Novel Integrated Conv Siamese Model for Land Cover Change Detection

Change detection, a crucial undertaking in remote sensing, involves the identification and characterization of disparities among multiple images of a given location obtained at distinct time intervals. Siamese networks have emerged as a promising approach for change detection, leveraging their ability to learn discriminative features and capture complex relationships between image pairs. This research proposes a Siamese-conv-net architecture for change identification that learns a shared representation for image differences. The CLCD (Cropland Change Detection) dataset with very high-resolution images is used for this work. These images are captured via GeoFen-2 satellite and are divided into 600 images of 512 \(\,\times \,\) 512 with 0.5 m resolution. The proposed method has given the outstanding f1-score of 76.6% and precision and recall of 78% and 75.3%. Experimental results on the benchmark dataset CLCD demonstrate that the proposed Siamese-conv-net achieves competitive performance in change detection, outperforming traditional methods.