Discrimination of river sandbanks for sand mining in high-mineral regions using multispectral images

Sand mining is a booming industry with significant economic and environmental implications. River sandbanks serve as critical sources for sand mining. However, unregulated sand mining poses ecological risks, necessitating precise identification of potential river sandbank regions to balance economic and environmental priorities. Satellite remote sensing enables non-invasive, large-scale detection of river sandbanks through their critical mineralogical and geomorphological features. In the past, semi-supervised and supervised techniques have been used to detect mining regions including sand mining. A few techniques employ multi-modal analysis combining different modalities such as multi-spectral imaging, synthetic aperture radar (SAR) imaging, aerial images, and point cloud data. However, the distinguishing spectral characteristics of river sandbank regions are yet to be fully explored. Further, such techniques may face limitations in scalability and dependency on labeled datasets. This paper provides a novel method to detect river sandbank regions for sand mining using multi-spectral images without any labeled data, i.e. in unsupervised manner, over the seasons. It specifically addresses the challenge of distinguishing them from spectrally similar high-mineral regions e.g., overburden dumps. The two discriminative features of river sandbank regions can be summarized as: proximity to river channels and unique mineral composition. The proposed work uses these distinguishing features to determine the spectral signature of a river sandbank region, which is robust to other high mineral abundance regions. It follows a two-step approach, where first, potential high mineral regions are detected and next, they are segregated using the presence of a river stream. Morphological operation over a water body index is employed to detect the presence of river stream. Primarily, Landsat 8 data is used for experimentation. The proposed technique provides average accuracy, precision, and recall of $$90.75\%$$ , $$85.47\%$$ , and $$73.5\%$$ , respectively over the seasons from Landsat-8 images without using any labeled dataset. By eliminating dependence on annotations or multi-modal inputs, the proposed work offers a cost-effective, scalable solution to detect sandbanks in dynamic riverine environments.