Monocular Camera Calibration And Pose Estimation Based On All Parameter Adaptive Particle Swarm Algorithm

At present,manipulators with vision are more and more widely used in various industrial productions,often with position-based visual servo structures.In order to realize the visual guidance operation of a manipulator,the hand-eye calibration is often required and then the accurate pose of the target object is needed.Therefore,in the whole process,it is important for visual servo in practical industrial applications to improve the accuracy of camera calibration and pose estimation of target objects.To solve these two problems,some improved methods of camera calibration and pose estimation of target objects with all parameter adaptive particle swarm algorithm were proposed in this thesis as follows.Firstly,the research on robot visual servo was done based on position-based visual servo,image-based visual servo,and hybrid visual servo,respectively.It is very important to get the accurate pose of the target object in position-based visual servo structures.Therefore,in order to solve the problem of pose estimation under monocular vision,the four main methods such as pose estimation based on artificial signs,pose estimation based on natural features,pose estimation based on sequence images and pose estimation based on model library were surveyed.Secondly,in order to solve the problem of local optimal solution in camera calibration method based on the traditional particle swarm algorithm,a method of monocular camera calibration based on particle swarm algorithm with all parameter adaptive mutation mechanism was proposed.The initial values of the internal parameters of the monocular camera were obtained by the Zhang zhengyou method,and then were optimized by particle swarm algorithm with all parameter adaptive mutation mechanism.During the iteration,the inertia factor was adaptively adjusted according to the change of the optimal fitness value.The learning factor was adaptively adjusted according to the different effects of the local and global optimal particles on the current particle.The mutation rate was adjusted adaptively according to the normalized average grain distance,respectively.At last,the accurate internal parameters of the monocular camera could be obtained.Thirdly,in order to solve the problem of the pose estimation of three-dimensional objects under monocular vision,a method for pose estimation of the target object based on particle swarm algorithm with variable-scale mutation mechanism was proposed.Through the corresponding relationship between feature points of the three-dimensional object in the camera coordinate system and the object coordinate system,the initial value of the pose would be given by the direct linear transformation and then be optimized by particle swarm algorithm with variable-scale mutation mechanism.During the iteration,the inertia factor was adaptively adjusted according to the change of the optimal fitness value.The learning factor was adaptively adjusted according to the different effects of the local and global optimal particles on the current particle.The mutation rate was adjusted adaptively according to the normalized average grain distance and a method of variable-scale mutation mechanism was given,respectively.Fourthly,in order to solve the problem of the pose estimation of three-dimensional objects under monocular vision,a composite method for pose estimation based on static extend kalman filter and particle swarm algorithm with dynamic mutation mechanism was proposed.The initial value of the pose would be given by the static extend kalman filter and then be optimized by improved particle swarm algorithm with dynamic mutation mechanism.At last,the experiment demonstrated the accurate pose was obtained by the propoesd method.