Genetic programming as an integration platform for visual routines

Genetic programming (GP) is emerging as a promising technique for an impressive variety of computational tasks. An important prerequisite for exploring new applications of GP is a versatile and distributed simulation environment for evolving genetic programs with flexibly specifiable processing primitives and types. We consider the field of computer vision as a particularly interesting and challenging domain and analyze some requirements for the application of GP to the example-based synthesis of computer vision algorithms. We then present an outline of the GENCODER2 (GENetic COde DEvelopeR) distributed simulation environment, which was designed with particular attention to these demands. As an application example, we report results for the genetic construction of algorithms for the detection and classification of cell bodies in microscopic images. The generated segmentation operators allow accurate viability determination for suspension cultures, which plays a fundamental role in cell culture technology