Road overflow real-time detection method based on deep learning

Abstract

The invention discloses a road overflow real-time detection method based on deep learning. The method comprises the following steps: 1) setting a camera preset position and calibrating a camera; 2) initializing a convolutional neural network model; 3) acquiring a real-time video stream; 4) checking the working state of the camera; 5) performing vehicle target detection on the region of interest byusing a convolutional neural network model; 6) carrying out vehicle multi-target tracking based on IoU matching; 7) analyzing road overflow; and 8) reporting the road overflow event and setting thesleep state. According to the road overflow real-time detection method provided by the invention, vehicle target detection is carried out by using a deep convolutional neural network (CNN), and vehicle multi-target tracking is carried out by combining a tracking algorithm based on IOU matching, so that road overflow detection can be accurately and quickly carried out.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a real-time detection method for road overflow based on deep learning. Background technique [0002] Road overflow refers to the area where vehicles line up outside the road entrance due to traffic accidents, road capacity and other issues at the road entrance. Road overflow will cause vehicles to pile up on intersection roads, affect the traffic of multiple driving methods, and even lead to the paralysis of the entire traffic network. Therefore, it is very important to detect road overflow accurately and in real time. [0003] Most of the traditional road overflow detection methods are based on the statistical data collected by the ground induction coil for overflow analysis, but the construction of the ground induction coil needs to cause damage to the road surface and the construction is complicated, easy to damage, difficult to repair, and may also misdetect the parkin...

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the shortcomings of low detection accuracy, slow detection speed, and poor robustness in the prior art, the purpose of the present invention is to provide a real-time detection method for road overflow based on deep learning, and use deep convolutional neural network (CNN) for vehicle target detection , combined with vehicle multi-target tracking based on IoU matching, this algorithm can perform road overflow detection accurately and quickly

Images