With Robust Weight Of Genetic Algorithm And Neural Network In Application Of GPS Elevation Fitting

Elevation is based on GPS technology to obtain geodetic reference ellipsoid height, but the project is needed is normal elevation, there is a certain difference between them, the difference is called the elevation anomaly. GPS geodetic elevation converted to normal elevation through elevation fitting. Abnormal elevation is obtained by known abnormal point elevation, then we obtain the unknown point normal elevation.BP neural network is widely used in GPS elevation fitting because of its self-adaptability characteristics, but BP neural network has its limitations that it easily falls into a very small value. It is applied that combining neural networks and genetic algorithms for limitations, it is based on neural networks, we search for the best weights and thresholds of the neural networks by the genetic algorithm, it can solve BP neural network easily falls into a very small value, network instability and other issues. In the process of GPS Height Fitting, the system error or a larger residual error may make mathematical models from distortion. Further robust estimation theory is reintroduced, because Selecting weight iteration robust factor can resist gross error, join it to genetic BP neural network can make the elevation fitting to resist to larger residual error, to make the results more accurate.BP neural networks, genetic algorithm, robust weight, the three combined constitute to the including robust weight of genetic neural network model. Apply MATLAB to programming, through experimental analysis of two engineering examples, by contrast the other three methods BP neural network method, GA-BP neural network method,containing robust weight of BP method described containing robust weight genetic neural network has the ability of anti-poor and more precise fitting.