Reducing False Alarms Using Genetic Programming in Object Detection

: This paper describes a refinement of an approach to locating objects in complex images with the objective of reducing the false alarm rate. The method uses genetic programming to evolve a detector using a threshold for defining an unclassified region. It is envisaged that incorporating a threshold during training will encourage programs to produce a high output when the object has been located. The objects required to be located are two difficult cephalometric landmarks that are either ambiguous in nature or located in a cluttered background. Results suggest that while increasing the threshold reduces the false alarm rate, it is to the detriment of the detection rate.

• Type: Conference
• Year: 2004
• Where: ICAI
• Authors: Andrew Innes, Victor Ciesielski, John Mamutil, Sabu John
• Keywords: Difficult Cephalometric Landmarks, False Alarm Rate, ICAI 2004, ICAI 2007, Unclassified Region