A Fast Recognition Method for Space Targets in ISAR Images Based on Local and Global Structural Fusion Features with Lower Dimensions
Algorithm 1. Isometric feature mapping (ISOMAP).
Input: The high-dimensional feature data set , where represents the th vector in the data set , which is spliced by the trace feature vector and the LBP feature vector, the neighbor parameter , and the fusion feature dimension
Process:
1: Fordo
2: Determine the nearest neighbor vertexes of ;
3: Set the distance between and the nearest neighbor vertexes to the Euclidean distance, and the distance to other vertexes is infinity;
4: End for
5: Call the shortest path algorithm (Floyd algorithm or Dijkstra algorithm) to calculate the distance between any two vertexes;
6: Input into the MDS algorithm;
7: Return the output of the MDS algorithm.
Output: Low-dimensional fusion feature of the high-dimensional feature data set .