Backlight illumination - Where precision starts with clear edges

In machine vision, lighting situations can be divided into two fundamental cases. Front lighting, where the object is illuminated from the camera direction, and backlighting, where the camera looks toward the light source and captures the object as a silhouette.

In this part of our illumination series, we want to focus on backlighting.

When we show engineers a backlight setup for the first time, many assume it's the simplest illumination technique out there. After all, you're just putting the object between a light source and a camera, right? It's basically shadow photography.

But here is what we have learned over years of working with measurement systems: what looks simple on paper is actually one of the most powerful tools we have in industrial image processing.

Beyond simple presence detection

Sure, backlight gets used for basic tasks like checking if a part is present, detecting positions, etc. And it does these jobs brilliantly.

Where backlight truly shines is in measurement applications. When you need to detect edges with absolute reliability and precision, nothing else comes close. That's why whenever we face a project requiring high-precision geometric measurements, backlight, if possible, is our first choice.

The proper lens: The game changer

We often use telecentric lenses, especially when the object plane is not entirely flat. Telecentricity eliminates those perspective errors that plague standard lenses. Your edges appear consistent across the entire field of view, no distortion. It also helps to get the edges more precisely, because the rays stay parallel.

But there's a catch many people miss: whatever lens you choose, your lens needs proper resolving power. It's not enough to match your sensor resolution. The lens must actually be capable of optically separating the features you're trying to measure. When you get the sensor, optics, and illumination properly matched, you can achieve measurement capability that's two to three times higher than your nominal pixel-per-millimetre resolution, that’s what still impresses me.

Real-world testing: What we've learned

In our lab, we use calibration plates etched with 0.15 µm accuracy for system analysis. There's a fundamental rule we follow: your calibration artifact should be at least ten times more accurate than your target capability. Ideally, twenty times.

Here's something that surprises many of our clients: exposure matters more than you'd think. Objects must not be overexposed. We found that slightly darker images, even ones that look odd to your eye, typically deliver more stable and repeatable measurements.

Our methodology is straightforward but thorough. We measure all known distances between calibration points across the entire image, repeating these measurements multiple times. This gives us statistically reliable data on how the system actually performs.

The numbers don't lie: Comparing backlight concepts

Using this approach, we can quantify exactly what different backlight setups deliver:

Our preferred light is directed backlight. The difference of a telecentric backlight over a directed, improves our measurement accuracy by another 0.3 µm. If you need that, perfect. If not, you can save that money.

The weakest performer? Diffuse backlight panels, especially when positioned close to the object. Edge sharpness degrades noticeably, which directly translates to reduced measurement accuracy.

The bottom line

Backlight illumination isn't a standard solution you just drop into place.
It's a precision measurement tool that demands proper design and understanding.
But when you get it right, the results go far beyond what you'd expect from pixel resolution alone.

In our world of industrial machine vision, the rule is clear:
when precision truly matters, backlight illumination is often the only viable path.

What's been your experience? Have you worked with backlight in measurement applications? We are ready to test possible solutions for you!