Multiple Target Recognition and Location Using Deformable Templates in Downward Looking Image
基于变形模板的多目标识别与定位

The typical target recognition and location is a key technique of the application system for Unmanned Aerial Vehicle (UAV), but there exists a certain difficulty in entirely automatic target recognition and location. A new method to recognize and locate multiple targets is presented in this paper. Through introducing the deformable template technique and the man-in-the-loop thought, this paper designs the multiple target recognition and location algorithm for the fused downward-looking image from UAV. The emphases of the method are laid on the construction of different deformable templates, the definition of corresponding energy functions, and the optimal search strategy. Results show that the technique of multiple target recognition and location using deformable templates is more effective and can accurately recognize and locate multiple typical targets.