Real-time Electronic Image Stabilization for Spherical Robot Based on Inertial Measurement Unit and Gray Projection Method

As a new kind of autonomous mobile robot, spherical robot will inevitably sway left and right during walking, and the amplitude will be large sometimes, which brings serious interference to visual systems on the robot, thus real-time electronic image stabilization (EIS) is required. Spherical robot can be used to patrol in complex environments with multiple foreground and background objects, commonly used image-based EIS algorithms often fail under these circumstances. Therefore, this paper proposes a real-time EIS algorithm suitable for spherical robot, according to the characteristics of it. The algorithm divides the task into three subtasks: rotation correction, vertical translation correction and horizontal translation correction. Information from inertial measurement unit (IMU) is used to correct the rotation. Vertical translation correction is based on IMU information, and supplemented by gray projection method. Motion estimation of horizontal translation uses IMU information combining with gray projection, and then mean filter is applied to smooth the translation. Experiments show that the algorithm can effectively stabilize the greatly oscillating images taken by spherical robot, so as to provide a solid foundation for subsequent vision-related algorithms.