ArqPy: A Python Toolbox for Remote Sensing Image Preprocessing and AI-Assisted Interpretation of Derived Products for Archaeological Prospection

Abstract

Highlights What are the main findings? ArqPy uses an object-oriented architecture that generates suitable derived products from WorldView-3 imagery for initial archaeological assessment while facilitating the future integration of additional satellite sensors. ArqPy integrates AI-based tools for complementary tasks, including MAE feature analysis of derived products, Z-PNN pansharpening, and SAM 3 text-guided segmentation of candidate crop marks. What are the implications of the main findings? The workflow improves reproducibility and comparability in archaeological remote sensing by making preprocessing and enhancement steps explicit and easily deployable on standard Windows systems. ArqPy generated complementary WorldView-3-derived products that supported expert identification of candidate crop marks, while SAM 3 provided text-guided segmentation for preliminary detection of crop-mark-like features. Abstract Remote sensing is widely used in archaeology, but the lack of standardised and readily deployable preprocessing workflows limits reproducibility and cross-study comparability, particularly for very high-resolution multispectral imagery. This study presents ArqPy, a Python toolbox designed to automate and standardise image preprocessing, enhancement and initial interpretation for archaeological prospection. The toolbox includes atmospheric correction, conventional pansharpening, spectral indices, principal component analysis (PCA), and spatial filtering. Its object-oriented architecture currently supports WorldView-3 (WV3) and WorldView Legion (LEGION) imagery and facilitates the future integration of additional sensors. ArqPy also incorporates complementary AI-based tools: Masked Autoencoder (MAE) feature analysis for exploring and ranking derived products, Z-PNN for deep-learning-based pansharpening, and SAM 3 for text-guided segmentation of candidate crop marks. The toolbox was applied to the preliminary inspection of crop marks at the Zar Tepe archaeological site in southern Uzbekistan before the 2026 field campaign. This case study illustrates the proposed workflow and demonstrates how ArqPy provides a reproducible and readily deployable environment through which archaeologists can access advanced image-processing methods. AI-based tools can support preliminary reconnaissance, but they cannot replace expert interpretation without further training and validation using site-specific archaeological data.