聚类 Resources

Showing items tagged with "聚类"

MeanShift, also known as mean shift, is widely applied in clustering, image smoothing, segmentation, and tracking. The shifted mean vector defines a family of kernel functions where the contribution of each sample's shift to the mean shift vector varies based on its distance from the shifted point. By incorporating a weight coefficient that assigns different importance to sample points, MeanSignificantly broadens its application scope. Target tracking using MeanShift is now a mature technique. Fundamentally, the MeanShift algorithm operates as a kernel density estimation method, often implemented through iterative gradient ascent to locate probability density maxima.

MATLAB 210 views Tagged
Download

Implementation of K-Means Clustering Algorithm: This algorithm partitions n objects into K clusters based on maximizing intra-cluster similarity while minimizing inter-cluster similarity. Limitations include potentially uneven cluster sizes and sensitivity to noisy data. Enhanced approach: k-medoids method selects representative objects (medoids) instead of centroids to define clusters. Implementation steps: 1) Randomly initialize K medoids; 2) Assign remaining objects to nearest medoids; 3) Iteratively optimize medoid selection by minimizing replacement cost. Code implementation typically involves distance calculations, cluster assignment loops, and convergence checks.

MATLAB 255 views Tagged
Download

Implementation of K-means clustering algorithm for image segmentation in MATLAB. The algorithm accepts parameter k as input and partitions n data objects into k clusters, ensuring high similarity within clusters and low similarity between different clusters. Cluster similarity is calculated using centroids (gravitational centers) obtained from the mean values of objects in each cluster.

MATLAB 237 views Tagged
Download

The k-means algorithm accepts parameter k as input and partitions n data objects into k clusters, ensuring high similarity within clusters and low similarity between clusters. Cluster similarity is calculated using centroid objects (gravity centers) obtained from the mean values of objects in each cluster. This MATLAB implementation uses the Iris dataset for testing and demonstrates centroid calculation, iterative assignment, and convergence mechanisms.

MATLAB 224 views Tagged
Download

This paper addresses the computational intensity and prolonged runtime of traditional FCM algorithms by proposing an enhanced FCM approach. The method involves dividing images into window-sized sub-blocks, extracting feature vectors at the sub-block level for coarse FCM clustering, followed by pixel-level feature extraction and fine segmentation specifically for edge sub-blocks. This hierarchical segmentation strategy significantly improves processing speed and segmentation accuracy through optimized computational resource allocation.

MATLAB 193 views Tagged
Download