DEM Matching Algorithm Using Gauss-Newton Method

DEM Matching technology without control point is one of critical steps for multi-temporal DEM analyses, and it is also one hot question in many others fields, such as pattern recognition, computer vision, computer aided design, and robot. By comparing DEM data sets acquired in different epochs, the change in earth’s surface related to geological disaster can be analyzed quantitatively. Currently with the rapid development laser scanner and spatial data processing technology, the efficiency of data acquisition, the data precision and spatial resolution become higher and higher, therefore, are required to strengthen its ability to deal with more complex terrain data sets and to enhance its processing efficiency.The existing studies indicated that topographic relief has an effect on DEM matching performance, but the specific effect has not been analyzed and discussed. This paper studied the effects of topographic relief on DEM matching performance through the simulation test, and preliminary findings prove that there are some possible rules between topographic relief and DEM matching performance. In addition, the effects of DEM scale on matching efficiency are discussed.Matching efficiency is one of the important factors on whether the DEM matching algorithm can be widely in real application. The matching time is mainly consumed by two steps:establishing the point correspondence and computing the objective equation. Presently, the efficiency of matching algorithm is enhanced by improving the correspondence of points. This paper improved the efficiency of DEM matching through increasing the efficiency of computing the objective equation, in which the Gauss-Newton method is used to replace the least-square technique. The discrete form of Gauss-Newton Method suitable for the discrete DEM is derived to compute the objective equation of DEM matching, and then the DEM Matching Algorithm Using Gauss-Newton Method is proposed in this paper. The simulated experiments are adopted to test the performance of proposed algorithm. Three indicators are considered, which include matching precision, pull-in range and matching efficiency. According to experimental results based on the simulated datasets, the proposed algorithm gives the accurate transformation parameters. Simultaneously compared with existing DEM matching algorithm, the average convergence rate of proposed algorithm improves about 42% and the matching efficiency reduces about 75%. The matching precision and pull-in range have improved. These experimental results indicate that the proposed algorithm is effective and reliable.