An intelligent flow recognition and statistics method for complex traffic intersections

Abstract

The invention belongs to the technical field of intelligent transportation, and discloses an intelligent traffic identification and statistical method for complex traffic intersections, constructs a video data set of complex traffic intersections, and trains a depth detection model; points the intersections with a mouse, automatically divides the detection area, and Numbering; use the trained deep learning model to detect the vehicle in the video screen; use the target tracking algorithm to predict the position of the vehicle in the next frame according to the detected vehicle and the tracking vehicle in the previous frame; according to the change of the intersection number of the vehicle, Count the traffic flow; repeat steps 3 to 5 until the video ends or is artificially stopped. The present invention can automatically identify vehicle types in various complex traffic intersection environments, count the quantity according to vehicle types, and count the traffic flow between different intersections, has the advantages of accurate statistical results and good real-time performance, and has high engineering practical value .

Description

technical field [0001] The invention belongs to the technical field of intelligent transportation, and in particular relates to an intelligent traffic identification and statistical method for complex traffic intersections. Background technique [0002] At present, with the continuous development of social economy, in order to meet the needs of people's daily travel, the demand for cars is expanding year by year. Coupled with the scarcity of land resources in modern cities, every inch of land is expensive, and traffic jams continue to appear and become more common and serious. Especially in the peak hours of commuting, road congestion is even more serious. This has seriously affected people's life and work efficiency in modern society, and has become a social problem that cannot be ignored and urgently needs to be improved or solved. The concept of urban brain, smart transportation and big data has given us the hope of effectively solving the problem of traffic congestion....

AI Technical Summary

Problems solved by technology

[0006] The difficulty of solving the above problems and defects is: in the previous technology, due to the limitation of computer computing power, it is often impossible to use the target detection algorithm based on deep learning with higher detection accuracy, so the previous technology often fails to meet expectations in terms of traffic statistics The goal

On the other hand, due to the lack of application of road division-related technologies in traffic flow statistics, many methods for counting vehicles can only be applied to straight roads, but cannot be applied to other complex traffic intersections such as T-junctions and crossroads. There are related technologies to achieve vehicle steering statistics at complex traffic intersections

Images