Sparse Stereo

The stereo matching module is used when there is a need to generate a sparse point cloud from images. It is able to extract keypoints from images, compute A-KAZE descriptors for each keypoint, and match the keypoints by applying a Hamming distance algorithim of their descriptors. Each of these sub-modules has various knobs that can be used to fine-tune the performance of the stereo matcher.

eBusPlayer connected to a Bottlenose camera tunable knobs for keypoint extraction, AKAZE descriptors, Hamming matching, and sparse pointcloud generation

A screenshot of eBusPlayer connected to a Bottlenose camera tunable knobs for keypoint extraction, A-KAZE descriptors, Hamming distance, and sparse point cloud generation.

Feature Points and Descriptors

Bottlenose extracts key points using either the FAST9n or the good-features-to-track (GFTT) algorithm. It has built-in support for computing Akaze descriptors for these extracted key points as well. Both features are exposed as chunk data. To enable this feature, chunk data transfer has to be enabled. Set Device control -> ChunkModeActive to enable these features. Once ChunkModeActive is enabled, the chunk selectors for FeaturePoints and FeatureDescriptors can be set.

Please contact us for sample code to decode the chunk data for both features.