Remote sensing image multi-class target counting method based on point positioning
The invention aims at the field of target counting and target positioning in remote sensing images, and discloses a remote sensing image multi-class target counting method based on point positioning. Firstly, features in a remote sensing image are extracted by using a feature extractor, then rough p...
Gespeichert in:
Hauptverfasser: | , |
---|---|
Format: | Patent |
Sprache: | chi ; eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The invention aims at the field of target counting and target positioning in remote sensing images, and discloses a remote sensing image multi-class target counting method based on point positioning. Firstly, features in a remote sensing image are extracted by using a feature extractor, then rough prediction of a target scale in the remote sensing image is completed through a scale adaptive feature enhancement module, and image features are fused and enhanced; and using a prediction head to predict the position, category, confidence and fine scale of the target in the image, and generating a candidate target corresponding to the remote sensing image. The candidate targets comprise repeated prediction of the same target, so that the candidate targets cannot be directly used as prediction results of counting and positioning of multiple types of targets. By using a non-maximum suppression post-processing means on the candidate targets, wrong prediction targets and repeated prediction targets in the candidate tar |
---|