Landmark direction angle estimation based on omnidirectional image

Abstract

In order to estimate landmark direction angle to localize robot in an artificial environment based on omnidirectional image, this paper introduces the method about low pass filter noise elimination, artificial landmark features detection and landmark direction angle estimation algorithms. Landmark feature detection proposed that red landmark pixels beyond the threshold were extracted as a small area and the center of gravity was calculated for the extracted small area representing the candidate of one landmark. The blue patch provides as compensation to further distinguish the landmark from other objects in a complex environment. Two type experiments were done. The landmark detection experiment proved that the low-pass filter noise elimination was effective. Another experiment was conducted on the level cement ground in a 5×5-m area in the sunlight. 16 test positions were chose to test. Results showed that the MAE and RMSE of landmark direction angles were 1.12 degree and 1.30 degree, respectively. The proposed direction angle estimation method is high accuracy, low time-consuming computation and good robustness.