Adaptation of the object recognition method based on random ferns for mobile robots

Due to the rapid development of the mobile robot technology, the vision-based object perception in changing scenes is of great practical significance. However, visual object recognition algorithms in constrained scenes, such as low image resolution, motion blur and background clutter is lacking. In this study, the performance of a representative object recognition algorithm called random ferns was evaluated on the mobile robot dataset. Meanwhile, some parameters of the random ferns are modified according to the characteristics of mobile robots, and then the effects of each parameter are analyzed to explore the relations to the final object recognition. The results show the adaptation of the random ferns object recognition for mobile robots was successful and provided a good reference for further application of the algorithm for mobile robots.