Space Coordinate Systems Registration Using A Self-evolution Genetic Algorithm

As the integrated measurement systems technology develops, an integration of standardIC and three-dimensional integrated sensor systems have increased continuously. Theregistration and correction problem of space coordinate systems widely exists in three-axismeasuring systems and registration errors greatly influence measurement meeting therequirement of high-precision. The registration space coordinate systems problem, attempts tofind the transformation relationship between multiple space coordinate systems. Then usingthe transformation relationship, the measurement data from multiple space coordinate systemscan be transformed into reference coordinate systems, in which collected data from differentcoordinate systems can be computed. It is unpractical to solve this problem by means ofimproving fabrication process because of the various noises existing along with theassembling measurement systems. A feasible method is to correct the registration errors byalgorithm.In this paper, a novel genetic algorithm, self-evolution genetic algorithm is developed toaddress this registration and correction problem. The self-evolution genetic algorithm whichhas no human experience needed for setting crossover rare and mutation rate, is the algorithmfollowing the natural evolution process. With the proposed GA, the misalignment errors canbe identified. The identified misalignment errors further compensate for the measuringsystems. The theorem of self-evolution genetic algorithm is introduced in detail, which helpsto find the mapping relationship between two3D-coordinate systems.The technology that can solve correction and registration of the vector field measurementsystem, including the development of the model, the selection of coding mode and geneticoperators is introduced in this paper. Finally, a large number of simulation experiments, in thenoise and no noise conditions, were operated to verify self-evolution genetic algorithmperformance in all aspects.Simulation results show that the algorithm is good in accuracy, speed and robustnesswhen it is compared with simple genetic algorithm.