v1.0

Grasp-Detection-Using-SIFT-Features and Improved RANSAC

Algorithm is capable of doing the following

1. Get the 2 grasp points (for a parallel plate gripper) from the user as two mouse-clicked points on the template image.
2. Extract all the matching SIFT features from the template and the scene image.
3. Filter out incorrect matches using method 1 explained in the paper SIFT Feature Point Matching Based on Improved RANSAC Algorithm.
4. Calculate the homography transformation matrix using the correctly matched point correspondences.
5. Transform two grasp points on the template image onto the scene image.

Note

However, this improved RANSAC algorithm was also not met the requirements of the project due to its unreliable behaviour on the test images. Sometimes it gave accurate results but sometimes it gave inaccurate results. In addition to that, sometimes the matching points given by it were not enough/ not correct to determine the homography matrix accurately.

Author @bimalka98