Building detection method of remote sensing images based on multi-scale and multi-feature fusion

Abstract

The invention discloses a remote sensing image building detection method based on multi-scale and multi-feature fusion, which includes down-sampling the high-resolution remote sensing image to obtain an image pyramid composed of images of different scales; calculating the edge images of the image pyramid; Multi-group feature calculation and fusion are performed on the scale edge image to establish a feature model; window selection is performed according to the feature model and local non-maximum suppression in the neighborhood to obtain the target window; the expansion / shrinkage calculation is performed on the target window in a small range to obtain a rectangular window ; Rotate the rectangular window according to the main direction of the target window to obtain an optimal target window, and extract buildings according to the optimal target window. Its remarkable effect is: multi-scale building detection is performed on Gaussian pyramid images, and the detection of buildings with different sizes, shapes, and orientations is universal; and the accuracy and efficiency of automatic building detection are effectively improved.

Description

technical field [0001] The invention relates to the technical field of remote sensing image processing, in particular to a remote sensing image building detection method based on multi-scale and multi-feature fusion. Background technique [0002] Buildings, as one of the main types of urban features, are thematic elements that must be highlighted in the large-scale basic geographic map of the city. Accurate building information can be used for land management, urban planning and other government departments to carry out land use status investigation and macro planning It provides important decision support for work such as digital city construction, investigation of illegal buildings and military reconnaissance and other fields. [0003] In high-resolution images, buildings have various shapes and sizes, and some are even obscured by neighboring trees, resulting in missing shapes. It is difficult to describe with a unified shape model, making automatic detection of buildings...

AI Technical Summary

Problems solved by technology

However, most of the existing features only take into account the texture information of the building, while ignoring the special structure of the building, etc., and it is difficult to solve the shadow occlusion of buildings and the different sizes, shapes, and orientations of buildings in high-resolution remote sensing images. question

At the same time, most of the current methods with better extraction effects rely on the support of LiDAR, DSM, GIS vector data and other auxiliary data, which complicates the detection process and data and reduces the extraction efficiency.

Images