An unsupervised machine learning algorithm for visual target identification in the context of a robotics competition

摘要

Computer Vision and Machine Learning are the key to develop autonomous robots. While engaged with a IEEE Open Challenge, in which the robots need to recognize a miniature of a cow, we saw a solution in these areas. The main contribution of this paper is the algorithm implemented to identify and follow a known object, the miniature of a cow. We are constructing an application based on Image Processing that can detect in images this previously known object. The method yields the limits and the mass center of the entity and appropriates known algorithms, as well as Shi-Tomasi Corner Detector, the clustering K-means and local binary patterns.