A highway traffic anomaly early warning method and system based on edge intelligence

CN122157494APending Publication Date: 2026-06-05CHENGDU TONGGUANG NETLINK TECH CO LTD
View PDF 6 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CHENGDU TONGGUANG NETLINK TECH CO LTD
Filing Date
2026-05-08
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing methods for identifying highway traffic anomalies rely on server-side information assessment or model training, making it difficult to perform specific identification based on different regions and time periods, resulting in insufficient accuracy.

Method used

By deploying lightweight models on edge devices and combining them with the alienation of models trained in the cloud, multi-source data fusion and abnormal event identification are performed. CNN models are used for feature extraction and Softmax classification. Thresholds are set to judge unknown anomalies and output them over the network.

Benefits of technology

It improves the recognition accuracy of edge devices, reduces communication latency, enhances the recognition efficiency and accuracy of the model, prevents overfitting, and supports model updates and optimization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122157494A_ABST
    Figure CN122157494A_ABST
Patent Text Reader

Abstract

The application discloses a highway traffic abnormality early warning method and system based on edge intelligence, and relates to the technical field of intelligent transportation. The method comprises the following steps: collecting cloud historical data, edge historical data and edge collected data to be detected; performing data fusion on picture data, point cloud data and road surface state data to obtain information tensors; establishing a cloud recognition model; adopting a model alienation method to alienate the cloud recognition model to obtain an edge execution model; and bringing the information tensors of the edge collected data into the edge execution model to identify abnormal events and realize the early warning function. Through the fusion of multi-source data, the multi-source data of each sample is fused into information tensors, the format is unified and the information is digitized and regularized, and after the model is well trained in the cloud, the model is downloaded to the edge device, and then the edge device is locally alienated according to the environment, so that the identification accuracy in the environment of the edge device is improved.
Need to check novelty before this filing date? Find Prior Art