Research On Video Stabilization Technology Of Humanoid Robot

With the development of science and technology and image processingtechnology updating, vision-based humanoid robots come into our field of vision.Because of the robot motion characteristics and lack of professional stabilizationequipment, images captured by the robot are usually accompanied with dithering,which is a big inconvenience to vision, and at the same time causes interferenceto post-processing of the image. So the video image stabilization technology ismore and more important. The aim of video stabilization techniques is to removethe jitter in the video part, to improve the image quality and the human visualperception. As image processing preprocessing stage, image stabilizationtechnology has great significance in the civilian and military.This paper studies the general flow of electronic image stabilization,through access to information, research of domestic and foreign about the statusof electronic image stabilization technology. I introduced the basic principles ofelectronic image stabilization and the main algorithm: block matching method,gray projection algorithm, the representative point algorithm, the feature quantitytracking algorithm etc. By understanding the motion characteristics of intelligentrobots, I designed a fast gray projection algorithm and tested by experiment. Itshows that for horizontal jitter and vertical jitter the fast gray projectionalgorithm is better and suitable for this kind of motion characteristics. Hav ingdepth study for the feature point matching algorithm，I Studied the SIFT operatorand Harris operator for feature point extraction in image stabilization applicationand improved feature point extraction method, the feature point selection process,based on relative error values instead of absolute error values. Finally, calculatedcompensation parameters by Kalman filter, motion compensation, and completedimage stabilization process. Improved method is lower than gray projectionalgorithm in time, but the rotational movement and the moving object can beprocessed accurately. Finally the electronic image stabilization technologyevaluation criteria proposed, subjective evaluation and objective evaluation,evaluation of the results achieved.Finally this paper introduced image stabilization system design of fully autonomous humanoid robot using dual-core architecture. Then describes thecharacteristics of the humanoid robot, using translational model and rotationscaling model for image stabilization system design Images captured fromcameras, processing images through image stabilization algorithm then the finaloutput the stable image.