Stereo vision is a technique that features fast image acquisition and a large field of view. It relies on ambient light and does not require special lighting conditions. As a result, stereo vision is suitable for measuring at long distances, including outdoors. A pair of cameras is used to capture the same 2D view of the target object from two different angles. Knowing the fixed positions of the two cameras, the software compares corresponding points in the two 2D images, identifies disparities, and uses triangulation to create a dense 3D point cloud in color.
A computer vision system captures images and then manipulates, changes, or enhances them for a given application
Types of 3D vision systems
In a 3D computer vision system, the visible scene image is no longer a flat picture (2D). It’s a three-dimensional point cloud of precise coordinates where the position of every pixel in space is known, followed by optional color (texture) information. The choice of a particular technique depends on the application. Rubedos combines passive stereo and TOF for robot navigation and material handling applications.
-
Stereo vision
-
Time-of-flight (TOF)
LiDAR (Light Detection and Ranging) is depth sensing method that uses light in the form of a pulsed, invisible laser to measure distances to objects in the scene. In LiDAR, laser light is sent from a transmitter and reflected from objects to be detected by the receiver. The TOF technique is then used to create a distance map of the objects in the scene. LiDARs are very fast, but have relatively low spatial resolution and do not provide color information.
Our 3D visual perception technology
Rubedos developed a universal bolt-on visual 3D perception system Rubedos® VIPER™ – the set of intelligent ‘Eyes’ for the next generation of robots which depend on the ability to acquire, organize, and interpret surrounding visual data in order to stay aware of the environmental situation and act both safely and adequately.
Rubedos® VIPER™ system employs the stereo vision technique. It is known to be computationally most intensive, however rapid advances in multi-core processors and 3D algorithms mean that stereo vision is fully capable of keeping up with performance requirements, while enjoying benefits of the lowest costs of hardware and maintenance (compared to the other three).
VIPER continuously measures environment depth in 3D and performs application-specific object detection and classification over medium to long range distances. With its built-in powerful yet energy efficient computing unit VIPER has been designed with the most challenging industrial applications in mind where flexible robot navigation and handling is key, such as material handling. At the same time VIPER is simple to deploy and maintain.
More than just a sensor
Rubedos VIPER system inventively offloads both low- and high-level image processing to its on-board processor, reserving its host robot’s main computing resources for navigation and other useful tasks.
The system is auto-calibrated, ingress protected, easy, and has zero maintenance overhead. It is designed to be extended with standard/custom perception apps for additional application-specific image processing, such as classification of a detected object. VIPER is neutral to the environment as it uses on a non-interfering passive optical sensing.
Rubedos® VIPER™ features:
-
On-board image processing (built-in NVIDIA Jetson TX2 GPU)
-
Extensible with standard and custom perceptions apps
-
Detects & tracks people, gestures, infrastructure
-
Supports industry standard APIs
-
IP65 ingress protected, industrial grade
-
Non-interfering passive optical sensor
-
Low energy consumption
-
Requires zero maintenance
