Clustering method and system for road object elements of crowdsourced map, and storage medium

A clustering method for road object elements of a crowdsourced map, the method being executed by a processor for mapping a high-precision map to assist in positioning and planning of an autonomous vehicle and comprising the following steps:

• S1:
• obtaining road object data of the crowdsourced map, classifying road object elements, randomly selecting a type of road object data, computing centroid coordinates of road objects, and putting the centroid coordinates of the road objects in a same category into a list type_dblist;
• S2:
• setting a bandwidth distance band_width, a threshold Inter0 of a ratio of an intersection area to a minimum area, a threshold epsilon of a distance between two centroid coordinates, and a threshold min_max0 of a ratio of the minimum area to a maximum area between an initial cluster center and a cluster sample point, as well as an initial value of a cluster category label cluster_id;
• S3:
• when the list type_dblist is not empty, randomly selecting a point from the list type_dblist as the initial cluster center, and if the cluster center is an invalid value, ending the current loop and selecting a new point;
• otherwise, putting the centroid coordinates and original data of the cluster center into a list sample_count1 and a list sample_count2 respectively;
• S4:
• drawing a circle with the cluster center as a center and the bandwidth distance band_width as a radius;
• circularly computing a centroid distance dis_e, a ratio Inter of an intersection area to a minimum area, and a ratio min_max of the minimum area to a maximum area between each point within the circle and the cluster center;
• putting the centroid coordinates of the points within the circle that simultaneously satisfy dis_e<epsilon, Inter<Inter0, and min_max<min_max0 and their original data into the list sample_count1 and the list sample_count2 respectively;
• after the points within the circle are looped through, if a length of the list sample_count1 is greater than a preset length, computing a mean offset value according to the points in the list sample_count1 to obtain a new cluster center;
• if the distance between the current cluster center and the new cluster center is ≥epsilon, performing step S5;
• otherwise, repeating step S4 with the new cluster center as a center;
• S5:
• if the length of the list sample_count1 is >the preset length, putting the clustering results stored in the list sample_count1 and the list sample_count2 correspondingly into a list groups1 and a list groups2, and increasing the cluster_id by 1;
• if the length of the list sample_count1 is ≤the preset length, skipping storing the clustering results into the list groups1 and the list groups2;
• solving a difference between the list type_dblist and the list sample_count1, using the result of the difference as a new list type_dblist, cycling steps S3-S5 until all samples of the same type are clustered;
• S6:
• obtaining centroid coordinates of a first cluster in the groups1, and computing a Euclidean distance between the centroid coordinates of the first cluster and the centroid coordinates of any other cluster;
• if the distance is less than the bandwidth distance band_width, merging the two clusters into a same cluster, and selecting the cluster_id of the longer cluster as a new cluster_id;
• looping through the list groups1 until all clusters in the list groups1 are determined;
• and
• S7:
• cycling steps S1-S6 until all types of samples are clustered.