A method for accurately identifying fueling station metering based on machine vision

By using an improved HorNet network and temporal feature analysis, the instability of gas station metering identification in complex environments was solved, achieving stable and consistent identification of metering data and refueling behavior, thus improving the accuracy and reliability of metering monitoring.

CN122244754APending Publication Date: 2026-06-19SMART OIL CUSTOMER (BEIJING) NETWORK TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SMART OIL CUSTOMER (BEIJING) NETWORK TECHNOLOGY CO LTD
Filing Date
2026-03-13
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing gas station metering identification methods are unstable in complex environments, lack comprehensive analysis of the refueling process, and lack constraints on the continuous variation of metering data, resulting in inaccurate metering identification results.

Method used

An improved HorNet network is used to guide high-order spatial interaction and cross-frame high-order stable structure in the metering area. Combined with behavioral state modulation, the metering data on the fuel dispenser display screen and the fuel nozzle operation behavior are jointly identified through temporal feature analysis, generating a temporally stable metering feature sequence and performing consistency judgment.

Benefits of technology

It improves the stability and environmental adaptability of metering identification, enhances the ability to identify abnormal metering, ensures the consistency between metering results and refueling behavior, and improves the reliability of metering monitoring.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122244754A_ABST
    Figure CN122244754A_ABST
Patent Text Reader

Abstract

This invention discloses a machine vision-based method for accurate metering identification at gas stations, comprising the following steps: acquiring and preprocessing video image data of the refueling pump operating area, locating the refueling area, and constructing a metering-related image sequence; inputting the metering-related image sequence into an improved HorNet network to generate a metering spatial feature sequence; performing cross-frame feature updates and anomaly jump suppression on the metering spatial feature sequence to generate a time-stable metering feature sequence; performing character enhancement, perspective correction, and digit segmentation on the display screen features to generate a metering digit sequence; modulating the metering features by combining the fuel nozzle operation action features and vehicle presence features to generate a refueling metering-related sequence; performing metering consistency judgment and anomaly identification on the refueling metering-related sequence, and outputting the accurate refueling metering identification result. This invention achieves accurate identification of refueling metering information, improving the stability of metering monitoring and anomaly identification capabilities.
Need to check novelty before this filing date? Find Prior Art