Abstract
We describe a method of low-coherence interferometry based optical profilometry using standard light-emitting diode (LED) illumination and complementary metal-oxide-semiconductor (CMOS) image sensors. A line-field illumination strategy allows for the simultaneous measurement of many points in space. Micron scale accuracy and resolution are achieved and demonstrated using a variety of targets.
Introduction
High-precision 3D imaging, particularly Optical Coherence Tomography (OCT), has traditionally been confined to specialized laboratory settings due to the high cost of components. Standard OCT systems often rely on expensive near-infrared (NIR) superluminescent diodes or swept-source lasers. While visible-light OCT offers superior lateral resolution, it typically requires costly supercontinuum lasers and slow point-scanning mechanisms.
Researchers from Mitsubishi Electric Research Laboratories (MERL) have addressed these barriers by developing a fast, low-cost, and high-resolution 3D sensing tool. By utilizing consumer-grade components—specifically a standard green LED and a CMOS camera—they demonstrate that micron-scale 3D profiling is achievable through clever optical design and advanced signal processing.
Methodology: Parallel Spectral-Domain OCT
The team implemented a "parallel" Spectral-Domain OCT (SD-OCT) system designed for industrial surface inspection and biological imaging. The core of the system is a line-field illumination strategy, which uses a cylindrical lens to project a line of light onto the sample rather than a single point. This allows the system to capture multiple depth profiles (A-scans) simultaneously across a row of pixels on a CMOS area-scan camera, significantly increasing acquisition speed.
Key technical innovations include:
- Wavenumber ($k$-space) Linearization: Because spectrometers sample light linearly by wavelength ($\lambda$) while depth extraction via Fourier Transform requires linearity in wavenumber ($k$), the researchers used a Hilbert transform-based calibration. This re-mapping prevents image blurring at deeper axial positions.
- Total Variation (TV) Inpainting: To handle noise and low-reflectivity areas, such as the edges of metallic surfaces or biological tissues, a TV-regularized algorithm was employed. This optimization approach fills in missing depth data while preserving sharp edges and removing artifacts.
Experimental Results and Validation
The system's performance was validated using several targets, including a custom machined steel "staircase" with steps ranging from $5 \mu m$ to $100 \mu m$, various coins, and biological samples like leaves.
The experimental results highlight the system's precision:
- Axial Resolution: The system achieved a resolution of $3.8 \mu m$, remarkably close to the theoretical limit of $3.5 \mu m$ for the 530 nm LED source.
- Calibration Impact: Without $k$-space linearization, resolution would degrade significantly (from $16 \mu m$ to $62 \mu m$) as the target depth increases. The calibration method maintains a stable resolution of $\sim 3.8 \mu m$ across the entire measurement range.
- Sensitivity: The system can detect surfaces with as little as $0.7\%$ contrast, enabling the imaging of fine structures like leaf capillaries or hidden anti-counterfeit holograms on currency.
Conclusion
This research demonstrates that high-performance 3D profiling does not require expensive laser sources. By combining off-the-shelf LED and CMOS technology with robust computational compensation, the authors have created a viable tool for cost-effective industrial inspection. While currently limited by the penetration depth of visible light and some vignetting at the edges of the line-field, future iterations using higher-power LEDs could expand its utility to transparent coatings and semiconductor wafer inspection.
Summary updated