How Active Stereo 3D Vision Supports Reliable Perception in AMRs and ASRS

Autonomous mobile robots, AGVs and automated storage and retrieval systems depend on 3D perception while moving through indoor spaces where routes, obstacles and nearby objects can change continuously. Navigation, spatial mapping and obstacle detection rely on depth data that remains consistent as the platform moves and the scene changes around it.

NetraVis, e-con Systems’ 1MP active stereo depth camera, is perfect for these robotics use cases. It combines RGB imaging, motion data and depth accuracy while handling depth processing within the camera before passing the resulting data to the Linux host.

In this blog, you’ll find out which 3D vision qualities are critical for AMR/ASRS and how NetraVis leverages RGB data, motion sensing and on-camera AI to enable depth imaging.

What AMRs and ASRS Need From 3D Vision

An indoor mobile platform may encounter a pallet close to its chassis and a route boundary several meters ahead. Both need accurate distance data. The readings also need to remain comparable over time, even when the platform or another object is moving.

Reflective and dark materials introduce another test, while the resulting depth map needs enough spatial detail for navigation and mapping.

Depth coverage

AMRs and AGVs may encounter objects at very different distances within the same route. NetraVis covers that operating area through three related parameters:

• 38 m to 6 m range in full-resolution mode
• 88° × 68° depth field of view
• 1080 × 720 depth output at 30 fps

The depth range covers objects close to the platform as well as those farther along its path. The field of view determines how much surrounding space is captured in each frame, while the 30 fps update rate helps the control software track changes in obstacle position as the robot moves.

Repeatability of depth accuracy

Navigation and positioning depend on depth readings staying consistent when the scene itself has not changed. NetraVis maintains depth deviation within 0.5% at 1 m, while high repeatability keeps successive measurements closely aligned under the same conditions. It gives the application a stronger basis for treating a change in measured depth as actual movement rather than variation in the sensor output.

Depth during movement

Motion affects the images used for stereo comparison. NetraVis depth sensors leverage reliable depth during movement, use global shutter mode and stereo synchronization, as well as expose the entire frame at once. This helps avoid spatial skew from rolling-shutter readout. The paired images retain the relative geometry needed for stereo matching when the robot or objects in view are moving.

That synchronized left-right pair becomes the input for depth calculation in the ASIC.

Depth on difficult surfaces

Reflective and low-texture surfaces can challenge indoor stereo depth sensing by making it harder to match features between the left and right images. NetraVis uses IR-assisted active stereo to add texture for depth matching on surfaces with limited visual detail.

Basically, black color alone isn’t the issue; the texture, lighting and IR reflectivity are also crucial.

Detailed spatial information

Distance data becomes useful to an AMR or ASRS once the application can turn it into a usable view of the surrounding space. NetraVis provides a 16-bit Z-depth map in millimeters for direct range measurements, while the SDK can also generate disparity, point-cloud and colorized-depth outputs.

Point clouds can be used for 3D reconstruction, while the other formats give developers different ways to work with or visualize the same depth information.

Bringing Active Stereo Into AMR and ASRS Development

Active stereo is valuable in AMR and ASRS environments because indoor routes can include dark areas, low-texture surfaces and objects that are difficult to distinguish through passive stereo alone. By adding active illumination to stereo depth sensing, NetraVis can generate accurate depth information for navigation, obstacle detection and spatial mapping under such conditions.

NetraVis_USB connects to a Linux host through USB 3.0 Type-C, giving the robotics application access to depth, RGB, IMU, point-cloud and inference data through the camera SDK. These outputs can feed different parts of the perception stack depending on the task. An upcoming GMSL2 variant will further expand connectivity options.

Depth data can enable route planning and obstacle detection, while RGB adds scene context and IMU readings help account for camera motion during movement.

Developers can incorporate point-cloud data where a 3D representation of the surroundings is required or use camera-side inference when object information needs to reach the host alongside the vision streams. ROS 2 packages also provide a path into robotics applications built around ROS-based navigation and perception components.

Maximize 3D Robotic Perception with e-con Systems’ Latest Camera

e-con Systems has been designing, developing, and manufacturing embedded vision solutions, from custom OEM cameras to complete ODM platforms, since 2003. Our vision solution, NetraVis, is a 1MP active-stereo depth camera for indoor robotics and automation applications.

Utilize our Camera Selector to easily explore our complete portfolio.

Looking for the right camera for your robotics application? Please write to camerasolutions@e-consystems.com to speak with our vision experts.

FAQs

1. Why is active stereo vision suited to AMRs and ASRS?
   Active stereo vision gives AMRs and ASRS depth information for navigation, spatial mapping and obstacle detection as they move through changing indoor environments. NetraVis_USB combines global-shutter stereo sensing with on-camera depth processing to maintain usable spatial data during movement.

2. What depth range does NetraVis_USB cover?
   NetraVis_USB covers a depth range of 0.38 m to 6 m in full-resolution mode, with an 88° × 67° depth field of view and 1080 × 720 depth output at 50 fps. These parameters cover nearby objects and objects farther along the robot’s route.

3. How does NetraVis_USB maintain depth data during movement?
   Its depth sensors use a global shutter, exposing the full frame at once and avoiding spatial skew associated with rolling-shutter readout. This preserves the image geometry required for stereo matching while the robot or objects in view are moving.

4. How does NetraVis_USB combine depth, RGB and motion data?
   A separate 1MP RGB sensor adds visual context to the depth stream, while Depth-to-Color registration maps spatial measurements to corresponding positions in the color image. A 6-axis IMU provides motion data, giving the application access to depth, visual and movement information through the SDK.

5. How can NetraVis_USB fit into an AMR or ASRS software stack?
   NetraVis_USB connects to a Linux host through USB 3.0 Type-C and provides depth, RGB, IMU, point-cloud and inference data through its SDK. ROS 2 support gives robotics applications a path to use these outputs within navigation and perception components.