Even after tweaking the camera settings to the max, the dark corners still won’t go away—this pitfall in visual projects might have been buried right from the start.

The machine is running, the production line is urging us on, yet the four corners of the screen remain stubbornly black and stable. The engineer stares at the screen, tweaking the parameters—each adjustment brings even more silence. The boss leans in and says, “Isn’t this just taking a picture? Why isn’t it done yet?”

This is exactly the kind of issue that makes people most anxious on-site.

It doesn’t look like much. But once you actually start running, it’s full of potholes.

In machine vision projects, many people immediately start asking about algorithms, lighting sources, exposure, and focal length.

But there’s a fundamental issue that is often overlooked:

Do the sensor size of the camera and the lens’s image circle actually match?

Choosing the wrong one not only affects image quality—it can also directly lead to dark corners, mismatched field of view, and wasted lenses.

Let’s get this matter completely clear today.

1. Don't rush to adjust the parameters—first, understand what a “target surface” is.

The so-called “target plane,” put simply, is that piece inside the camera.Image sensor.

It is responsible for photo-sensitive imaging and also determines how large a scene the camera can “capture.”

The “target area size” we usually refer to simply means the physical size of this sensor.

It will affect two key things:

First, the imaging range of the image. Second, how exactly should you choose a lens?

At many visual project sites, the problem isn't that the algorithms aren't working hard, nor is it that engineers aren't staying late.

Rather, from the very first step of selection, the direction was already off track.

Second, “1 inch” isn’t 25.4 mm—many people have fallen into this trap.

When looking at industrial camera parameters, you’ll often come across these terms:

1Inch, 2/3 inch, 1/2 inch, 1/3 inch...

Beginners easily and instinctively interpret it as:

“1 inch—that’s 25.4 mm, isn’t it?”

Sorry. In industrial cameras, it’s not calculated that way.

The term “inch” here originates from the naming convention used in the era of older-style video tubes. It’s more of a standardized term left over from historical practices.

So,The nominal size is not equal to the actual physical size..

This is also why, in some projects, even though the parameters seem similar at first glance, the field of view after installation turns out to be completely different from what was expected.

The on-site environment won't take the time to let you slowly learn from your mistakes.

Here are several common target surface specifications; please refer to the table below:

In addition to the diagonal, the target surface also comes in different aspect ratios.

For example,43, 16:9.

Different aspect ratios will result in different final image display ranges. Therefore, in a project, you can't just look at the “fraction of an inch”—you also need to consider the actual dimensions and imaging requirements.

3. A lens also has a “sensor size,” which determines whether it can cover the camera’s sensor.

The camera has a target plane, and the lens also has a corresponding target-plane size.

The lens's image circle size refers to the area that this lens can cover.Maximum sensor size for full-frame imaging coverage.

Here's the key point.

What happens if the camera’s sensor size is larger than the lens’s supported size?

The answer is very straightforward:

Black corner.

The four corners of the screen will become dark. In severe cases, the corner areas will simply become invisible. The imaging range is forcibly restricted.

Upon taking a look at the screen on-site, everyone’s first reaction might be:

“Is the exposure incorrect?” “Is the light source uneven?” “Is the lens not screwed on properly?” “Is the camera broken?”

After troubleshooting for a long time, I finally realized the root cause wasn't in the algorithm., nor in the parameters.

Instead, the lens wasn't covering the camera at all.

Conversely, if the lens’s image circle is larger than the camera’s sensor size, vignetting generally won’t occur.

But it’s not entirely without cost.

This is because the camera uses only the central portion of the lens's imaging area. The lens's imaging potential is being wasted.

So this matter has a very simple yet crucial principle:

Lens target size ≥ Camera target size

Remember this phrase—many visual projects can avoid a significant amount of detours.

4. Don’t reverse the selection process: For most projects, it’s recommended to choose the camera first and then select the lens to match.

What visual projects fear most isn't difficulty.

You got it wrong from the very beginning.

In most scenarios, the recommended order is:

First, select a camera based on the detection requirements, then match a lens according to the camera’s sensor size.

For example, you’ve chosen a device.1Inch target surfaceof the industrial camera.

That shot will have to choose support.1Inch or larger target areathe model. For example, supports1.1InchOr a lens with an even larger imaging range.

This way, you can avoid dark corners in the image.

But if you take one...2/3 inchesthe lens, to match with1InchWith that camera, the problem becomes obvious.

The lens can't cover it. Naturally, the image is incomplete.

At this point, obsessing over exposure, gain, and algorithm thresholds is essentially an effort in the wrong direction.

5. It’s okay to have a limited budget—what’s worrying is “not cutting where you should, but stubbornly cutting where you shouldn’t.”

It's perfectly normal for a project to have a limited budget.

If you are using...1/3 inchesor1/2 inchesFor a camera like this, you don't necessarily have to pair it with a large-format lens.

Not only does it waste resources, but it may also increase the size of the lens.

But conversely, if you’ve already switched to a full-frame camera, the lens must keep up.

Otherwise, the image quality will be affected.

One sentence:

No matter how good the lens is, if the target surface doesn't match, the effect will be for nothing.

It’s like buying a large-sized canvas but then using a small-sized projector. It’s not that the canvas is inadequate, nor is it that the projector isn’t bright enough. It’s simply that the two aren’t compatible.

Six, with the same 25mm focal length, why does the field of view change when you switch cameras?

Here’s another frequently asked question from the现场:

“My lens focal length hasn’t changed—so why does the field of view change after I switched cameras?”

The reason also has to do with the target surface.

For lenses with the same focal length, the angle of view will differ when used with cameras that have different sensor sizes.

For example, the same...25mm focal lengthThe lens:

Used in1Inch cameraUpward, the field of view will be wider. Used in...1/3-inch cameraUpward, the field of vision will become narrower.

This phenomenon is also often interpreted as a change in “equivalent focal length.”

So, if you’re replacing the camera model in your project, don’t just focus on whether the interface will fit.

It also needs to be recalculated:

Is the field of view sufficient? Should I change the focal length? Is the detection area off-center? Will edge information be lost?

Many on-site failures aren't due to equipment being unusable.

After the model was changed, the optical relationships were not recalculated.

7. To sum it up in one sentence: Only when the lens covers the camera can the image be complete.

The matching between the camera’s sensor and the lens’s image plane is not a minor detail that can be ignored.

It is the first step in visual system design.

It can be simply summarized in three sentences:

· The camera's sensor is too large, and the lens isn't wide enough: vignetting is likely to occur.

· The lens’s imaging surface is too large, while the camera is too small: a waste of resources.

· The target surface size is just right: the effect is more stable, and the cost is more reasonable.

Shorter version:

Only when the lens can fully cover the camera can the image be complete; only when the camera is matched with the right lens can the cost be reasonable.

Of course, in real-world projects, lens selection isn't based on sensor size alone.

You also need to consider parameters such as field of view, resolution, distortion, aperture, and interface.

But matching the target surface size is just the starting point.

If the starting point is wrong, no matter how carefully you adjust later on, you might still be taking a roundabout route.

It’s not that visual projects can’t be done—it’s often that we underestimate the complexity of the problem right from the start.

The next time you’re at the现场, the black corners are glaring, the field of view is off, and no matter how much you adjust the image, it still feels awkward—don’t rush to doubt the algorithm just yet.

Let me ask first:

Do the camera and lens really match?