A road scene matching method and device

Abstract

The invention provides a road scene matching method and device. The method comprises: inputting the road scene picture to the target detection module, obtaining the target detection result corresponding to the road scene picture output by the target detection module; based on the delaunay triangulation algorithm, according to the target detection result, generating the adjacency diagram corresponding to the road scene picture; Input the adjacency graphs corresponding to any two road scene pictures to the graph matching module, and obtain the similarity between the two road scene pictures output by the graph matching module; if the similarity is greater than the preset threshold, the two road scene pictures are respectively The corresponding target in the feature matching is carried out, and the matching result is obtained. The method and device provided by the embodiment of the present invention improve the accuracy of scene matching, and the speed far exceeds the traditional scene matching method. It does not need to manually set and select parameters to widely adapt to road scenes, and at the same time solves the problem of the relevance of deep neural networks. .

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a road scene matching method and device. Background technique [0002] At present, among the high-precision maps that autonomous driving relies on, high-precision maps are the most basic safety guarantee for autonomous driving, and real-time updates are required to ensure the current situation of high-precision maps. The real-time update of high-precision maps is inseparable from crowdsourced collection, and crowdsourced visual data collection requires scene matching to determine data updates. [0003] In the current technical solutions, there are many ways to achieve scene matching, including scale-invariant feature transformation matching algorithm, thin-plate spline matching algorithm and accelerated robust feature matching algorithm. [0004] The traditional image processing method mainly uses the description operator of local feature points to detect the feature poi...

AI Technical Summary

Problems solved by technology

[0006] In the existing scene matching technology, the traditional image processing method can achieve scene matching better, but the calculation speed is very slow, and the parameters need to be selected, which cannot meet the current situation and scene uncertainty required in the real-time update of high-precision maps. sexual condition

Images