Technical Implementation of Guangzhou's "Open Four Stop Four" Traffic Restriction System

To alleviate traffic congestion in Guangzhou, the city has enforced a "Open Four Stop Four" rule for non‑local vehicles since July 1, 2018. This article analyzes the technical realization of the rule, focusing on three major functional modules.

1. Electronic Police Image Capture and Storage

The system leverages mature electronic surveillance solutions (e‑police, checkpoint cameras, security monitoring, traffic monitoring) to collect image or video data. Approximately 170 devices were deployed for the "Open Four Stop Four" program, though any traffic camera can provide the necessary data. The architecture includes an electronic checkpoint layout and a capture system (see Fig. 2‑1 and Fig. 2‑2). Massive image/video data can be stored using Tencent Cloud File Storage (CFS) to ensure elastic scaling and shared access.

2. Out‑of‑Town License Plate Recognition and Ingestion

Captured images are processed with machine‑learning based image/video key‑frame techniques to extract out‑of‑town license‑plate information, which is stored in the LimitedPassDB.originalInfoTable . The recognition technology is widely used in parking lots and highway checkpoints. A system diagram (Fig. 3) illustrates the workflow. Tencent Cloud’s "Intelligent Ti" machine‑learning platform can be used for model training, evaluation, and prediction, supporting pipelines such as image upload to COS, preprocessing, Faster R‑CNN training, and deployment for real‑time vehicle identification.

3. Violation Detection

The core module determines illegal vehicles based on a bitmap encoding of daily travel patterns. The allowed travel codes (e.g., 100001, 1100001, …) reduce complexity. The process includes:

Recording each out‑of‑town vehicle sighting as a trace record with fields like flagStr , timestamp , and isIllegal . The first sighting sets flagStr to "1".

Appending "1" to flagStr for subsequent daily sightings; appending "0" at day‑end for days without a sighting.

Running a daily scheduled task to evaluate flagStr . If the flag contains "0000", the record is cleared (vehicle complied with the "stop four" rule). Otherwise, if the flag does not match any of the predefined compliant patterns, the vehicle is marked illegal, an illegal record is added, a fine is issued, and the flag is shifted left by one character.

The pseudo‑code for the daily evaluation is:

if '0000' in flagStr:
    // flagStr = ''
    // isIllegal = 0
    deleteTraceRec()  // clear traceInfoTable entry
else:
    if flagStr not in (11110,11000,11100,10100,10010,10110,11010,111100,111000,101000,100100,101100,110100,1111000,1001000,1011000,1101000):
        isIllegal = 1
        addIllegalRec()   // add record to illegalInfoTable
        sendBill()        // issue fine
    flagStr = flagStr[1:] // drop the oldest flag

This technical breakdown demonstrates how image acquisition, AI‑driven license‑plate recognition, cloud storage, and rule‑based logic combine to enforce the "Open Four Stop Four" traffic restriction.