GISdevelopment.net ---> AARS ---> ACRS 1989 ---> Poster Session 1

On the Appliation of the auxilary data in the composition of the multiple data:
An example for Wuhan City

Guan Zhe-Qun, Shu Ning
Wuhan Technical University of Surveying and Mapping
Wuhan China


Abstract
In this paper the application methods of the auxiliary data for the composition of the multiple remote sensed data are discussed
  1. The auxiliary data and geographical network system are encoded by sampling and quantizing with adaptive mathematical model. After digitizing the auxiliary data, a independent band is obtained, and then composite with the remote sensed data which is geometrically rectified.

  2. The rational filtering operators in the restructured remote sensed data by introducing the 'information raster file which is derived from auxiliary data facilitate the image analysis and features selection.

  3. When the geometrical and spectral resolution of remote sensed data different from each other, the problem under study is find a auxiliary data which could provided the satisfactory hierarchical relation between them.
Finally, on the basis of the landform auxiliary data some experimental have been carried in wuhan area.

Introduction
The auxiliary data is useful to the recognition of an object. There are several important reasons. The first, many images do not contain concrete objects and sometimes image features tend to be ambiguous. Edge detection techniques, for example, mark the areas where intensity changes take place. Such line segments can indicate the objects surface boundaries, and may also indicate shadow edges. As a result, interpretation of an image often requires a combination of auxiliary data. The second, many images do not directly show us the information on some natural environment factors ( such as oil) , so when we analyse image features according to the formation cause of environment, we must use auxiliary data. The third, the resolution of many images are not often high enough to recognize objects which contain contingency and complexity, so we must apply auxiliary data in enrich the information of image.

A common approach for using the auxiliary data is that they and geographical network system are encoded by sampling and quantizing with adaptive mathematical model. After digitizing the auxiliary data, a independent band is obtained, and then composited with the remote sensed data which is geometrically rectified.

In this paper, on the basis of above, we mainly discuss the following application methods of the auxiliary data.
  1. The rational filtering operators in the restructured remote sensed data by introducing the information raster file' which is derived from auxiliary data facilitate the image analysis and features selection..

  2. When the geometrical and spectral resolution of remote sensed data are different from each other, the problem under study is to find a auxiliary data which could provide the satisfactory relation between the various kinds of remote sensed data.
Finally, on the basis of the landform auxiliary data, some studies have been carried out. And the analysis of the composition of SPOT and TM images in Wuhan area has proved the flexibility of the proposed methods.

Information Raster File Formation
Some environmental data, which compactly relate to the theme extracted by image analysis techniques, cannot be directly obtained from images. Without these data, the results of the image analysis aren't satisfied. TO use these data in the image analysis, one often forms information raster file' which is a theme file, after the classification of the environmental data. We know that the objects corresponding to a certain image feature aren't usually unique, such a fragment feature in a touristy area may be a forest land viewed and admired, but may also be a forestland serving as a foil. The function of the information raster file is to sift out some object from many objects corresponding to a images feature. In the remote sensing of the touristy resources of the Dong take area information raster file, which synthesizes the specific of the stop, the direction of the slope the altitude and the morphology, ahs been formed by the landform feature time and the weighted distance transform structured as follows:
  1. Weighted Distance Transform
    Let each of the eight neighbors of any pixel P be indicted by the corresponding cardinal point in the compass, and let it and 12 be respectively the two distance weights selected for the horizontal vertical and for the diagonal neighbor of P. Unless differently stated, the letters used to denote pixel, will also indicate the associated distance label.

    The weighted distance transform )WDT) of area F can be computed within two sequential raster scans ( Performed in forward and in backward fashion, respectively),during which the following operation are applied to every P in F,

    Forward scan, f1(P) = min(W + t1, NW + 12, N + t1, NE + 12)
    Backward scan, f2 (P) = min(P, E+t1, SE + t1, SW + t2)
    here, W, NW, N, NE, E, SE, S, SW represent west , northwest, north, northeast, east, southeast, south, southwest respectively.

    It has been point out that with t1 fixed to 1, the difference between the Euclidean distance and the weighted distance is minimized when t2=1, 351, so we take the choice of t1=3 and t2=4 as weights.

    The distance d*, according to which the WDT is computed , can be defined by suitably adding the distance do which represents the shortest distance in the eight neighbours and distance to which is the shortest distance in the four neighbours, refer to FIG. 1.

    Fig.1. A minimal path joining p with q
    Relation among d*, d8 and d4


    Hence, d* is a distance in the mathematical sense. To measure the distance d* between two pixels p and q, we note that along the shortlest path connecting p and q any horizontal vertical move ( diagonal move) has a weight t1=3 ( t1 =4). Since horizontal/vertical can be expressed with2d8 ( p.q.) -d1 (p,q) and diagonal moves are d4(p,q)-d8 (p,q) . So d= (p,q)= 2d8(p,q) td4(p,q).

  2. Information Raster File Formation
    Select specific contours as the type lines which depict landform, At the same time, depict mountain-backbone lines (see FIG.2.) We computer the distance along the shortest path connecting p and q, according to the WDT and in the direction from backbone to countrour (See FIG.3.).

    In FIG. 3, 0 depicts backbone lines and 1 is the type line of the landform For each pixel we may define three functions in FIG.3.

    S=S(p) S is slope
    D=D(p) D is the direction of slope.
    0=0(p) 0 is the position implying distance between backbone and p.

    In terms of the three essential considerations, we obtain the local landform attribute vector of each pixel.

    Q = (S, D, O)

    We may form information raster file by using the attribute vector. On the basis of the composition of the image and the raster file, the touristic resources of the Dong make area can be better recognized and analysed as follows.

    The area in FIG. 4 corresponds to the area in FIG. 3.. The tables in FIG. 4. are the results of image interpretation. In terms of FIG. 4., the area is a forestland, hence it is suited to become a touristic spot in the local point of view, but it is difficult to determine more exactly the action of the forestland on the touristic area.

    Under these circumstances, we can filter FIG. 4. by FIG.3. FIG. 5. is the results using such kind of filtering in terms of which we can exactly divide this area into three subareas for their different functions.

Fig.2. Contour and backbone lines


Fig.3. Landform weighted distance transform


Fig.4. Touristic spot analysis by image interpretation


Fig.5. Touristic spot analysis by information raster file


The Hierarchical Anaslysis of the Touristic Resources by Using Multiresolution Image
It is not enough to investigate touristic resources in a local area, since a landscape being a touristic spot, depends on not only the local attributes of this spot, but also the global attributes of the area. In this section we use the low resolution images in investigating the global attributes of the Dong take, and use the high resolution images in investigating the local attributes of its subarea. The global attributes are connected with the local attributes of its subarea. The global attributes are connected with the local attributes by the hierarchical system of the landform, as is described in the following.
  1. Hierarchical scheme
    The scheme for the extraction of attributed relation representation form images, is actually a parallel hierarchical shceme that consists of several layers, Each layer, associated with a image has a field of cells which are connected in accordance with a certain neighborhood configuration, as defined by the relationship between landform types.

    A image hierarchy can be represented by a tree. In a tree of that kind, subimages at lower levels which originate from target scale images are jointed to allude to subimages at higher levels which originate from smaller scale images. The ith node at the 1th level of the tree corresponds to the subimages S(I,t,) . The links between nodes indicate set inclusion. Hence, a link between a subimage S(k, t+1) (ancestor or parent) and its disjoint subparts S(I, t) (descendants or sons) indicates that S(I,t) < S ( k, t+1). A multiresolution image hierarchy, P, therefore corresponds to a node set (S1, SS2,…., Sn). Which associated with the relationship between the species and the category of landform types. Consequently, image local representation in some area can be extracted from higher resolution images; image global representation can be extracted from tower resolution images; and the relationship between the local and global representation can be established according to the incorporation relation of landform types.

  2. Hierarchical Analysis for the Touristic Resources of the Dong Lake Area
    The images are real pictures of land surfaces at the moment of scanning or photographing. A great deal of information of area, such as the temperature of terrestrial overburden landuse and environment associated with touristic resources, could be derived from images. But because the difference in resolution and characteristics of different images, the effect isn't satisfying to directly composite them on the basis of the unified scale. We regard the composition of the multi-images which listed in Table. 1. as a hierarchical mapping which explained in section 3.1.

    Table 1. The sensing system used in Wuhan dong take
    Landsat TM 120 (m) 6 (spetral ) Sensitive to temperature
    SPOT multispectral 20 1 reduced level of pigment absorption
    SPOT multispectral 20 2 strong chlorophyll absorption
    SPOT multispectral 20 3 high vegetation reflectance
    Airphoto 1, 10000   false color infrared

    FIG. 7. Depicts the landform hierarchical system with respect to Mo hill located on the southern lakeshore of the Dong lake. In terms of the top level in FIG. 7. Mo hill belong to takeshore landscope, so it has development significance in touristry. But the top le el in FIG. 8 indicates the serious destruction of the primitive landscope. Overall, the Dong take area hasn'st been a good touristic area. In terms of the medium level in FIG. 7. Mo hill belongs to hill-terrace landscope where/ The touristic spots can easily be laid out. But the medium level in FIG. 8. . shows us that this region has been divided into several parts by different departments, so the touristic spots and has been divided into several parts by different departments, so the touristic spots and the action radius are quantified. In terms of the bottom level in FIG. 7. we know for certain that the stope facing the Dong take is sharp in the direction of northeast, and the stope is less in the direction of southwest. So the former is suited to lay out the touristic spots where trurists appreciate scenery; the latter is suited to lay out the spots where tourists take a west or find entertainment. On the basis of the classification depicte din the bottom level in FIG. 8., We find that the touristic spots watch the Landform in Mo hill region.

Fig.6.


Fig.7. The landsorm hierarchicat system in Wuhan Dong take area.


Fig.8. The image hierarchy of the Dong take


Conclusion
  1. We propose a information raster file, which systhesizes the specific of slope, direction of slope, attitude and morphology by introduction the weighted distance slope, direction of slope, attitude and morphology by introducing the weighted distance transform. The experiment in the investigation of the touristic resources of the Dong lake area, proves that it is useful for sifting out the object.

  2. This paper indicate the merits of the hierarchical analysis. The first, it can give full play to the diffe4rent resolution images. The second, it provides a methods of search of the objects from local attributes to global attributes, as is useful to computer image analysis.