Combination filter for keypoint detection and pattern recognition in traffic visual surveillance

The detection of keypoints is very important for several computer applications. There are several ways for keypoint detection and pattern recognition. Earliest work proposes a trainable filter called COSFIRE filter which is a combination of bank of log gabor filters. This proposesd work interested in configuration of more sophisticated combination filter using bank of log gabor filters with bigger support. The specific function that use here to combine filter response is weighted geometric mean, essentially multiplication, which has specific advantages regarding shape recognition and robustness to contrast variations. The proposed work tends to demonstrate the effectiveness of filter in traffic visual surveillance including detection and recognition of traffic signs in complicated scenes.