A fast and accurate information extraction method for high-resolution images

Abstract

The invention discloses a fast and accurate information extraction method for high-resolution images, including high-resolution remote sensing image input and preprocessing; according to the distribution law of the image's surface coverage, the image is partitioned and processed; various types of ground objects in the image are processed according to the partition Different types of target features are selected for one-time feature extraction to obtain multiple extraction results of various types of target features; Sensitivity analysis is performed on each extraction result to obtain various types of target features The sensitive feature group of the target ground object is used, combined with the layering strategy to determine the layered extraction sequence of the ground object, and the surface object information in the partitioned image is extracted layer by layer in the order of rough extraction and then fine extraction. Its remarkable effects are: reducing the cost of data, strengthening the research and application of domestic image data; solving the problem of difficult feature selection; improving the applicability and automation of feature information extraction.

Description

technical field [0001] The invention relates to the technical field of remote sensing image processing, in particular to a method for rapidly and accurately extracting information from high-resolution images. Background technique [0002] High-resolution remote sensing images can provide a large number of surface features, and the rich detailed information, spatial information, geometric shape of ground objects, and the relationship with adjacent ground objects can be better reflected within the same surface object category. However, the spectral statistical characteristics of high-resolution remote sensing images are not as stable as those of medium and low-resolution images, and the spatial distribution of ground objects is complex. The spectra of different ground objects are confused with each other, and the traditional spectral classification methods cannot obtain satisfactory information extraction results. [0003] With the continuous development of remote sensing tec...

AI Technical Summary

Problems solved by technology

However, by analyzing and summarizing the current research status and practical applications at home and abroad, it is found that there are still some problems in the extraction of land cover information: ① In actual production work, domestic Gaofen-2 satellite images are rarely used, and most of them are imported from abroad. high-resolution images, the cost is high, and it causes a great waste of domestic remote sensing image data; ②The application research of high-resolution image surface object information extraction methods is insufficient, due to the limitations of its experimental area, application field and the complexity of implementation However, many of the proposed research methods cannot effectively improve their production efficiency in practical applications; ③ Difficulty in feature selection

Due to the continuous emergence of high-resolution image features, feature selection is more difficult. At present, there is no optimal feature analysis method that can effectively select feature features; ④ The degree of automation of information extraction is low

In actual production work and application, the use of mathematical statistics theory combined with manual interpretation and field investigation is still the dominant method for large-scale remote sensing information extraction. Although this method has the advantages of mature algorithms and full use of human-computer interaction, However, there are also shortcomings such as long time-consuming, strong dependence on personnel involved in interpretation and analysis, low repeatability, and difficulty in obtaining large-area land cover information quickly, accurately and objectively.

Images