Machine Vision: Exploring Context With Genetic Programming
Daniel L. Howard, Simon C. Roberts, Conor Ryan · 2002
This paper proposes an advanced method of object detection using a data crawler. Starting from a preliminary 'object identification', the data crawler scrutinizes the object's surroundings, and places flags where its interest has been aroused. Next, the binary map defined by these flags is analyzed statistically to quantify the context in which the object appears. The crawler's overall output indicates whether the object is a required target or a false alarm. The crawler's navigator program, its lag-placement program and its target detection program, are all controlled by a tree-based Genetic Programming (GP) method with fixed architecture Automatically Defined Functions (ADFs).