Method for extracting and simplifying three-dimensional road model

Abstract

The invention discloses a method for extracting and simplifying a three-dimensional road model, which comprises the following steps of: obtaining a multi-patch reference surface in a model of road elements to be extracted, constructing a hexagonal bounding box and orthogonally projecting the hexagonal bounding box to two dimensions, and greatly simplifying the operation difficulty of road sketching in a complex three-dimensional scene under the condition of retaining color and depth information, and the corresponding space and texture coordinates are obtained according to the road extraction area, so that the effect of automatically constructing the road model is achieved. According to the method, the original manual visual interpretation work is replaced by a computer space resolving technology, so that the extraction efficiency and accuracy of the road elements in a large-scale three-dimensional scene are effectively ensured.

Description

technical field [0001] The invention relates to the technical field of three-dimensional modeling, three-dimensional element extraction and simplification, in particular to a method for extracting and simplifying a three-dimensional road model. Background technique [0002] High-precision map scene modeling technology, especially road model extraction technology, has always been one of the most basic and important technical focuses in the field of autonomous driving. However, the current main methods of lane modeling still rely on a large number of manual operations. In large-scale scenes based on the real world (such as aerial oblique photogrammetry models), the amount of data is often very large, the model processing is very difficult, and road element extraction requires Professionals draw road network boundaries under various custom viewports. This method is extremely inconvenient to operate in a large scene. The selection of viewport position and posture requires a lot ...

AI Technical Summary

Problems solved by technology

However, the current main methods of lane modeling still rely on a large number of manual operations. In large-scale scenes based on the real world (such as aerial oblique photogrammetry models), the amount of data is often very large, the model processing is very difficult, and road element extraction requires Professionals draw road network boundaries under various custom viewports. This method is extremely inconvenient to operate in a large scene. The selection of viewport position and posture requires a lot of time and energy. At the same time, due to the viewport of other elements Occlusion and coverage often lead to missing or repeated extraction of road network elements

Images