A Feature Point Filtering Method of Oblique Image with Double Constraints of Semantic and Structural Information

Abstract

The invention discloses a method for filtering oblique image feature points with double constraints of semantic and structural information, and relates to the field of image processing. The present invention comprises the following steps: identification of vegetation area and extraction of line features: extracting vegetation information in oblique images by means of difference vegetation index in visible light bands, and simultaneously calculating feature line features of ground objects in images by means of improved line detection algorithm; classification of feature point binary groups Model: take the results of vegetation area recognition as the semantic constraints of feature point filtering, and the results of line feature recognition as the structural constraints of feature point filtering, establish a feature point classification model, and divide feature points into four categories; feature point progressive selection algorithm: build Gaussian pyramid, determine the threshold of the number of feature points, and use this threshold as the approximation condition, traverse the Gaussian pyramid layer by layer to select feature points until the number threshold is met. The invention effectively improves the filtering quality of feature points.

Description

technical field [0001] The invention relates to the field of image processing, and more specifically relates to a method for filtering oblique image feature points with dual constraints of semantic and structural information. Background technique [0002] In recent years, the use of ISFM (Incremental Structure From Motion) method emerging in the field of computer vision to realize the research of oblique photography 3D scene reconstruction has attracted extensive attention from scholars at home and abroad, and has formed feature matching, spatial adjustment, dense The three steps of network construction are the core reconstruction process. Among them, feature matching is the basis, and algorithms such as SIFT (Scale-invariant feature transform) and SURF (Speeded UpRobust Features) are usually used to extract feature points, and then the features of the same name are used for matching. However, in the 3D reconstruction process of large-scale oblique photography, the feature ...

AI Technical Summary

Problems solved by technology

The former takes global single-scale feature points as the research object, and uses machine learning (such as bag-of-words model, vocabulary tree, neural network, etc.) to classify and filter feature points. Therefore, in practical applications, more scholars choose the latter for automatic filtering of feature points, that is, to establish a differential Gaussian pyramid , feature point filtering on the global multi-scale

Images