Research On3D Measurement System For Rocket Characteristic Stages

As a new development of the rocket models, many new designs and materialsare applied to the xx-x launch vehicle. xx-x rocket fairing, for example, has adoptedcomposite materials for the first time, so it needs the ground separate test to verifythe separation reliability. Rocket package-connection-unlocking device (packet forshort) is used to realize the connection and separation between the satellite and therocket. The device is an important stand-alone product, which requires a highunlocking reliability, and the packet must be separated within a limited envelopeafter unlocking. In order to learn and evaluate the mechanism of the new rocket atthe stage of design, recording and analyzing the real motion state of these importantparts is necessary. A rocket three-dimensional feature measurement system designedin this thesis is used to record the kinematic process of the important parts, toanalyze three-dimensional position information and complete the three-dimensionalreconstruction of the measured target. In this thesis, with the main topic ofimproving the measurement accuracy of three-dimensional measurement system, thispassage focuses on the vision measurement methods and related issues based onmulti-constraint fusion algorithm, and proposed a method of three-dimensionalreconstruction using the measurement data.The camera calibration has been the key point of the vision measurementsystem. The distance constraint method is used to realize multi-camera calibration online in this thesis. According to the characteristics of the central projection, arelational model between the pixel coordinate and spatial distance is built, and theinternal and external parameters and image distortion coefficient of the camera arecalibrated in high-precision. In the calibration process, the camera parametercalibration and three-dimensional coordinate measurement are completed step bystep. Experiment results show that the calibration method is of good stability and ahigh calibrationaccuracy.Vision measurement algorithm based on multi-constraint fusion is proposed,which is the integration of distance constraint in the field of view, the collinearconstraint and coplanarity constraint. This algorithm is introduced to the collinearequations which describes the central projection imaging model, and the objectivefunction with constraints is built. The paper gives the normal equations of targetpixels, the pixels with constraints and the pixels to test, and obtain three-dimensionalcoordinates of the test points, the orientation parameters of each camera and imagedistortion coefficient with the Newton iterative method. Multi-constraint fusionalgorithm reduces the impact of image distortion, and improves the accuracy ofvisual measurement system. Compared with the traditional bundle adjustmentalgorithm and the algorithm based on a single constraint, the algorithm proposedimproves the vision measurement system accuracy greatly.The accuracy of the pixel coordinate center location affects the finalmeasurement result directly. Therefore, to select the appropriate pixel centercalculation method and ensure a high processing speed are also a focus of this thesis.In this paper, combining the border search and gray-scale weighted-centroid, a highprecision locating of the target point center is achieved. First, search the boundary ofthe target point using border search algorithm to determine the imaging area, andthen determine the position of the pixel by using gray-scale weigthed-centroidmethod.The pros and cons of the point matching results directly affect the stability andaccuracy of the calculation procedure of the final space coordinates, so, the matching algorithm chosen should match with high accuracy and matching efficiency. Pixelpoint matching is a complex issue, and the matching method should vary accordingto specific scenes. This article introduces how to obtain the corresponding match ofthe same name point between the image pairs by encoding meta information, andcalculate the transformation matrix F, to make use of the measurement of similarityand compatibility, and get the initial match results of the same-name but non-codingmeta between the image pares, and finally through the fuzzy criterion, similarguidelines and distance constraint error criterion to remove the mismatch.In the process of the3D reconstruction, by means of using RBF network basedon micro genetic algorithm, the three-dimensional reconstruction of the3D pointcloud data sets is accomplished, and the three-dimensional reconstruction of theparallax image proves the superior processing power of this algorithm on largedatasets. The learning of the micro genetic algorithm based on radial basis networkavoids the premature phenomenon during the network evolution, while theconvergence speed and accuracy have also been improved.Based on the theoretical studies above, a multi-camera vision measurementsystem for the rocket characteristic three-dimensional measurement is designed. Inthis paper, the design of the vision measurement system includes: the control targetand the constraint control, the high-speed camera groups, the shoot controlequipment, computer, lighting, massive data memorizer, landmarks, as well asthree-dimensional measurement software. Use shoot control equipment via externaltrigger mode to make several cameras shoot in sync, and finally multi-constrainedintegration algorithm is applied to complete the measurement of thethree-dimensional position and surface of the target.