Rotated object recognition based on corner feature points for mobile augmented reality applications

Object recognition technology has been a major issue in mobile environments. However, it is difficult to recognize any object on mobile devices, because mobile devices do not have enough performance on CPU and memory components. Thus, a new fast handling algorithm optimized for mobile devices is required to recognize objects. In this research, we propose new methods to recognize any object, especially rotated objects. Our method is designed to recognize any rotated object through corner point data. Corner data can be replaced by grouping those points having similar features as a representative one. And, corner data that are nearest from edge points are chosen to minimize any change of pixel information when rotating any given object. Also any specific pattern of the selected corner data and pixel information around the selected corner data need to be collected and stored for later matching operation. Experiment result shows that the proposed method can provide 96% accuracy. And, our algorithm shows highest performance. Therefore, our methods can be adapted to recognize any rotated object for performance and accuracy.