A staged genetic programming strategy for image analysis
Daniel L. Howard, Simon C. Roberts · 1999
The problem addressed is that of automating the task of visual inspection of images to detect objects of interest. Detectors are functions of the pixel information. When moved across an image these discriminate objects from non-objects according to the value that is computed. The detector is evolved off-line in two evolution stages. This strategy results in practical evolution times. It produces fast automatic detectors that can be interpreted to understand object detection principles.