A road condition prediction method and device

Abstract

The invention discloses a road condition prediction method and device. The method comprises the following steps: converting road condition information of a road section into a topological structure diagram; nodes of the topological structure diagram represent road sections; edges represent connection relations of the road sections; taking the topology map as a first input of a prediction model, the prediction model comprising a convolution layer, a pooling layer and a full connection layer arranged in order; taking N-1 downsampling result output by the pooled layer output as the Nth inputof the volume layer until N reaches the set number of times, N is a positive integer greater than or equal to 2. The Nth downsampling result of the pooling layer is inputted into the whole connectionlayer to perform the whole convolution operation, and the road condition prediction result of the road section output by the prediction model is obtained. The technical proposal provided by the application can effectively extract valuable features for road condition prediction by performing convolution neural network on the topological structure diagram obtained by road condition conversion, thereby realizing accurate road condition prediction.

Description

technical field [0001] The invention relates to the field of real-time traffic technology, in particular to a method and device for predicting road conditions. Background technique [0002] With the continuous increase of vehicles, the roads are becoming more and more congested. Therefore, in large cities, how to achieve accurate road condition prediction has important guiding significance. Road condition prediction refers to the use of historical road condition information and static road network information to predict future road condition information. Future traffic information helps to improve the excellent rate of ETA (estimated travel time), and helps users avoid future congestion. [0003] A solution for road condition prediction in the prior art is to use a CNN network structure for road condition prediction. CNN is a convolutional neural network, which belongs to a transformation operation in deep learning, and is often used to extract the local area of ​​a regular...

AI Technical Summary

Problems solved by technology

Stationarity is one of the prerequisites for the application of CNN; therefore, the conventional LSTM-CNN network structure cannot effectively extract features that are valuable for road condition prediction, and thus it is difficult to achieve accurate road condition prediction

Images