Rubedo CVM

Rubedo Computer Vision Module (CVM) provides high definition images, accurate measure of the environment depth, and can extract objects of interest in real-time.

With its own powerful graphics processing unit (GPU) Rubedo CVM has been designed for the MOST CHALLENGING APPLICATIONS, including autonomous vehicle control, aerial mapping, collision avoidance, security, and surveillance.

datasheet download

Ready for industry 4.0?

Industry 4.0 revolution is all about unleashing the power of intelligent machines so they can leave their safety cages and seamlessly blend into our dynamic lives. Unless they are versatile, modular, and reusable in different scenarios, they will not comply with the new standard.

 

industry 4.0

The new generation of robotic machines strongly depend on the ability to acquire, organize, and interpret surrounding visual data in real-time in order to stay aware of environmental situation and act both safely and adequately.

 

Solution

Rubedo CVM is a light-weight long-range depth sensing system based on passive stereo vision. It outputs a hi-res side-by-side video that contains two synchronized video streams, and creates a depth map of the environment in real-time using its highly capable built-in graphics processing unit (GPU).

Rubedo CVM is ideal for robot applications because it offloads both low-and high-level image processing to its on-board processor, reserving the robot's main computing resources for robot navigation and other tasks. Specifically, Rubedo CVM handles all aspects of initial image disparity calculation, including realtime lens correction, camera rectification, correspondence searching, filtering, and error removal. Depending on application, it can perform further real-time image processing, such as obstacle detection or analysis.

 
 

features

image modalities

Multiple real-time image modalities

2D/3D video, point cloud output, etc.

ROS integration

ROS and OpenCV integration

Supports industry standard APIs

image processing

On-board image processing

Uses internal GPU for image processing

object tracking

Natural object tracking

Tracks humans, gestures, infrastructure, etc.

learning capability

Deep learning capability

Supports image recognition and tagging

benefits

Rubedo CVM is significantly cheaper, lighter, and less power hungry than conventional 3D LIDAR technology used today.

Standard version of Rubedo CVMis optimally suited to 3D depth sense in robot navigation or object sensing for high moving speeds both indoor and outdoor.

Special version of Rubedo CVM with ~2x smaller baseline is optimally suited for a variety of industrial automation applications.

application examples

Rubedo flexible software and tailor-made sensors enable 2D/3D computer vision application accross all domains.

service robotics

Service robotics

GPS-free navigation, interaction

material handling

Material handling

Navigation, manipulation

agriculture

Agriculture

Crop inspection

surveillance

Surveillance

Intruder detection and localization

manufacturing

Manufacturing

Quality inspection

healthcare

Healthcare

Navigation, interaction

technical data

technical data

Description

Default configuration is optimally suited for application in outdoor self-driving vehicles or drones with direct integration into their real-time control loop. Additional features such as object detection and tracking can easily be enabled via supplied SDK. All image processing steps are performed on the on-board processing unit.

Rubedo CVM provides both high and low level access to the device through its API: from controlling common parameters (video mode, frame rate, exposure, WB, gain) to outputting 3D point cloud (useful for visual localization, obstacle detection, and path planning for your autonomous vehicle) and detecting relevant environmental features (trees, cars, faces, etc.).

Output

Video (side-by-side)   1080p @ 30fps, 720p @60fps
3D Point Cloud   2-20fps (adjustable, affects accuracy)
Occupancy Grid   2-20fps (adjustable, affects accuracy)
Depth Map   2-20fps (adjustable, affects accuracy)

Camera

Lens   2.8mm, f/2.0
Sensor Type   1/3" @ 4M per sensor, rolling shutter
Field of View   109°/88°/72° (D/H/V)
Depth Range   1m - 25m (standard accuracy),
25m - 50m (lower accuracy)
Vision Engine   NVIDIA TX1,
ROS Kinetic Kame (Ubuntu 16.04)

Physical

Dimensions   300x160x50 mm
Stereo Baseline   250mm
Weight   <400g
Power Consumption   1,2A @ 12V
Operating Temperature   0°C to 40°C
Communication Interface   Ethernet, 802.11ac
Mounting Interface   1/4"- 20 UNC
Note: custom versions tailored to specific applications are available on demand (depth range and accuracy, IP protection level).