Genetic programming for image analysis: orientation detection
Simon C. Roberts, Daniel L. Howard · 2000
The problem addressed is that of detecting the orientation of objects in infrared lines-can (IRLS) imagery obtained by low-flying aircraft. A novel method for orientation detection is described, based upon the minimisation of pixel-value deviation for linearly-spaced pixels. Genetic programming (GP) is used to discover the optimum manipulation of pixel information for this task. Results are compared against those produced using the conventional method of second-order central moments for principal axis detection.