A computational method for rapid orthographic photography of lake sediment cores

Abstract

AbstractPhotographs of sediment cores are an important dataset, often containing visual evidence for environmental change via sediment composition and structure. These photographs may be used to stratigraphically correlate adjacent cores or for automated image analysis, and can facilitate collaboration amongst researchers through sharing and annotation of the image files. Here we describe the use of computational photogrammetry (also known as Structure from Motion–Multi-View Stereo) to generate orthographic imagery of sediment cores. Computational photogrammetry is a rapid and economical technique, typically requiring only a few minutes for each metre of core, using consumer-grade digital camera equipment. The photogrammetric methodology corrects for topographic distortion caused by sediment surfaces that are not perfectly flat, and can also record features of the scene surrounding the core, such as notes, colour reference cards and measurement tapes or rulers. As the photogrammetric process also generates a three-dimensional reconstruction of the sediment core, spatial-based analysis can be used to identify damaged or non-representative sections of the core that are to be avoided during image analysis. Using an intermittently laminated sediment core from Lake Surprise, Australia, we tested 22 scenarios using control points in a variety of configurations, as well as calibrated and uncalibrated cameras, to identify techniques that can reconstruct the core accurately and generate orthophotos. Multiple techniques were able to achieve suitable accuracy. In particular, targets placed on the table alongside the core, combined with a calibrated camera, achieved high accuracy and enabled a simple, rapid, and repeatable method for generating high-quality sediment core images.