means of transportation

Snow Tunnel Turtle King

Snow Mountain Tunnel real-time traffic condition signboard

How to read the Kanban board

Going south and going north are counted separately. Three numbers can determine whether to set off now and which road to cut before entering the tunnel.

The happy turtle king represents smooth traffic flow

speed

The current average driving speed in this direction. When the speed limit is close to 90, it is smooth. When it drops below 40, it usually means that the road ahead has started to be blocked.

Worried King Turtle, the representatives began to line up

queue length

Estimated array length (km). This number reflects the traffic jam earlier than the vehicle speed - the vehicle speed has not dropped yet, but the queue is already getting longer.

Turtle King wearing a helmet, representing suggested driveway

Suggested lanes

Based on the traffic flow on both sides, it is determined whether to go left or right. Lane changes are prohibited in the tunnel, so decide before entering the tunnel.

Basic information of Xueshan Tunnel

full length
12.94 km
speed limit
90 km/h (interval speed)
Opening hours
24 hours
driveway
Two lanes in one direction, no lane changes allowed
large car
Drive in the outside (right) lane
Smooth ride
about 9 minutes

FAQ

What is the speed limit in the Snow Mountain Tunnel? Is there a speed test?

The speed limit is 90 kilometers per hour and there are

Why can’t I change lanes in a snow tunnel?

The space between lanes in the tunnel is

Is it faster to take the left or right lane in a snow tunnel?

There is no fixed answer, it depends on the current flow of traffic. Large vehicles are required to drive in the outside (right) lane, so there are often large vehicles on the outside and the speed is lower; but in traffic jams, the inside may also be slower. The upper signboard will directly give suggestions based on real-time vehicle speed and traffic flow status.

What is the total length of the Snow Mountain Tunnel? Approximately how long will it take?

full length

When are traffic jams most likely to occur in snow tunnels?

Consecutive holidays and weekends

How are these numbers calculated?

Speed and lane recommendations come from computer vision analysis of highway CCTV images, not manual reporting or simply reposting official data. Below is the entire process.

Current status:看板數值為展示用範例資料;即時分析管線正在整合中,尚未上線。

The actual output screen of the analysis program: The CCTV image of the Snow Mountain Tunnel is superimposed with a red central dividing line and a yellow passing counting line. The two vehicles are marked by blue and green tracking boxes with vehicle numbers respectively. The real-time speed of the left and right lanes is displayed in the lower left corner.
  • 中央分隔線——左右車道由它分開
  • 通過計數線——車輛跨過就計一次
  • 追蹤框與車輛編號——同一台車跨影格的身分

Actual output of the analysis program (China 5 North 25K+335)

Background processing YOLOv10 detects vehicles Cars, motorcycles, buses and trucks frame by frame ByteTrack multiple target tracking Give each car a fixed number so that the displacement is meaningful Inverse perspective pixel conversion to meters Use vanishing point and lane width to restore true distance Median regression to calculate vehicle speed Immune to detection frame jitter and will not be pulled away by a single hop Separate statistics for left and right lanes left lane 90 km/h 30.4 vehicles/minute There are few vehicles on the way and the traffic flow is smooth. right lane 82 km/h 27.9 vehicles/minute Large vehicles drive on the outside and the speed is low It is recommended to take the left lane The vehicle volume will be judged first. If the vehicle speed difference is too large, the vehicle speed will override it. (Values are examples)
Why not just use the "pass rate" to decide? 因為 Flow = Density × Velocity,所以「通過率低」有兩種完全相反的意思: Open and smooth - there are few cars, so the passing rate is naturally low (want to leave) It’s so crowded that I can’t pass – there are too many cars but I can’t move (I want to avoid it) 單看通過率分不出來,所以改以車量與車速判斷。
How a CCTV frame becomes "left or right": The left side is the actual output of the analysis program, and the right side is the result of each of the four processing and two lanes.

Vehicle detection: YOLOv10

Frame the cars, motorcycles, buses and trucks in the picture frame by frame. The resolution of highway CCTV is only 352×240, and cars in the distance often have less than 20 pixels. Therefore, the detection threshold must be lowered, and subsequent geometric checks can be used to filter out misjudgments.

Multi-target tracking: ByteTrack

Merely framing a vehicle cannot calculate the speed - you must know where the "same car" is in the front and rear frames. The tracker is responsible for assigning a fixed ID to each vehicle so that the displacement is meaningful.

Inverse perspective: Convert pixels to meters

A single shot has no depth information. Assuming that the road surface is a plane, the displacement on the screen can be restored to the actual distance on the road surface by using the vanishing point where the lane lines intersect and the standard lane width of the national highway.

Lane Dividing: Center Divider

Taking the central double white solid line in the tunnel as the boundary, the vehicle falls on whichever side of the line belongs to which lane. The speed and volume of vehicles are counted in the two lanes respectively.

Vehicle speed estimation: median regression

Speed is the slope of time versus distance. The detection frame will jitter at low resolution, so take the median of the slopes of all sampling points instead of the least square - the median will not be pulled away by a single jump point.

Suggested lanes: priority given to traffic volume, speed vetoed

No pass rate judgment is required. Because flow rate is equal to density multiplied by speed, "low passing rate" may mean that it is empty and smooth, or it may be that it is too crowded to pass. Instead, it will be based on the average number of vehicles in the lane. If the speed difference is too large, it will be directly rejected by the vehicle speed.

The vehicle speed is an estimate of the monocular camera, which will be affected by calibration errors. It is for reference only before traveling and is not equivalent to the official speed measurement data.

Other common tools

The source of road condition data is traffic information published by the highway authorities. The values are system estimates and are for reference only before traveling. Please refer to on-site signs and traffic control.