SURF算法 Resources

Showing items tagged with "SURF算法"

This section details our implementation of the Speeded-Up Robust Features (SURF) algorithm and its adaptation for mobile devices, including key code modifications for platform compatibility. We analyze the trade-off between matching accuracy and the speed of nearest-neighbor search, demonstrating an order-of-magnitude acceleration with minimal impact on performance through optimized data structures and approximate matching techniques.

MATLAB 220 views Tagged
Download

The SURF (Speeded-Up Robust Features) algorithm, as a recently developed feature extraction method, surpasses or approaches previously proposed similar methods in three key aspects: repeatability, distinctiveness, and robustness, while demonstrating significant advantages in computational efficiency. This implementation utilizes the SURF algorithm for image detection, coordinate transformation, and image stitching. The core implementation involves using the Hessian matrix for image detection to identify feature points, followed by refinement through Fast Nearest Neighbor (NN) matching, Random Sample Consensus (RANSAC) algorithm, and Levenberg-Marquardt (LM) parameter optimization for precise feature matching. Finally, coordinate transformation is performed to unify the coordinate systems and achieve seamless image stitching.

MATLAB 224 views Tagged
Download

Image object detection using SURF (Speeded Up Robust Features) - This algorithm extracts SURF features from both source and target images by employing Hessian matrix for feature point detection. The SURF operator approximates second-order Gaussian filtering using box filters, constructing a Fast-Hessian matrix where the determinant expression is ΔH=Dxx(x)Dyy(x)-(0.9Dxy(x))². It utilizes Haar wavelet responses with Gaussian weighting to determine dominant feature orientations, resulting in a four-dimensional descriptor vector that undergoes normalization.

MATLAB 266 views Tagged
Download