行人预测直接影响到自动驾驶系统的安全性和可靠性,尤其是密集人群场景的行人预测.传统的人群预测方法通过对人群进行优先级分类,再按照不同优先级对行人逐个预测.但是,在密集人群场景,因为预测目标较多,即使正确划分了优先级,单纯依靠这种方法也会带来很大的处理时延.使用空间密度聚类(Density-Based Spatial Clustering of Applications with Noise,DBSCAN)算法对密集人群场景进行分析,选择合理的聚类算法参数,在聚类结果的基础上,结合传统优先级分类算法,简化人群的处理,并提出多线程并行处理的方法,以提高聚类算法的效率.通过对DBSCAN算法应用于人群预测的分析,丰富了人群预测的方法,为优化自动驾驶行人预测的表现提供了重要参考.
Analysis of the Method of Simplifying Crowd Prediction by DBSCAN Algorithm
Pedestrian prediction has a direct impact on the safety and reliability of the autonomous driving system,especially in dense crowded scenes.Traditional crowd prediction methods classify crowds based on priority and then predict pedestrians one by one according to different priorities.However,in dense crowd scenarios,due to the large number of prediction targets,even if priority is correctly assigned,relying solely on this method will bring signifi-cant processing delays.This article uses Density-Based Spatial Clustering of Applications with Noise(DBSCAN)al-gorithm to analyze dense crowd scenes and select reasonable clustering algorithm parameters.Based on the cluster-ing results,traditional priority classification algorithms are combined to simplify crowd processing,and a multi-threaded parallel processing method is proposed to improve the efficiency of clustering algorithms.By analyzing the application of DBSCAN algorithm in crowd prediction,the methods of crowd prediction have been enriched,pro-viding important references for optimizing the performance of autonomous driving pedestrian prediction.
Crowd predictionDensity-Based Spatial Clustering of Applications with NoiseCrowd clusteringAlgorithm optimization
万方数据
维普
