Improved object recognition algorithm based on template matching
Xiaoling Ding, Yibin Li, Xin Ma · 2015
To realize the goal of real-time object detection fmobile robotics, we present using an improved te mplate matching algorithm that does not require a time con suming training stage, and can handle texture-less objects and recognize the same object from different perspectiv e. At its core, we use the dominant gradient orient ations only and use a limited set of templates to represent an obje ct. To recognize an object from different perspecti ve, we use the affine transformation. The experiments demonstrate that it can detect texture-less objects in complex situations and recognize the same object from different perspectiv e in real-time.