Stereo Vision

1. Principle

Stereo vision mimics the visual mechanism of human eyes. Two cameras placed at a fixed baseline capture images of the same object simultaneously. By matching corresponding points in the two images and analyzing their positional differences (disparity), the 3D coordinates of each point can be computed using triangulation.

2. Key Technologies

Camera Calibration: Determines intrinsic parameters (focal length, principal point, distortion coefficients, etc.) and extrinsic parameters (relative position and orientation between cameras).

Feature Extraction and Matching: Extracts distinctive and repeatable features from the left and right images, then finds correspondences between them.

Disparity Calculation and 3D Reconstruction: Computes disparity based on matched features and reconstructs the 3D shape of the object.

3. Advantages

Intuitive principle with relatively low cost.

Capable of capturing 3D information of large scenes.

4. Limitations

Highly dependent on accurate feature matching; may fail in low-texture or repetitive-pattern regions.

High computational complexity may affect real-time performance.

5. Applications

Robot Navigation: Helps robots perceive 3D structures of the environment and plan paths.

Autonomous Driving: Identifies the 3D positions and shapes of roads, vehicles, and pedestrians for safe operation.

3D Reconstruction: Builds 3D models of buildings, landscapes, and other large structures.

II. Structured Light

1. Principle

A predefined pattern—such as stripes or coded textures—is projected onto the object surface. The pattern deforms according to the object’s shape. By analyzing the deformation, depth information of each surface point is computed to reconstruct the 3D geometry.

2. Common Types

Fringe Projection: Projects a series of parallel fringes and computes depth from fringe deformation.

Coded Patterns: Techniques such as Gray code and phase-shifting encode and decode spatial information for depth acquisition.

3. Advantages

High measurement accuracy, especially for close-range and small-area scanning.

Insensitive to object surface color and texture.

4. Limitations

Sensitive to ambient light interference.

Limited measurement range.

5. Applications

Industrial Inspection: Measures 3D dimensions and shape deviations of components.

3D Scanning: Quickly captures high-precision 3D models of human bodies, objects, etc.

Face Recognition: Obtains accurate 3D facial geometry for improved security and accuracy.

III. Time-of-Flight (ToF)

1. Principle

A ToF sensor emits light pulses toward the target and measures the time it takes for the light to return after reflection. Using the known speed of light and the measured time, the distance to each point is calculated to generate 3D information.

2. Advantages

Captures depth information of an entire scene rapidly with high real-time performance.

Large measurement range.

3. Limitations

Depth accuracy is relatively lower and decreases with distance.

Sensitive to ambient light and multipath reflections.

4. Applications

AR/VR: Enables real-time scene perception and interactive experiences.

Logistics & Warehousing: Quickly measures the volume and position of goods.

Autonomous Driving: Provides distance perception of surrounding objects.

IV. Laser Triangulation

1. Principle

A laser beam projects a spot or line onto the object surface. A camera observes the position of the spot on the image plane. The height or depth of the object surface is calculated based on the geometric (triangular) relationship among the laser source, the object, and the imaging system.

2. Advantages

High measurement accuracy, suitable for precise surface profile inspection.

Relatively simple structure and easy to implement.

3. Limitations

Typically limited measurement range.

Performance depends on the reflectivity and roughness of the object surface.

4. Applications

Surface Profile Measurement: Such as inspecting surface roughness of mechanical parts.

Inline Inspection: Real-time monitoring of product dimensions and shape variations on production lines.

These 3D machine vision methods each have their own strengths and limitations. In practical applications, the optimal method—or a combination of multiple methods—should be selected based on specific requirements and environmental conditions.

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.