|
| |
Vision Robotics Technology
Stereo Image Analysis
The images are segmented and features are matched between the images. We have incorporated several well-known methods including:
- Image Correlation - Each area in the left image that has an interesting configuration of pixels is correlated with areas in the right image.
- Edge Segmentation - A Canny edge operator is applied to both images to extract edges. Lines and corners in the left image are matched to the right using multiple criteria including Bayesian filtering.
- Spatial Matching - Image data from different camera locations is correlated to confirm matches.
Each pixel match between the left and right image can be used to calculate the point in 3-D space that corresponds to the matched object. These matches are used to create the 3-D grid. In addition line and corner matches are used to create a landmark database for localization and perimeter mapping.
|
