1. Introduction and Overview

The core of industrial vision inspection lies in acquiring clear and reliable images through high-performance image sensors, and achieving defect detection and dimensional quantification via precision measurement technologies. This article analyzes the key technologies from two dimensions—sensor imaging and precision measurement—and explores their applications in real-world industrial scenarios.

2. Key Technologies in Sensor Imaging

2.1 Camera Selection: Area Scan vs. Line Scan

Camera selection should be balanced between area scan and line scan cameras based on the inspection object and application scenario:

Area Scan Cameras: Capture a complete two-dimensional image in a single exposure. They are suitable for static objects or array-based inspection scenarios.

Line Scan Cameras: Acquire images line by line and are ideal for continuously moving materials such as web materials, battery electrodes, glass, and films. The line rate must be precisely synchronized with the production speed to avoid image distortion or motion blur.

2.2 Imaging Dimensions: 2D vs. 3D Vision

2D Imaging: Based on planar images, 2D vision detects surface defects such as scratches, contamination, and missing prints through grayscale or color contrast. It is mature, cost-effective, and fast (hundreds of frames per second), but lacks depth information.

3D Imaging: Uses structured light, stereo vision, or similar methods to acquire height information, enabling detection of volumetric defects, deformation, and coplanarity. It is less sensitive to lighting and color variations, with measurement accuracy reaching sub-millimeter levels. However, it involves higher system complexity, cost, and computational load.

Hybrid Approach: In practice, 2D and 3D vision are often combined—2D for rapid surface screening and 3D for inspections requiring depth information.

2.3 Imaging Spectrum and Resolution

Visible Light Imaging: The most widely used approach, suitable for most surface defect inspections.

Infrared Imaging: Capable of penetrating materials such as silicon, enabling detection of internal cracks and voids (e.g., subsurface crack inspection in semiconductor wafers).

Ultraviolet Imaging: With shorter wavelengths, UV imaging can surpass diffraction limits and enhance optical resolution, making it suitable for micro- and nano-scale structure inspection (e.g., mask defect inspection).

Super-Resolution Imaging: Uses multi-frame reconstruction or deep learning algorithms to recover high-resolution details from low-resolution images, breaking through hardware limitations of sensors.

2.4 The Critical Role of Illumination

Illumination directly affects image contrast and defect visibility:

2.5 Sensor Performance Trade-offs

Reliable detection of micro-defects requires comprehensive consideration of sensor parameters:

Resolution and Pixel Size: Determine the minimum detectable defect size.

Frame Rate / Line Rate: Must match production speed to avoid motion blur.

Dynamic Range: Enables differentiation between highlights and shadows, preventing over- or under-exposure.

Signal-to-Noise Ratio (SNR): Affects the detection of subtle intensity variations.

Spectral Sensitivity: Selection of visible, infrared, or ultraviolet bands based on defect characteristics.

3. Precision Control Technologies in Measurement

3.1 Vision Calibration: Establishing Accurate Coordinate Mapping

Checkerboard Calibration (e.g., Zhang’s Method): Solves camera intrinsic and extrinsic parameters, corrects lens distortion, and maps pixel coordinates to world coordinates.

Laser Plane Calibration: Used in line-structured light 3D systems to accurately determine the laser plane equation and convert laser pixels into spatial depth.

Multi-Camera Calibration: Aligns relative positions and orientations among multiple cameras for large field-of-view stitching or multi-angle collaborative measurement.

3.2 Image Processing and Algorithmic Enhancement

Subpixel Edge Extraction: Uses interpolation, curve fitting, or moment-based methods to improve edge localization accuracy to one-tenth of a pixel or better, reducing quantization error.

Lens Distortion Correction: Applies radial and tangential distortion coefficients obtained during calibration to ensure geometric linearity in measurements.

Deep Learning-Based Dimension Regression: Convolutional neural networks learn end-to-end mappings from images to dimensions, suitable for complex or irregular features, often combined with traditional geometric methods.

3.3 Error Compensation in 3D Reconstruction

3D measurement accuracy can be enhanced through an “error calibration–modeling–compensation” workflow:

Measure known reference artifacts to obtain spatial error distribution.

Fit an error model using polynomial functions or neural networks.

Apply compensation to measured point clouds, significantly reducing system errors (studies report diameter error reductions of approximately 52%).

4. Typical Application Scenarios

4.1 Lithium Battery Manufacturing

Challenges: High-speed electrode production (up to 120 m/min) with micro-scale defects such as burrs and coating misalignment.

Solutions:

High-resolution, high-frame-rate CXP area scan cameras achieving micron-level detection accuracy.

Combination of backlight and high-angle illumination to enhance edge visibility.

Hybrid line scan and area scan systems for high-speed scanning and multi-view inspection.

Key Factors: Sensor resolution defines the minimum detectable defect size; frame rate and synchronization determine whether production speed can be met.

4.2 PCB and Electronics Manufacturing

Challenges: Diverse defect types (missing silkscreen, solder defects), small components, and reflective surfaces.

Solutions:

High-pixel color area scan cameras with color calibration for fine print defect detection.

3D AOI systems combining RGBW illumination and stereo cameras to capture solder joint geometry.

Coaxial illumination to suppress reflections and improve contrast.

Key Factors: Camera resolution limits visible defect size; dynamic range and color/depth perception affect inspection completeness.

4.3 Automotive Component Manufacturing

Challenges: Structural defects affecting strength (burrs, short shots) and strict dimensional tolerances (hole spacing, gaps).

Solutions:

High-resolution cameras with specialized illumination (low-angle ring light, polarized light) for surface defect detection.

High-resolution cameras or structured light 3D sensors for inline dimensional measurement with accuracy down to 0.1 mm.

Key Factors: Sensor resolution and 3D accuracy directly determine compliance with tight tolerances.

4.4 Consumer Electronics Manufacturing

Challenges: Micro-scale appearance defects (cracks, bubbles) and stringent dimensional accuracy requirements.

Solutions:

High-resolution cameras combined with AI algorithms for micron-level crack detection.

Infrared imaging to assist in detecting internal bubbles and delamination in glass cover panels.

High-precision vision measuring systems for 100% inline dimensional inspection.

Key Factors: Pixel size, SNR, and spectral selection determine sensitivity to subtle defects and dimensional variations.

5. Conclusion

In industrial vision inspection systems, sensor imaging technology determines the sensitivity and reliability of defect detection, while precision measurement technology ensures accurate dimensional quantification. In practical applications, system designers must comprehensively consider defect types, dimensional accuracy requirements, production speed, and environmental conditions. By selecting appropriate camera types, resolutions, spectral bands, and illumination schemes—and combining them with advanced calibration, algorithms, and compensation techniques—efficient and reliable vision inspection solutions can be achieved.

You may contact us at chenguo@mindvision.com.cn to gain more in-depth technical insights and practical applications in the fields of machine vision and optical imaging.