SMART STEREO CAMERA

The Autonomos Smart Stereo Camera is a fully-integrated, realtime-capable solution for three-dimensional object recognition. It combines efficient 3D preprocessing of image data on an FPGA with the flexibility of processing the data further with software (ARM). As an independent 3D sensor it is versatile and easily upgraded, thus forming the bedrock of our camera-based assistance systems.

« 2D | 3D »

FUNCTIONAL PRINCIPLE

Using a mathematical method, a disparity map can be calculated from two offset camera images. This depth map contains information about the distance of single pixels. The distances are color-coded in the adjacent image (red = near, blue = far). By moving the circle in the image on the left-hand side you can view the 2D and 3D view of the same situation.

FROM PICTURES TO OBJECTS

SPATIAL VISION

A disparity map can be calculated from two offset camera images (similar to human spatial vision).

3D POINT CLOUD

The disparity map can be transformed into a 3D point cloud with depth information for improved processing.

ANALYSIS

The computer detects connected structures in the point cloud and can localize objects and distinguish them from each other.

SMART STEREO CAM – TECHNICAL SPECS

  • Dual core ARM processor with Autonomos Software Framework
  • Image processing pipeline and dense stereo image processing on an FPGA
  • Global shutter HDR sensors with 752×480 pixels and up to 60 FPS
  • Fast point cloud diagram calculation (latency < 1 frame)
  • Low energy consumption: 5W
  • Dimensions: 100mm x 40mm x 30mm (variable baseline)
  • Interfaces: Gigabit Ethernet, USB, CoaXPress

COMBINING HARD- AND SOFTWARE

EFFICIENT PREPROCESSING ON HARDWARE (FPGA)

  • Image capturing (up to 60 FPS)
  • Debayering color images
  • Removing lens distortion
  • Rotating the image layers in 3D
  • Image rectification (line-by-line image arrangement)
  • Separable 2D image convolution
  • Recognizing and describing distinctive features
  • Calculating a dense disparity map
  • Providing a 3D point cloud of a scene

FLEXIBLE POSTPROCESSING IN SOFTWARE (ARM)

  • Object recognition and tracking
  • Calculation of movement maps and areas of navigability
  • Automatic post-calibration (extrinsic parameters)
  • Visualization of 3D point clouds
  • Visual SLAM
  • Defect recognition

ASK A QUESTION OR
SEARCH FOR A CUSTOMIZED SOLUTION

CONTACT US