A study on the application of Gravitational Search Algorithm in optimizing Stereo Matching Algorithm's parameters for star fruit inspection system

Abstract

This paper reports the result obtained by implementing Gravitational Search Algorithm for tuning Stereo Matching Algorithm's parameters for the application star fruit inspection system. The hardware for the inspection system is built by CvviP from Universiti Teknologi Malaysia using only single camera. The implemented Stereo Matching Algorithm used on the system comes from the built-in Matlab library. Each agent of Gravitational Search Algorithm in the search pace represents a set of candidate numerical value of the stereo matching's parameters. The sum of absolute error of the gray scale value of both images is used to indicate the fitness function. Benchmarking has done by comparing the result obtained with the previous literature that implements Particle Swarm Optimization. The result indicates that the application of Gravitational Search Algorithm as parameters tuner for stereo matching's parameters tuning is essentially on par with the Particle Swarm Optimization Algorithm.