Automated Classification of Stone Projectile Points in a Neural Network

We have built a working prototype for online classification of stone projectile points in a neural network. The initial application involves specimens drawn from the North American Pacific Northwest cultural area. The computing system environment for hardware is designed for 1386 architecture. Software is coded in VB.net and C# for the DLLs. The current database design is not software specific; however, it requires a robust relational database server. The auto-classification system consists of three stages. Stage 1 is the classification system, with software that allows users to submit images of artifacts or actual specimens that are digitized by lab staff. Stage 1 generates projectile point classifications with specimens assigned to recognized types and is a .NET standalone application. Stage 2 consists of release of a typological descriptive report to system users, including a full image inventory of submitted and classified specimens with attached statistical probabilities of type assignment. Stage 3 is a webbased application hosted on the Technology Innovation Center system that serves as the educational system for public access and study. This paper presents the practical difficulties and successes encountered in automating stone projectile point classification in a neural network, which offers potential for development of a creative, thinking classification system and a rich, accessible, secure reference database.