Abstract:

A method of target recognition from images based on complex natural scenes is presented. Firstly, multi-scale and multi-view dynamic features are represented, and the contour characteristic views of the target are prepared. Then, an input image is processed using morphological grayscale reconstruction. The processed image is segmented and the regions of interest are obtained. Lastly, the target is recognized by contour extraction for the regions of interest and matching based on Hausdorff distance metric. The experiment results show that the proposed method can effectively recognize the target from complex natural scenes.