Fast Gabor Filters for Object Recognition of Mobile Robot

Gabor filters have been used extensively in areas related to feature extraction of images due to their localization in space and bandlimited properties. Since 2-D Gabor filters have more complexity of computation, they have been used in static image decompositions rather than for mobile robots. In an attempt to reduce complexity of computation of 2-D Gabor filters for mobile robots, in this paper, a method of fast Gabor filters is presented. In the method, 2-D Gabor filters are decomposed into 1-D Gabor filters along non-orthogonal axes with different variances first, and those 1-D Gabor filters are recursively implemented, then the image group after fast Gabor filters is extracted feature by Principle Component Analysis(PCA), last the image would be classified by support vector machine(SVM). Experiment results indicate that, mobile robots can reach recognition rate of more than 92% and speed of quasi realtime image processing of 8 frames per second by the method.