Intersection retention event detection method and system based on machine learning

Abstract

The invention relates to an intersection retention event detection method and system based on machine learning, and belongs to the field of computer vision. The intersection retention event detection method comprises the steps of: collecting multiple consecutive frames of images of a to-be-detected intersection in a set time period; according to each frame of image, obtaining position information and speed information of each vehicle target frame in a real coordinate system; for an ith frame of image, obtaining a multi-dimensional feature vector of the ith frame of image according to the position information and the speed information of each vehicle target frame in the ith frame of image and the adjacent first three frames of images, i being greater than 3; based on a retention detection model and according to the multi-dimensional feature vector of the ith frame of image, determining a retention state of the to-be-detected intersection at the time point corresponding to the ith frame of image; and determining a retention event of the to-be-detected intersection according to the retention state of the to-be-detected intersection at each time point. The determined multi-dimensional feature vector has continuity and robustness between a former frame and a latter frame, the misjudgment of a prediction result can be reduced, and the accuracy of intersection retention event detection is improved.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a machine learning-based detection method and system for stranded events at intersections. Background technique [0002] For traffic congestion, scholars at home and abroad have given different definitions from different perspectives. Generally speaking, traffic congestion is a traffic operation situation in which traffic demand exceeds traffic supply. Vehicles are on the roadway outside the intersection without signal lights. Congested intersections are defined as intersections where vehicles that are blocked and have a queuing length of more than 250m, or vehicles that fail to pass through the intersection at an intersection controlled by signal lights for 3 green lights; congested road sections are defined as vehicles blocked on the roadway and the queuing length exceeds 1km. [0003] At present, the existing traffic state (including traffic congestion) detection algorithm can b...

AI Technical Summary

Problems solved by technology

[0004]However, the state of traffic stranded in the actual scene is a dynamic process of change, and most of the existing technical schemes for stranded detection, whether direct or indirect, are only for a single The state under the time node is classified and judged, without considering the information association between the previous and subsequent frames, resulting in poor robustness of the algorithm, and false positives or missed negatives are prone to occur in some special scenarios

Images