Depth Camera / RGB Camera

A depth camera is a vision sensor that outputs both a color image and a per-pixel distance to the scene (depth map). Main technologies are structured light, ToF (time-of-flight) and stereo vision; it is the core of short-range perception and semantic recognition for robots.

What problem does it solve

A 3D LiDAR answers where not to drive, but not what the object is — pedestrians, pallets, cartons, QR codes and faces require semantic recognition. The RGB stream drives face following, QR docking, pedestrian detection and object classification; the depth stream drives short-range obstacle avoidance, cliff detection and arm grasping, effectively giving the robot a 2D color image plus a third dimension of distance.

On a Tbot chassis, the depth camera is usually an addition to a 3D multi-line LiDAR: the LiDAR handles long-range mapping and large-obstacle perception, while the camera handles short-range semantics and human-interaction scenes. Indoor structured light or iToF is accurate and cheap at 0.3-3m; outdoors in strong sunlight, structured light and ToF are overwhelmed by infrared sun noise, so stereo or pure LiDAR is preferred.

How to read the key specs

Range and accuracy: structured light is accurate near range (0.2-1.5m) but fails far; ToF covers 0.3-5m; stereo reaches 10m+ but depends on texture. Choose by working distance.
Resolution and frame rate: depth resolution is usually 320x240 to 640x480 at 30fps; docking and grasping tasks need higher resolution and lower latency.
Ambient light: structured light actively projects an IR pattern; direct outdoor sunlight over-exposes it. Passive stereo is more stable under strong light.
Interface and drivers: USB3 vs Ethernet, and whether ROS/ROS2 drivers, SDK and calibration files are provided; this directly affects integration effort.

Common misconceptions

Myth 1: A depth camera can replace LiDAR. Its FOV is narrow, far-range accuracy is poor and outdoor performance is unstable; it cannot replace a 360° 3D LiDAR. It complements short-range and semantic tasks.

Myth 2: Any camera works outdoors. Structured light and iToF produce large holes in the depth map under direct sun. Outdoors choose stereo or ToF with active IR filtering.

Myth 3: Only RGB resolution matters. High-res RGB does not mean accurate depth. Look at depth hole rate, edge accuracy and drift under extreme temperatures.

Recommended models

Tbot chassis can be optioned with a depth camera for short-range obstacle avoidance, person following and QR docking: Tbot-N5-3D, two-arm wheeled humanoid (face following and grasping), cleaning robots for cliff detection. Full sensor configuration: specification overview.

Need a robot solution with visual perception?

Tell us the environment (indoor/outdoor), short-range avoidance or recognition needs and quantity; our pre-sales team replies with sensor configuration and a quote within one business day.

Get a quote

Depth Camera / RGB Camera

What problem does it solve

How to read the key specs

Common misconceptions

Recommended models

Related articles