Short-term traffic flow prediction algorithm based on multi-modal deep learning

By employing a multimodal deep learning-based algorithm, which utilizes Gram angle field transform and convolutional neural networks to fuse image and temporal features, and combines temporal convolutional networks and multi-head attention mechanisms, the accuracy of short-term traffic flow prediction is improved, the complexity of traffic flow prediction is addressed, and optimization decision-making in intelligent transportation systems is supported.

CN122416725APending Publication Date: 2026-07-17ZHEJIANG SCI RES INST OF TRANSPORT +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610537424.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-04-22
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Short-term traffic flow forecasting is affected by factors such as time, space, weather, and holidays, resulting in low accuracy and difficulty in effectively alleviating traffic congestion.

Method used

A multimodal deep learning-based algorithm is used to generate image features through Gram angle field transform. By combining convolutional neural networks and temporal convolutional networks, image and temporal features are fused, and multi-head attention mechanism is used to enhance information at key time steps for traffic flow prediction.

Benefits of technology

It significantly improves the accuracy of short-term traffic flow forecasting, better supports intelligent management and dynamic guidance of the traffic system, and alleviates traffic congestion.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122416725A_ABST
    Figure CN122416725A_ABST
Patent Text Reader

Abstract

基于多模态深度学习的短时交通流预测算法，其具体步骤如下：S1，将交通流量、占用率和平均速度三个特征分别进行格拉姆角场变换，生成对应的格拉姆角场图，并将其分别作为红、绿、蓝三个通道合并成一张完整的彩色图像；S2，将步骤S1中得到的彩色图像输入卷积神经网络以提取深层空间特征，得到特征向量，并将特征向量与时间特征进行融合得到最终的序列特征；S3，将步骤S2中得到的最终的序列特征送入时间卷积网络捕捉时序依赖，最后通过多头注意力机制进一步强化关键时间步的信息，最终输出交通流量的预测结果；其中所述时间卷积网络采用深度扩张卷积。
Need to check novelty before this filing date? Find Prior Art