different when placed in different fields of view.

So, before selecting a model, don’t rush to ask about the number of megapixels. First, clearly define the product dimensions, inspection area, minimum detectable defect size, installation distance, and allowable tolerance. Otherwise, the parameters you calculate might just “seem reasonable” without actually being accurate.

Resolution refers to the number of dots.

Resolution refers to the number of pixels in an image—for example, 2448 × 2048, which amounts to roughly 5 million pixels.

It tells you how many small grids this image has been divided into, but...It won’t directly tell you how many millimeters each small grid corresponds to on-site..

Many beginners directly equate “high pixel count” with “high precision,” which easily leads to bias at this stage.

What really matters is the extent of the field of view to which these pixels are assigned.

Field of view determines the actual size corresponding to each pixel.

Field of view is the actual range that a camera can capture in a single shot.

For example:

Field of view: 100 mm, image width: 2000 pixels → Each pixel corresponds to 0.05 mm.

Field of view: 200 mm, image width: 2000 pixels → Each pixel corresponds to 0.1 mm.

The conclusion is simple:With the resolution remaining constant, the larger the field of view, the larger the actual size corresponding to each pixel..

So, before asking about precision, you must first ask about the field of view. The extent of the area you want to observe determines how much detail each pixel can capture.

Accuracy ≠ Pixel Equivalent

Knowing how many millimeters correspond to one pixel is just the first step.

The actual detection accuracy is also affected by many factors:

Lens distortion

Light source effect

Edge sharpness

Workpiece attitude

Algorithm stability

Calibration method

Theoretically, one pixel corresponds to 0.05 mm, but that doesn't mean you can necessarily achieve 0.05 mm accuracy on-site.

For example, in an edge-measurement project, the theoretical pixel equivalent is 0.05 mm. However, if the edge has burrs, the light source reflects unevenly, or the workpiece is placed at an incorrect angle, the edge position extracted by the algorithm will fluctuate.

Therefore, in engineering, you can't just pat yourself on the back based solely on pixel equivalence. Pixel equivalence serves as a calculation basis—it’s not a deliverable promise.Generally, you need to allow for a margin of error; otherwise, even a slight fluctuation on site will cause the accuracy to fall short.

The three concepts must be inferred jointly.

The correct selection sequence:

1.First, determine the detection target and the area size.

2.Confirm the minimum size or defect that needs to be distinguished.

3.Calculate the approximate pixel equivalent to see if there’s sufficient headroom.

4.Verify the imaging performance by considering the lens, working distance, depth of field, and light source.

5.Use real samples for boundary testing—don't rely solely on theoretical values.

Many projects don't fail due to a lack of precision—it's that the boundaries between vision and precision weren't clearly defined from the very beginning. Even after changing the camera, the lens, and the algorithm, the system still remains unstable.

Don't turn selection into guessing parameters.

Resolution, accuracy, and field of view—these three essentially answer the same question:Within what range do you need to see things of what size?.

Think it through clearly, and you’ll have a clear direction when choosing your equipment. If you’re not clear-headed, and just focus on camera pixel counts, you’ll end up with huge images and massive files—but your project won’t be stable in the end.

So don’t just ask, “Are X million pixels enough?” You should rather ask:

How wide a range do I need to look at?

What is the minimum that needs to be distinguished?

Can it be filmed stably on site?

These conditions, taken together, represent the starting point for selection.