Study On Stereo Vision Planning For Measurement Of Large Complex Curved Surface Parts

With the increasing demand for product accuracy and production efficiency of large-scale equipment manufacturing industry, the three-dimensional shape inspection technique as an efficient approach to quality control is playing an increasingly important role in manufacturing process. Binocular stereo vision measurement with advantages of high measuring precision and better flexibility in three-dimensional shape inspection system for large parts, has been widely recognized. However, for the measurement of large parts with complex curved surfaces, traditional handheld devices of binocular stereo vision measurement tend to have problems in local missed detection, low accuracy and low work efficiency. Therefore, in this thesis, a new kind of automatic vision measuring device based on binocular stereo vision is designed, the stereo vision planning problem is extensively studied in vision measurement and its corresponding resolution is given.First, the structure scheme and hardware configuration of vision measuring device are determined, and the upper computer and lower computer system of the measuring device is developed. The upper computer system mainly contains communication parameter setting block, image display and acquisition block, input and display block of camera path and pose; the lower computer system mainly completes data acquisition joint control.When the measurement coverage rate and feature resolution rate of planning objectives are given in vision measurement network, the decision variables and the constraint conditions of the planning are identified, the math model of multiobjective planning in vision measurement network planning is designed, which lays foundation for the solution of Pareto optimal set. Furthermore, the application of vision measurement net planning based on multi-objective genetic algorithm is studied. To be specific, the coding mode, the population processing methods, the constraint condition processing method, the individual fitness functions and control parameters of the algorithm are designed. Taking a large propeller blade as example, a group of Pareto global solutions satisfies the two planning objectives and the constraints is obtained through measurement network planning, and the effectiveness of planning result could be verified by this experimentA shortest route planning method meets regional constraints is proposed to solve the route planning problem, which uses two layers search, regional route finding and interregional route finding, to realize the shortest route planning. Moreover, compared with the method of exhaustion, the validity of the proposed algorithm has been proved.Finally,a vision measuring device is designed, and stereo vision planning method is provided for 3D measurement of large complex curved surface parts,which is of great significance for improving the accuracy of measurement of large parts.