Goals

• [x] Merge camera views into panoramic view
• [ ] Inference with Jetson AGX Xavier
• [ ] Get occupancy predictions for the collision risk (Crossover scenario)

Camera’s field of view update

Making the most of having two cameras, the field of view was extended to get better coverage.

Before

After

Image Stitching

The next step was to find a stitching method which takes images from each camera and combine them into a single panoramic image. The method used was developed by OpenStitching:

GitHub - OpenStitching/stitching: A Python package for fast and robust Image Stitching

It comprises 5 main tasks:

• Feature detection: Identification of keypoints in each image.
• Feature matching: Identification of corresponding keypoints.
• Transformation Estimation: Determine the transformation to align the images (Homography)
• Image Warping: Application of the transformations on the images.
• Blending: Process to smooth out the transition area.

The images below show the result of this method:

Left camera image

Right camera image

Panorama image