Analysing The Motion Gesture Of Ship Model Based On The Visual Image

Shipbuilding is a one-piece or small batch manufacturing industry and the cost is veryhigh. In order to reduce the risk, a large number of the ship model tests for valuation theperformance of full-scale ship were carried out during the design process and before the startof construction. Based on the results of ship model test, the full-scale ship performance wasextrapolated and evaluated with a certain similar relationship.After evaluation, the design wasoptimized for meeting the demand of ship. Therefore, the model test plays a important role inshipbuilding industry.Machine vision technology is a new non-contact measurement solution to estimate thethree-dimensional space motion of ship model which is developed in the recent years.Three-dimensional space motion which includes a displacement and rotating of threedimension. The camera takes pictures of the ship model in continuous to obtain the motionthrough image processing. The measurement precision has been improved. The motion ofship model can avoid the influence by the signal transmission cable of traditional measuringequipment.The thesis made the following research work:First, it discusses the process of target position and target tracking by computer visiontechnology. On the basis of theoretical knowledge, the paper put forward a feasible solutionfor tracing and motion measurement of ship model.Second, analyzing the imaging model of the camera, the relationship of spatial point inthe earth coordinate system, the camera coordinate system, the image coordinate system andthe image pixel coordinate. The camera calibrated with Zhang’s camera calibration method bycheckerboard calibration plate.Third, in the project, the Mean Shift tracking algorithm was modified in two aspects:using similarity coefficient and combining with Kalman prediction to improve the MeanShift algorithm.For solving the problem of the size of the target in the image will be changed with thedifferent distance between the target and the camera, when the model moved in a large area,the algorithm will use similarity coefficient, which can be described as a Gaussian function,to adjust the size of tracking core. Combining the Kalman prediction, the algorithm usestarget position history to predict the location of current time to solve the case of the target isshaded. Some simulation results about the improvements were given in this paper.In the end, the paper discussed the problem of motion analyzing in visual measurement,and proposed a method for determining the monocular vision pose known body coordinatesbased on the ship model. The3D pose of the ship model is regarded as the angle of rotationbetween the ship model coordinate system and the earth coordinate system. The nonlinearequations containing the ship model coordinate is deduced of the camera imaging model.Through the algorithm simulation, the algorithm has certain feasibility, which can be used for measuring the pose of ship model system based on vision.