Vehicle track prediction method and system based on Gaussian mixture model

Abstract

The invention discloses a vehicle track prediction method and system based on a Gaussian mixture model. The vehicle track prediction method comprises the following steps of (1) expressing an acquired track T to consist of time t, current time speed v and angular deviation phi corresponding to N in a 2D plane of a vehicle coordinate system; (2) adopting a phi<V> expressing method to describe the track, so as to obtain a unified expression of the track T; (3) predicting the track T based on the distribution of the Gaussian mixture model, and predicting the speed c<v> and angular deviation c<phi> at a future time; and (4) through the statistical characteristics of calculating condition distribution p(X<f>|X<h>), predicting the c<v> and / or the c<phi>, wherein the X<f> is an approximate future track, and the X<h> is a historical track. Through the adoption of the vehicle track prediction method disclosed by the invention, the tract situation of the vehicle can be predicted in advance, and the future track distribution can also be predicted, so that potential danger can be avoided in time.

Description

technical field [0001] The invention relates to the field of computer vision technology and image processing of an advanced driving assistance system, in particular to a vehicle trajectory prediction method and system based on a Gaussian mixture model. Background technique [0002] With the advent of the automobile era, the number of motor vehicles has increased greatly, and more and more attention has been paid to automobile safety technology. Among them, the advanced driver assistance system (ADAS) plays a vital role in the field of automobile safety. [0003] In the advanced driver assistance system (ADAS) based on computer vision technology, accurate and reliable prediction of vehicle trajectory is of great significance for improving the comfort of the system, predicting potential threats in advance and protecting driver safety. Currently, intersections are still a big challenge for driver assistance systems. According to statistics, the most common vehicle accidents o...

AI Technical Summary

Problems solved by technology

Depending on the accuracy and complexity of the underlying kinematic model, model-based predictions (e.g., assuming constant velocity and yaw angle) can lead to large deviations from the true trajectory and prediction failures, especially in the case of steering maneuvers

For example, in some methods Morzy M. Mining frequent trajectories of moving objects for location prediction. In: Proc. of the 5th Int'l Conf. on Machine Learning and Data Mining in Pattern Recognition. LNCS 4571, Heidelberg: Springer-Verlag, 2007.667-680 .It is a combination of prefix tree PrefixSpan and frequent pattern mining FP-tree algorithm proposed by Morzy et al. to mine dynamic motion rules of moving objects, but the time cost of building prefix tree and FP-tree is high

Among other methods, Pan TL, Sumale A, Zhong RX, Indra-Payoong N.Short-Term traffic state prediction based on temporal-spatial correlation.IEEE Trans.on Intelligent TransportationSystems,2013,14(3):1242-254.Pan proposed an optimal linear predictor based on multivariate normal distribution. The disadvantage of this method is that the prediction will cause delay and cannot be applied to real-time monitoring of traffic flow.

Images