A traffic light localization method based on multiple clustering

Abstract

The invention relates to a traffic light positioning method based on multiple clustering, comprising: step 1, collecting traffic light data in an area; step 2, grouping the traffic light data to obtain data sets corresponding to different headings; step 3, respectively Block the data of each data set to obtain the data blocks corresponding to different collection vehicles and collection periods in each data set; step 4, traverse each of the data sets, and determine the number of traffic lights in the data set based on the DBSCAN clustering algorithm , taking the obtained number of traffic lights as the number of cluster centers, and based on the K-Means clustering calculation, the cluster center is the traffic light coordinate result. Based on the DBSCAN clustering algorithm to find the number of traffic lights, and based on the K-Means clustering algorithm to find the cluster center multiple clustering, the accuracy and positioning accuracy of the number of traffic lights obtained by traditional single clustering are improved.

Description

technical field [0001] The invention relates to the field of map generation under crowdsourcing mode, in particular to a traffic light positioning method based on multiple clustering. Background technique [0002] Among various map service technologies, whether the position of traffic lights on the map is accurate is of great significance to map navigation and assisted driving. [0003] The traffic light positioning method in the prior art is mainly to first identify the traffic light with a camera, and then rely on the radar to detect the relative position of the traffic light and the car and calculate the GPS coordinates of the car itself. However, the recognition accuracy of traffic lights detected by radar is not 100%, and the position of traffic lights needs to be calculated by the car's own GPS after recognition. Due to the possible error between GPS and relative position, the accuracy of traffic lights will be poor. . [0004] In addition, when the error of the coll...

AI Technical Summary

Problems solved by technology

However, the recognition accuracy of traffic lights detected by radar is less than 100%, and the position of traffic lights needs to be calculated by the car's own GPS after recognition. Due to the possible error between GPS and relative position, the accuracy of traffic lights will be poor.

[0004] In addition, when the error of the collected data is large, direct clustering using traditional clustering algorithms (such as the DBSCAN algorithm) will have greater instability, which may lead to inaccurate numbers of traffic lights and poor positioning accuracy

Images