UTILIZING IMAGE RECOGNITION TECHNOLOGY FOR FORAMINIFERAL ASSEMBLAGE ANALYSES

Analyses of foraminiferal assemblages involve repetitive microscopic assessment of sediment samples. Because long hours of assessment are required, many samples are never processed. Image recognition software is widely used in a wide range of contexts from law enforcement to medical research. The software systematically matches features within sample images against an image library. Presently, most medical and oceanographic applications utilize flow-through systems in which samples are suspended in water or similar liquid and pass through a beam of light where the images are captured. To directly apply flow-through technology to sediment analyses, one challenge is finding an appropriate liquid matrix that will suspend and distribute the sediment so that individual particles can be identified; this necessity would raise costs and potentially produce hazardous waste. Moreover, flow-through systems are primarily designed to use transmitted rather than reflected light. Identification of foraminifera generally utilizes reflected light, because most shells are relatively opaque. Our strategy is to directly image the sediment samples using reflected light, and then apply recognition software to the sample images. A library of high quality digital images to be utilized by the identification software can be developed by photographing foraminifera identified conventionally from samples of interest. A longer-term goal would be to develop a set of library images for specific regions and depth ranges. Foraminiferal research has the advantage of extensive publications and reference collections upon which to base regional libraries. Because the foraminiferal assemblages of the Gulf of Mexico are well known, sediment samples from the west Florida shelf are being used to develop a reference set of images for assemblage analyses. Recognition software will then be trained to automate assemblage counts and those results will be compared with results from traditional picks and counts. Other advantages of acquiring digital images of entire samples or subsamples ultimately will be the ability to collect quantitative data such as diameter and length, allowing size-frequency assessments of foraminiferal populations while automating grain-size analyses without requiring separate processing.