Research On Trajectory Prediction Of Flying Spherical Objects Based On Stereo Vision

Grasping moving object by Mechanical Arm is widely used in industrial production, aerospace and military application fields. It is also the challenging research subject in the development of machine vision and the mechanical arm control technology.Binocular stereo vision system is the equal of the robot’s eyes, it affords Mechanical Arm the direction of the target object, so, binocular stereo vision system is the important part of the process grabbing motion target by Mechanical Arm. Come from that, the research is carried to finish the calibration of the binocular stereo vision system, realize the conversion of coordinate from Binocular stereo vision system to Mechanical Arm, meanwhile realize the tracking and forecast of the trajectory of the flying target by binocular stereo vision system. The main contents and results of this dissertation are presented as follows:Put forward a kind of localization method direat at globular flying object based on the camera calibtation, without matching and correction, by the means of analyzing the camera imaging model. This method make full use of the relationship between the coordinate system of imaging model, based on the genreal visual system hardware conditions. This method get the three-dimensional coordinates of target objects in the world coordinate syetem directly from the target in two camera imaging plane by the constraints. This method realize the matching of the center of the globular object by gray centroid method, improbing the system efficiency and accuracy. Compared with the simple straight forwad binocular stereo vision positioning method, this method does not requre system hardware conditions, and it’s application is more wide.Solved the hand-eye calibration between the mechanical arm and the camera by mathematical theory methods. This method aim at the eye-to-hand separation, transfering the rotating translational relationship between mechanical arm coordinate system and the camera coordinate system to the relationship between a new coordinate system to the two coordinate system. Realizeing the convertion the three-dimensional coordinate from binocular stereo vision system to mechanical arm by the rotating translational relationship from mechanical arm to binocular stereo vision system. By this means, we can provide position for mechanical arm, experiment proved that hand-eye calibration can guide the mechanical arm close to the traget object effectively.Structuring a binocular stereo vision system on the basis of The existing laboratory equipment, solving the synchronize of two cameras obtaining the target scene. We use QueryPerformanceFrequency(&) and QueryPerformanceCounter(&) to control two cameras to work at regular intervals, realizing the synchronize of two cameras obtaining the target scene and getimage. Improving the accuracy of the track prediction.In order to get the track of traget object, we attempt to use neural network and motion model to predict. But the experiment proved that those two method could not predict the three-dimensional coordinate effectively. So we put forward a mean combining neural network and notion model to provide compensation to motion model. By this way, we improve the precision of the prediction greatly.