In modern manufacturing, dimensional quality control is a critical factor that determines product reliability and brand reputation. Traditional measurement methods using calipers, micrometers, or dial indicators are still in use, but they reveal many limitations: low speed, dependence on the operator, the risk of scratching surfaces, and incompatibility with automated production line speeds. Measurement using camera vision (machine vision metrology) has emerged as a superior alternative, allowing inspection of thousands of parts per minute with accuracy down to a few microns and without any physical contact.
BeeVision is a consulting and implementation partner for camera vision metrology systems in Vietnam, with deep specialization in 2D and 3D solutions built on the Cognex platform — a world-leading brand with more than 1.6 million systems installed worldwide.
Operating Principles of Camera Vision Metrology
A camera vision metrology system uses an image sensor combined with specialized lenses and image-processing algorithms to determine the size, position, shape, and profile of an object. The entire measurement process is non-contact — meaning the camera only captures images, and the software calculates the actual dimensions based on a previously established pixel-to-real-world calibration factor.
Key Advantages Over Traditional Methods
- Non-contact: Eliminates the risk of scratching or deforming parts — particularly important for electronic components, semiconductors, and pharmaceuticals.
- High speed: Capable of measuring hundreds to thousands of parts per minute, ideal for 100% inline inspection on production lines.
- Micron accuracy: With sub-pixel edge detection algorithms and telecentric lenses, errors can be smaller than 1 µm over a 50 mm field of view.
- High repeatability: Removes the subjective influence of the operator — every measurement yields the same result if conditions remain unchanged.
- Digital data storage: Results are automatically saved to a database, supporting traceability and SPC (Statistical Process Control) analysis.
2D Vision Metrology Systems
2D measurement is the most widely adopted approach. It uses a single camera to measure features in the X-Y plane such as outer diameter, width, length, distances between holes, angles, and positions. This is the optimal choice for inspecting flat parts or features that lie within the same observation plane.
Sub-Pixel Edge Detection Algorithms
The heart of 2D measurement is the algorithm that detects edges below the pixel level (sub-pixel). Instead of locating boundaries on a discrete pixel-by-pixel basis, the algorithm interpolates the brightness gradient to determine edge positions with accuracy down to 1/10 or 1/20 of a pixel. Combined with a high-resolution sensor and proper calibration, the system can achieve measurement accuracy of just a few microns.
The Role of Telecentric Lenses
Telecentric lenses are an indispensable component in high-precision measurement applications. Unlike conventional (entocentric) lenses, which introduce perspective distortion when an object moves along the Z-axis, telecentric lenses maintain a constant magnification regardless of distance — a property that completely eliminates errors caused by changes in part height.
| Optical Configuration | Accuracy (Trueness) | Repeatability | Typical Applications |
|---|---|---|---|
| Standard lens + CMOS camera | 5 – 20 µm | ±0.02 mm | Visual inspection, code reading, positioning |
| Telecentric lens + coaxial illumination | < 1 µm over a 50 mm FOV | ±0.01 – 0.05 mm | High-precision dimensional measurement, mechanical machining |
| Cognex In-Sight L38 3D system | ±5 µm | High | Profile measurement, flatness, assembly |
| Laser profiler | A few µm on the Z-axis | Very high | Welds, warpage, complex profiles |
3D Vision Metrology Systems
When parts have complex profiles with multiple height levels, or when measurements are required for features such as flatness, warpage, volume, and weld profiles, 2D measurement is no longer sufficient. This is where 3D Vision systems shine — the camera reconstructs a point cloud containing the full X, Y, and Z coordinate information of the part’s surface.
Cognex In-Sight L38 — A 3D Camera with Integrated AI
Cognex In-Sight L38 is the world’s first 3D vision system with artificial intelligence (AI) integrated directly on the device. The product combines deep-learning-based detection of subtle features with high-precision measurement, enabling solutions to problems that previously required multiple separate devices. The In-Sight L38 is especially valuable in applications such as:
- Measuring the height and flatness of electronic boards, connectors, and IC pins.
- Inspecting weld and adhesive bead profiles in automotive and electric vehicle battery assembly.
- Locating 3D parts for vision-guided robotics during pick-and-place operations.
- Measuring volume, warpage, and coplanarity of complex mechanical components.
Common 3D Measurement Technologies
- Laser triangulation: Projects a laser line onto the surface, and the camera measures the displacement to calculate height — Z-axis resolution can reach the micron level.
- Structured light: Projects a structured light pattern and analyzes its deformation to construct a point cloud — fast and well-suited to complex parts.
- Stereo vision: Uses two cameras, like human eyes, to calculate depth through triangulation.
- Binocular telecentric stereo: Combines two telecentric lenses to achieve high accuracy across all three axes.
The Cognex Ecosystem in Vietnam
Cognex is an American brand with more than 36 years of in-depth experience in machine vision. In Vietnam, Cognex products are being widely deployed across the automotive, electronics, semiconductor, medical, and food and beverage industries. Some of the most representative product families include:
- In-Sight 2000 series: Cost-effective 2D vision sensors suitable for basic inspection applications.
- In-Sight 9000 series: High-resolution 2D cameras for high-precision measurement applications, achieving up to 0.01 mm accuracy.
- In-Sight L38: A 3D camera with integrated AI — an all-in-one solution for the smart factory.
- VisionPro & Designer: Professional vision development software with more than 100 algorithm tools.
How to Evaluate a Measurement System
Even the best camera vision system must be quantitatively evaluated before being put into production. BeeVision always applies international standard evaluation procedures to ensure that the system meets the customer’s exact technical requirements.
1. Calibration
The first step is to convert pixel coordinates into real-world units (mm or µm) by means of a calibration target. This process simultaneously corrects optical distortions such as radial distortion and trapezoidal (keystone) distortion — errors that can persist even with telecentric lenses (around 0.1% inherent distortion).
2. Type I Gauge Study (Bias Evaluation)
A single operator measures the same calibrated reference part of known size 30 times in a row to determine the system’s bias and resolution. This is the most fundamental check to ensure that the system measures the true value accurately.
3. Gage R&R (Repeatability & Reproducibility)
The standard Gage R&R method requires 3 operators to measure 10 different parts, with each part measured 3 times. The variance analysis of the results helps to assess:
- Repeatability: The intrinsic error of the equipment — same operator, same part.
- Reproducibility: The error caused by different operators (for automated systems, this factor is nearly zero).
- %GR&R: The ratio of measurement error to the product’s tolerance — typically required to be < 10% for the system to be approved for production use.
Real-World Applications of Camera Vision Metrology
- Automotive industry: Measuring pistons, camshafts, gaskets, body welds, and wiring harness connectors.
- Electronics & semiconductors: Inspecting IC pins, bumps, PCB coplanarity, and wafer dimensions.
- Pharmaceutical industry: Measuring pill sizes, injection vials, and liquid fill levels.
- Precision mechanical engineering: Measuring cutting tools, gears, and bearings to micrometer-level tolerances.
- Packaging industry: Inspecting bottle dimensions, cap sealing integrity, and label positioning.
BeeVision — Consulting on the Optimal Measurement Solution
Every measurement challenge has its own requirements for accuracy, speed, environment, and budget. Choosing between 2D and 3D, between standard and telecentric lenses, or between area-scan and line-scan cameras requires hands-on experience and a deep understanding of optics, mechanics, and software. BeeVision partners with customers from the initial survey all the way to stable system operation:
- Analyzing technical requirements and defining FOV, target accuracy, and production line speed.
- Recommending the right hardware configuration — camera, lens, illumination, and mechanical fixtures.
- Developing measurement algorithms and integrating with the factory’s PLC and MES systems.
- Performing calibration, Type I Gauge Study, and Gage R&R to demonstrate performance.
- Providing operator training, maintenance, and technical support throughout the system’s lifecycle.
With a team of experienced engineers and a strategic partnership with Cognex, BeeVision is committed to delivering camera vision metrology solutions that are highly reliable, cost-effective, and tailored to manufacturing conditions in Vietnam.
Contact BeeVision:
Website: bee.vision
Email: hive@bee.vision