Sensor Types
Differences between sensor types
3D Sensing Type | Example Sensors | Principle of Work | Limitations |
---|---|---|---|
Stereovision | Realsense D405 | 3D structure inferred from two 2D cameras separated by baseline distance. Similar to human 3D perception. | Error quickly grows with distance (quadratically). Baseline determines working range (minimal distance and accuracy). Problem with 3D estimation in texture-less area. |
Active Stereovision (pattern projection) | Realsense D435/D455 | Stereovision supported by pattern projection. Improves 3D estimation in low texture areas and supports stereo algorithms. Sensor doesn't expect particular pattern. | Supporting projection may not work at larger distances or in sunlight (in such case sensor still works like stereovision). |
Neural Stereovision | ZED 2i, ZED-X, ZED-X Mini | Stereovision supported by neural depth estimation improving accuracy. | Improved but less predictable accuracy. Typically still performs worse in texture-less areas and at larger distances. |
Time of Flight (ToF) | Kinect4A, Femto Bolt | 3D measurement based on time required for emitted light signal and its return. | May have trouble with low reflectivity (like black) objects, strong ambient light (sunlight), reflective surfaces. Sensor potentially interferes with other ToF sensors. |
Structured Light | MotionCam-3D | Sensor expects particular projected patterns and computes 3D from pattern deformation. | May have trouble with strong ambient light (like sunlight), reflective surfaces and black objects. Rapid pattern projections potentially interfere with other sensors. |