Center Positioning Of Target In Monocular Position And Attitude Measurement

The subject comes from a shooting training system. When shooting, the gun has laser sighting. There is need to know the location of the laser point on the target. The location is used to correct the actual location when firing a gun. And because the target will move, the attitude information is needed. The entire model can be used monocular pose measurement system to simulate.Position and attitude measurement is an important topic in modern control, navigation and many other areas of research. Position measuring system of computer vision in industry and engineering has been widely applied and receives a lot of attention. Among these measuring techniques, monocular vision has advantages of non-contact measurement, simple structure. Compared to multi-sensor measurement, it can avoid difficulty of small field of view and the complexity of the calibration process.Based on the composition of position and attitude measurement system model of monocular vision, this paper discusses the principles of its composition. And that will be camera calibration, monocular visual space coordinates of the position and attitude measurement problems (PNP issues) and description of changes in target position and attitude. System uses a linear method for solving three-dimensional coordinate based on four coplanar feature points. Position and attitude method in system is described based on the measured coordinates results. In this paper, all simulation results are by means of OpenCV image database platform, so the programming is made easier. Coordinate error in the algorithm designed in this paper is limited to the highest accuracy within 0.03mm, which lays a good foundation for the following experiment.Target center positioning is positioning captured photos using image processing methods. The system introduces the two algorithms on spot positioning:the gravity of region and the least square fitting. When the algorithms are used on one picture, the time are both between 14-16ms with the conclusion that the least square fitting is faster than the gravity of region. Experiments show that the two objectives of target location error are no more than 0.2116mm.To distinguish characteristics spots belong to different system, this paper presents a center location algorithm based on different geometry:First segments regions in the original image to get the edge of the target. After that corners were detected with Harris corner detector. The paper compares this method to original Harris and improved Harris in accuracy and efficiency way. Experiment results show that this determination of center algorithm advances both accuracy and efficiency, and its error is within 0.1164mm.