Sorting of Archaeological Samples in D-band

Identifying the making of prehistoric pottery is one of the valuable means of tracing technical traditions and human mobility. We are interested here in the settlement in the Western Mediterranean through the sorting of pottery samples. The aim is to discriminate between two building techniques, looking for the features formed by the air bubbles embedded in the pottery samples. Current methods make use of heavy characterization systems such as the CT-Scan or the synchrotron. Microwave imaging offers an interesting alternative due to the possibility to have compact and portable acquisition systems. In this paper we investigate for the first time the sorting of potteries based on sub-mmWaves measurements. We process the images obtained from measurements with a non-linear SVM. Given the `mm' size of the desired aggregates, we work in D-band and take advantage of frequency and spatial diversities to have large training and test datasets. The first results show that the accuracy reaches up to 70% and depends on the choice of samples repartition among training and test datasets.