Recognition of bridges over water in infrared image based on fractal theory and background knowledge mining

To recognize a large bridge over water under complicated background in long rang aerial photography infrared images, a novel three-step strategy based on fractal theory and background knowledge mining is presented. First, all regions in accord with the contextual description of the infrared images about bridges over water and regions of interest (ROI) around the pixel 8-neighbors are detected based on fractal dimension and histogram active threshold, at the same time latent bridge targets are segmented. Then, image is performed over all sub images with morphological filters. Finally, the bridge target is recognized based on the characters matching. Experiments on the image tracker based on TMS320C6416 are conducted, and the results demonstrate that the approach performs well in automatic recognition accuracy and computational efficiency.