Method of adjusting green light passing time based on faster-rcnn

Abstract

The present invention relates to a kind of method based on Faster-RCNN regulation green light transit time, and the steps are as follows: (1) build waiting vehicle recognition model; First carry out feature extraction to obtain feature map, then extract candidate regions to obtain the features with candidate frames of different sizes map, then map to obtain small feature maps, and finally perform full connection operations on the small feature maps to obtain vehicle condition monitoring maps with recognition boxes; (2) train waiting vehicle recognition models; first establish a training set, then train feature extraction networks, and then share Fast-RCNN network and RPN network parameters, and finally perform model pruning; (3) Collect the vehicle situation monitoring map when the red light counts down from 20 to 40 seconds in the traffic intersection video, and input it into the waiting vehicle recognition model, and output it Vehicle status monitoring map with identification frame, count the number of waiting vehicles by counting the number of identification frames, and adjust the green light passing time according to the number of waiting vehicles. The model structure of the invention is simple, the method consumes less time, and can realize self-adaptive timing.

Description

technical field [0001] The invention belongs to the technical field of multi-target recognition, and relates to a method for adjusting green light passing time based on Faster-RCNN. Background technique [0002] The current traffic light timing is fixed, which will cause the green light time to be too short in some cases, resulting in incomplete clearance or too long to cause vehicles in other directions to wait too long, resulting in congestion. There are currently two solutions, one is to re-allocate the traffic flow forecast at the intersection, and the second is to perform adaptive timing according to the number of vehicles waiting at the current intersection. Compared with the two methods, although the adaptive timing of vehicles waiting at the current intersection has higher requirements for technology and processing equipment, it is more helpful to alleviate traffic congestion. [0003] For vehicles waiting for a red light, it can be modeled as vehicle detection in a...

AI Technical Summary

Problems solved by technology

Aiming at the fact that the fixed traffic light timing will reduce the traffic efficiency, the invention uses an algorithm based on identification and statistics of waiting vehicles to carry out adaptive timing; for the recognition algorithm existing in the traditional RCNN network and Fast-RCNN, it takes a long time and calls a third party The problem of time-consuming software, combined with the special application background of identifying waiting vehicles, the present invention invents a statistical method for waiting vehicles at red lights based on the Faster-RCNN algorithm and using model pruning optimization

Images