| The hexapod robot is chosen as the experimental platform in this paper. Both the hexapod robot controller and the infrared module for detecting moving target have been completed. By means of modeling for the mechanical structure of the hexapod robot, the kinematics characteristics of the robot are detailedly analyzed and a computing method for rotation angle of leg joint is obtained. Finally, the walking in straight line and turning gait around fixed-point of the hexapod robor can be realized using the designed gait arithmetic.Focus on the static background and the moving background, the algorithms for moving target detection and tracking are respectively designed. The algorithms are ultimately programmed and realized using the vision library OpenCV2.1 in the VC2008 environment.The algorithm based on adaptive background updating is applied to real-time detecting moving targets and the combination of Mean-shift algorithm and fast Kalman filter prediction is used for realizing target tracking in the static background.In the moving background, the question of the background motion compensation is analyzed. The motion estimation algorithm of SIFT feature point matching is designed based on scale invariant feature transform. Firstly, the stable SIFT feature points are extracted and 128-dimensional eigenvector is produced for describing feature points. Secondly, establishe the KD-tree toward SIFT feature points set, take the Euclidean distance as judgment criterion of the similarity and adopt the optimal binary search algorithm to look up all the matched feature points between the two adjacent frames. Then, get rid of outliers that belong to the targets using RANSAC algorithm, solve the accurate affine model parameters and realize the background motion compensation. Finally, the moving targets in the moving background are detected using the method of inter-frame difference and the motion history image. |
PDF Full Text Request