A Independent Approach to Multiclass Object Detection Using Genetic Programming


Authors: Mengjie Zhang, Victor Ciesielski, Peter Andreae
Source: GZipped PostScript (1048kb); Adobe PDF (1243kb)

This paper describes a domain independent approach to the use of genetic programming for object detection problems in which the locations of small objects of multiple classes in large images must be found. The evolved program is applied as a template, in moving window fashion, over the large images to locate the objects of interest. The paper develops three terminal sets based on domain independent pixel statistics computed from a square input field and considers two different function sets based on standard arithmetic and transcendental operations. The fitness function is designed based on the detection rate and the false alarm rate. A program classification map and a multiple binary map are developed for multiclass object classification problems to determine the class of an object in an input field. We have tested the method on three object detection problems of increasing difficulty with multiple different classes of interest. On images of easy and medium difficulty all objects are detected with no false alarms. On difficult images there are still significant numbers of errors, however the results are considerably better than those of a neural network based method for the same problems. This work not only extends genetic programming to multiclass object detection problems, but also shows how to use a single learned/evolved genetic program for both object classification and localisation. The results also show that mutation plays an important role in the genetic programming evolutionary process for multiple class object detection tasks. The object classification map developed in this approach can be used as a general classification strategy in genetic programming for multiple class classification problems.

[Up to Computer Science Technical Report Archive: Home Page]