Object Detection of NAO Robot Based on a Spectrum Model

NAO robots often need to detect objects to accomplish its task. At present, color segmentation is popular with NAO robot vision tasks because of its lower-end specification. A spectrum segmentation algorithm is proposed to realize real time detection in this paper. Spectral model is the foundation of human visual system, which can separate objects of distinctive color characteristics from complex illumination. Compared with current methods, color threshold in our method is trained by objective color and background color, which can automatically separate foreground and background. In addition, this paper employs Support Vector Machine (SVM) to recognize segmented regions to increase detection accuracy. Experimental results demonstrate effectiveness of the proposed method.