Training and detection methods for remote sensing image change detection models

Abstract

This application discloses a training method and a detection method for a remote sensing image change detection model. The method includes: constructing a training dataset containing sample image pairs, their location coordinates, land cover types, and labels; filtering initial negative samples in the training dataset based on constraint feature information of unchanged land cover types to construct a training negative sample set; constructing a training positive sample set using initial positive samples from the training dataset; training the training positive and negative sample sets using a neural network model; and dynamically adjusting the ratio of training positive and negative samples based on the precision and recall of the validation dataset until the model performance meets preset thresholds corresponding to precision and recall, thereby obtaining a remote sensing image change detection model. This application, by introducing a negative sample filtering mechanism and dynamically iterating the positive and negative sample ratio, effectively improves the model's generalization ability and robustness in complex scenarios, significantly enhancing the accuracy of remote sensing image change detection.

Images