Building contour extraction method fusing image features and dense matching point cloud features

Abstract

The invention provides a building contour extraction method fusing image features and dense matching point cloud features. The building contour extraction method comprises the following steps: step 1,generating a three-dimensional dense image matching (DIM) point cloud and an orthoimage from an aerial image; step 2, fusing the DIM point cloud and the aerial image to detect a mask of a single building; 3, carrying out rough building contour line detection based on line matching; and 4, regularizing the boundary of the building by fusing the matching line and the contour line of the rough building mask. According to the invention, the ground object classification precision can be improved; the detection precision can be improved; a regularized building contour line can be generated.

Description

technical field [0001] The invention relates to the field of surveying and mapping science and technology, in particular to a method for extracting building outlines by fusing image features and densely matched point cloud features. Background technique [0002] Buildings are important features in urban areas. How to extract buildings from scenes has always been the focus and focus of research by scholars at home and abroad. Oblique aerial images have the advantage of "visible facades without occlusion", which can not only provide information on building roofs, but also provide information on building facades. Therefore, it is of great research value and social benefit to study how to extract buildings from oblique aerial images. [0003] At present, the literature on the fusion of DIM (Dense Image Matching, Dense Image Matching) point cloud features and image features is mainly concentrated in the detection field related to individual vegetation, while there are few litera...

AI Technical Summary

Problems solved by technology

[0003] At present, the literature on fusing DIM (Dense Image Matching, Dense Image Matching) point cloud features and image features is mainly concentrated in the detection field related to individual vegetation, while there are few literatures involving fusion of image and DIM point cloud features to detect buildings

The first method can detect the building facade on the oblique aerial image, and use the detected facade to generate a building model, and then use the DIM point cloud to verify the building, so as to achieve the purpose of building modeling; The two methods can use the line segment information in the image and the DIM point cloud for building verification; in these two methods, the facade is the most important feature, which leads to the problem that such algorithms still have insufficient ability to detect small buildings

The third method can use the DIM point cloud to generate a height map, a height standard deviation map, and a height gradient map, and at the same time segment the original image at different scales, extract the information of the segmented image patches, and then sequentially detect the grass according to the defined rules , trees, facades, roofs and roads; different from the first two methods, this method makes full use of the features provided by the image to detect various ground objects, which makes up for the shortcomings of the first two methods to a certain extent; however, this There are still some defects in the method: on the one hand, this method is a building detection method based on hierarchical semantics, the quality of the feature threshold seriously affects the accuracy of object classification, and the selection of the threshold is more difficult; on the other hand, the The method only uses 9 features to detect 5 types of ground objects, resulting in a low detection accuracy, and the roof mapping accuracy is only 81%.

[0004] In view of the above problems, the present invention provides a building outline extraction method that fuses image features and densely matched point cloud features to solve the problem that the quality of feature thresholds in the prior art seriously affects the accuracy of ground object classification, and the detection accuracy is not high. The problem

Images