|其他摘要||Change detection is man-made structures a fundamental process of urban monitoring. Although the traditional change detection methods get a lot of useful change information from the two-dimensional remote sensing images at different times, it can’t obtain artificial structure changes in the vertical direction sufficiently. It is difficult to obtain sufficient artificial structures changes in the vertical direction because the different conditions and sensor acquisition parameters.
In addition, it is hard to distinguish spectral difference between buildings and other surface objects due to the similarity of their spectral features, such as building roads and bridges. The elevation information from digital surface models is particularly important to monitor changes, especially for change detection of artificial structures. At present, a lot of change detection studies are based on lidar data, but the frequency of lidar data is less and it is almost unusable when used to detect the changes in remote areas or at large-scale areas. Aerial imagery has the characteristics of convenient, fast, affordable and easy to be obtained, and can be served as a convenient tool especially for unexpected disaster relief.
The digital surface model from high resolution aerial images can acquired three-dimensional information of artificial construct. The true digital orthophoto provides the shape, texture, gray information of observed objects. We used improved algorithms and techniques to accomplish quick and accurate detection of comprehensive changes of urban man-made structures in two-dimensional and three-dimensional aspects. The main novel points of this paper are described in the following:
(1) The acquisition and optimization of digital surface model and true digital orthophoto model. First, we produce the initial digital surface model by using a probabilistic relaxation matching algorithm to obtain global consistency of feature points, lines, surfaces and matching model. Then we obtain refined digital surface model by the elevation value constraint method for point cloud data points precision ground and artificial constraint classification structures.
(2) The change detection of posting extraction of artificial structures. This method is different from the most conventional methods. We get the artificial structures profile with elevation values from digital surface model. At the same time, we get the texture information, location information, shape information as two-dimensional from true digital orthophoto model. Two-dimensional and three-dimensional information was combined in a topology design. At last, we obtain the artificial structures results of performing change detection in different periods.
(3) Direct the change detection. It is a process of using the change information of elevation information from digital surface model and position, shape, texture changes from true digital orthophoto model. This method not only guarantee in the change of three-dimensional information, and also to ensure the changes in the two-dimensional information.
The efficiency and robustness of the proposed technology and the method are evaluated via comparative experiments and real practical applications. The method makes up the problem that the artificial structure can't be detected by the change of shape, position and gray value when the gray scale, shape and position information is detected. Found several tests post classification change detection of too much manual intervention, slightly higher accuracy, but classified detection methods need to deal with a lot of data, a relatively time-consuming and labor-intensive, for the demand of real-time and quick, direct, automated method more practical.|