A crowdsourcing mapping method and device based on color feature distribution and image semantics

Abstract

The present invention relates to a crowdsourcing mapping method and device based on color feature distribution and image semantics. The method includes: acquiring a crowdsourcing image, sequentially performing enhancement processing and grayscale conversion on the crowdsourcing image to obtain a grayscale image; The grayscale image is divided into grids, and the grid to be counted is determined according to the semantic information and the dynamic target of the grayscale image; the color characteristics of the original image corresponding to each grid to be counted are counted, and according to the The color feature determines the grid required for crowdsourcing mapping; the crowdsourcing map is constructed according to the image corresponding to the grid of crowdsourcing mapping. The present invention provides judging basis for multi-batch data fusion with less amount of data through image color feature distribution information, color statistical information and image semantic information, effectively increasing the stability of crowdsourced map data matching.

Description

technical field [0001] The invention belongs to the field of high-precision map making, and in particular relates to a crowdsourcing map building method and device based on color feature distribution and image semantics. Background technique [0002] As an important data source in the field of autonomous driving, high-precision maps can effectively assist autonomous vehicles in positioning and path planning. However, the collection and production cycle of professional high-precision maps is long, resulting in insufficient data updates. Crowdsourcing is required. Methods A crowdsourced layer was established to improve the real-time performance of high-precision maps. [0003] The crowdsourced map is based on the image semantic data of the crowdsourced collection vehicle for 3D reconstruction to generate local map data, and a large number of local maps can be fused to construct a crowdsourced map. The process of data fusion requires good matching between local maps. Due to ce...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problem of poor stability of crowdsourcing map data matching, the first aspect of the present invention provides a labeling method for target detection, which acquires crowdsourcing images, performs enhancement processing and grayscale conversion on the crowdsourcing images in sequence, Obtaining a grayscale image; dividing the grayscale image into grids, determining the grids to be counted according to the semantic information and dynamic targets of the grayscale images; counting the color of the original image corresponding to each of the grids to be counted Features, according to the color feature to determine the grid required for crowdsourcing mapping; construct a crowdsourcing map according to the image corresponding to the grid of the crowdsourcing mapping

Images