Traffic overflow state recognition method based on video

Abstract

The invention discloses a traffic overflowing state recognition method based on video. The traffic overflowing state recognition method based on the video comprises the following steps that two cameras are erected on a target crossroad, the first camera monitors the queuing behavior of vehicles, in the overflowing occurrence direction, inside the crossroad, and the second camera monitors a signal lamp in the direction crossing the overflowing occurrence direction; pre-processing is carried out on video images monitored by the first camera, and video features including the vehicle existence ratio and the movement pixel ratio are extracted; image recognition is carried out through a blur recognizer with the feedback function; monitoring images of the second camera are processed, and positioning and recognition of the signal lamp are achieved; the signal lamp recognition result is output; the traffic overflowing result is obtained according to the true and false overflowing decision rule; traffic overflowing state display and alarm are carried out. According to the traffic overflowing state recognition method based on the video, real-time monitoring is carried out on traffic information inside the crossroad through the cameras, automatic recognition of the traffic overflowing state is achieved through the blur recognition technology with the feedback function, and support is provided for prevention and relief of traffic jams.

Description

technical field [0001] The invention relates to a video-based traffic overflow state recognition method. Background technique [0002] Traffic overflow is a traffic phenomenon in which some vehicles occupy the upstream intersection due to the bottleneck effect of the downstream intersection, which makes the queuing of vehicles on the road section exceed the length of the road section. The direct consequence of it is to cause traffic congestion. If it is not controlled after the occurrence, the congestion phenomenon will spread from a single intersection to the surrounding area, leading to chain congestion, and even gridlocks, resulting in traffic paralysis in a large area of ​​the road network. [0003] At present, the theoretical research and technical application of traffic overflow at home and abroad are mostly concentrated in the field of its occurrence mechanism and control methods. The detection of traffic overflow is mainly realized by traditional coil detectors. The ...

AI Technical Summary

Problems solved by technology

Its direct consequence is to cause traffic congestion. If it is not controlled after the occurrence, the congestion phenomenon will spread from a single intersection to the surrounding area, leading to chain congestion, and even gridlocks, causing traffic paralysis in a large area of ​​the road network.

[0003] At present, the theoretical research and technical application of traffic overflow at home and abroad are mostly concentrated in the field of its occurrence mechanism and control methods. The detection of traffic overflow is mainly realized by traditional coil detectors. The literature [GeroliminisN, SkabardonisA. ].TransportationResearchBoard (TRB) 89thAnnualMeeting,Washington,D.C.2010:10-3498.]Proposed an overflow identification method based on coil detector occupancy data, and gave the corresponding threshold, but the coil detector has complex installation and maintenance procedures, High cost and other shortcomings, and the accuracy of traffic overflow detection results through coil detectors is largely dependent on the layout distance of the detectors, so this method has great limitations in practical applications

Literature [Zhang Lidong, Jia Lei, Zhu Wenxing. Traffic overflow recognition algorithm based on fuzzy theory[J]. Computer Application, 2012,32(8):2378-2380.] proposed a fuzzy theory-based overflow recognition method to obtain vehicle Queuing ratio and average vehicle speed and other indicators and fuzzy reasoning, so as to realize the overflow state identification in the simulation environment, but the above indicators are difficult to obtain in practical applications, so this method also has limitations in practical applications

The core link of traffic overflow state identification is the detection of vehicle queuing state inside the intersection, literature [He Xiaofeng, Yang Yuzhen, Chen Yangzhou. Detection of vehicle queuing length based on video image processing [J]. Traffic and Computer, 2006, 24(5) :43-46.] Proposed a video-based method of motion detection first and then presence detection, which can quickly detect the length of the vehicle queue, but cannot give the dynamic formation process of the vehicle queue, and cannot provide the necessary conditions for the identification of the overflow state

At present, manual observation is mostly used in China to identify traffic overflow, and the efficiency is low

Images