A fast image processing based kin recognition method in a robot swarm

In this paper we develop a reliable and fast visual recognition algorithm to detect kin object in a robot swarm, where every robot is equipped with a zebra pattern. Our idea is based on Fast Fourier Transform (FFT), which is sensible for this kind of pattern and has a relatively low complexity. We sampled the camera images of the robot by a vertical sampling method, and the sampled one dimensional vectors were the inputs of the FFT. The basic idea of the pattern localization is that there is a distinguishable peak in the FFT spectrum if the sample went through the horizontal zebra pattern. If this peak is larger than an empirically determined threshold value the algorithm finds the location of a kin robot. To test our idea more experiments were accomplished with Surveyor SRV1 robots, which have on-board camera system. The captured pictures of the observer robot were sent to and processed in a PC using MATLAB program.