Automated feature extraction for prospection and analysis of monumental earthworks from aerial LiDAR in the Kingdom of Tonga

Abstract

Recent LiDAR (Light Detection and Ranging) survey in Tonga has documented a dense and complex archaeological landscape, particularly on the principal island of Tongatapu. Among the features revealed by the LiDAR are a profusion of earthen mounds, most of which are associated with residence, sporting, or burial in the period 1000–1850 CE. For identification and mapping of the mounds we use and evaluate two automated feature extraction (AFE) techniques, object-based image analysis and an inverted pit-filling algorithm (“iMound”). Accuracy of these methods was measured using an F1-score (harmonic mean of precision and recall). Variable AFE results indicate that continual and iterative fine-tuning is required. Successful mapping of some 10,000 mounds on Tongatapu reveals a distinct spatial structure that relates to traditional land division and tenure.