Heterogeneous remote sensing image content-oriented digital elevation data reconstruction method

Abstract

The invention provides a heterogeneous remote sensing image content-oriented digital elevation data reconstruction method, which comprises the following steps of establishing a remote sensing image content heterogeneous description operator, describing differences among remote sensing image contents based on the operator, and establishing a regression prediction model by utilizing the difference description and the characteristics of a remote sensing image, and large-area region digital elevation data reconstruction is realized. By means of the method, the regression prediction model can welladapt to heterogeneity between image contents, the model can obtain a good prediction result under the condition that the image contents have large difference, and large-area area digital elevation data reconstruction oriented to heterogeneous remote sensing image contents is achieved.

Description

Technical field: [0001] The present invention provides a digital elevation data reconstruction method for heterogeneous remote sensing image content, which is used for reconstructing and acquiring data in the field of geographic information systems, and specifically relates to the field of landform and surface remote sensing image technology. Background technique: [0002] Obtaining digital elevation data of large areas on the surface of the earth is helpful for large-scale geographic three-dimensional modeling, and is of great significance for the construction of basic geographic information data, virtual reality simulation, disaster simulation and prediction. Therefore, it is very necessary to obtain fine digital elevation data information of large areas of the surface. [0003] In order to obtain fine surface digital elevation data, the current methods are divided into two categories: the first type is to obtain fine surface digital elevation data by means of manual measu...

AI Technical Summary

Problems solved by technology

[0003] In order to obtain fine surface digital elevation data, the current methods are divided into two categories: the first type is to obtain fine surface digital elevation data by means of manual measurement and aircraft laser radar measurement. The advantage of this method is that the data accuracy is high, but It requires a lot of financial and time support, and is more suitable for the acquisition of data in relatively developed urban areas. For large areas, most units cannot afford the corresponding financial and time requirements

The second type is to first download the low-resolution digital elevation information of the corresponding area as the basic three-dimensional data, find a small area for manual measurement to obtain fine elevation data, introduce high-resolution remote sensing images of the area, and use artificial intelligence algorithms to establish Based on the regression prediction model between high-resolution remote sensing images, low-resolution digital elevation, and high-resolution digital elevation information, the entire region is predicted; this method will be better applied in small areas and ranges However, when faced with a large area, due to the large area, its internal data corresponds to multiple remote sensing images, which will make the traditional method face two problems: Problem 1, the remote sensing images of the corresponding area may come from different There will be differences in the satellite shooting results, data resolution, and pixel value response range, and these differences will eventually transmit errors to the regression prediction model, which will lead to the failure of 3D reconstruction; problem 2, in the case of large areas, There will also be differences in shooting time and angle between different images, and there may even be seasonal differences in different locations (for example, some areas have large differences in altitude, the top of the mountain is covered with ice and snow below 0 degrees, and the foot of the mountain is covered with vegetation above 10 degrees) , these large discrepancies can cause the prediction process using traditional methods to succeed in some regions, while data reconstruction fails in other relatively heterogeneous regions

Images