Aircraft recognition based on convex-concave analysis

The problem of aircraft recognition in a single image is analyzed. A novel method combined convex-concave feature and hierarchical analysis is introduced to do pattern recognition for aircraft using the feature of shapes and regions. Silhouettes obtained by image segmentation are described by chain code. Then convex-concave feature is computed from chaincode. Aircraft features are represented hierarchically with a sequence of convex-concave curves. Fast matching is performed with fundamental convex-concave feature, after which useful information (such as pose of the aircraft) can be obtained. The refined convex-concave feature is used to get detailed classification with additional information. Experimental results show the effectiveness of the algorithm for aircraft pose estimation in a gray level image.