This project builds around OpenCV library object detection based on the Scale- Invariant Feature Transform (SIFT) algorithm in a scene. The target image is preprocessed by making it grayscale, performing Gaussian blur, and histogram equalization. The images of scenes in the specified directory undergo similar pretreatments. SIFT algorithm detects keypoints and calculate descriptors from both target and scene.
The script next uses the feature matcher, FLANN, from the theory of descriptor match Lowe. It searches through scene image shapes, finding the best match iteratively based on the number of valid matches. The final output is displayed as the match of the keypoints along with the best matched scene image. The given script can be considered as a simple illustration of object recognition using SIFT and FLANN in computer vision.