Geometric Correction and Radiation Correction #

In the practical application of remote sensing images, it is necessary to transform them into a geographic coordinate system, that is, geometric correction. Generally, the method of geometric correction is to establish polynomial fitting formulas by collecting ground control points. They can be extracted from the vector database of GIS, and then determine the corresponding coordinates of each point in the image, and establish correction formulas. After the correction is completed, the vector points can be superimposed on the image to judge the effect of the correction. In order to accomplish the above functions, the system needs to be able to process raster and vector data comprehensively.

Some remote sensing images, such as Layover, Shadow, Foreshortening and so on, will be geometrically distorted due to the influence of terrain. DEM data is needed to correct and interpret them. In addition, due to the change of illumination caused by topographic fluctuation, it can also be shown in remote sensing images, such as the brightness difference between shady and sunny slopes. DEM can be used for radiation correction to improve the accuracy of image classification.

Image Classification #

For remote sensing image classification, the most obvious advantage of integration with GIS is the selection of training area. Through vector/raster integrated query, the image statistical characteristics of polygon area can be calculated, the classification effect can be evaluated, and the classification method can be improved.

In addition, in image classification, vector data can be rasterized and used as “remote sensing image” to participate in the classification, which can improve the classification accuracy. For example, considering the vertical zoning characteristics of vegetation, when classifying vegetation in mountainous areas, DEM can be used as a classification variable.

Selection of Region of Interest #

In some remote sensing image processing, only one region is needed to extract some features, which requires intersection between raster data and vector data.

Extraction of Lines and Other Geographic Elements #

In image processing, there are many Edge Detection filter operators, which can be used to extract the boundaries of regions (such as land and water boundaries) and linear objects (such as roads, faults, etc.). The results can be used to update existing GIS databases. This process is similar to the vectorization of scanned images.

Generation of DEM Data #

Using Stereo Images and radar images, higher precision DEM data can be generated.

Land use change and map updating #

The most direct way to update spatial database with remote sensing data is to take the corrected remote sensing image as background map and edit and modify the vector data according to it. The results of classification of remote sensing image data can be added to the GIS database. Because the result of image classification is raster data, the raster-to-vector operation is usually needed; if not, the raster data can be directly used for further analysis, then the system needs to provide the grid/vector intersection retrieval function.

Because remote sensing images can be regarded as a special type of raster data, it is not difficult to implement integrated tool software for remote sensing and GIS - the key is to provide very convenient raster/vector data interoperability and conversion functions. However, it should be noted that due to various factors, the information extracted from remote sensing data is not absolutely accurate. In typical land use classification, a classification accuracy of 90% is a considerable result, so it requires field investigation and verification - GPS can be used for positioning in this process.

In addition, the scale issue should also be considered, that is, the correspondence between the spatial resolution of remote sensing images and the scale of GIS data. For example, in practice, a common problem is: when performing geometric correction on TM data with a ground resolution of 30 meters, what scale of topographic map is needed to collect ground control point coordinates, and what scale of land use data can the classification results be used to update? Based on experience, the appropriate scale is from \(1:5万\) to \(1:10万\) . If it is too large, the accuracy of remote sensing data will be insufficient, and if it is too small, it will be a waste of remote sensing data.