obstacle detection size expansion algorithm [closed]

edit

I try to follow this paper (Monocular Vision-Based Obstacle Detection/Avoidance for Unmanned Aerial Vehicles) to do obstacle detection, I worked with SIFT algorithm in order to detect the keypoints in both images (two consecutive frames), then the Brute-Force Algorithm to match them,. Then I should extract the keypoints that its keypoints are expanding in the second image, this is by comparing the size property of the keypoints and save it in a new vector of keypoints. Finally, I want to use the functions to create the the convex hull around the final keypoints after filtering, and drawing it using findContours function. How to create this convex with key points?