All Collections
Methodology & Requirements
Requirements for Traffic Camera Video Processing
Requirements for Traffic Camera Video Processing

How to make your video suitable for processing from fixed traffic cameras?

Pavel Severa avatar
Written by Pavel Severa
Updated over a week ago

GoodVision Video Insights provides greater than 95% accuracy of collected traffic data. Our processing supports several types of vehicles, motorcycles, bicycles and pedestrians – all can be on the scene together. All you have to do is to provide a quality video input. So how to make sure that your videos have sufficient quality to achieve the best results? There are certain aspects of the video files which affect the accuracy of video analytics. The recommended settings are described in these requirements.

Requirements table:


Recommended settings



Camera height

8-12 meters (min 5m, max 30m)

Object distance

Up to 70m

Object size

20x20 px


1920 x 1080 px (min. 320x240 px, max 4096 x 2160 px)

Frame rate

FPS 25 - 30 (min 10, max 240)


5000+ kbps (SDR),

10000+ kbps (HDR)

Shutter speed

1/48 s or 1/50 s (min 1/24 s, max 1/260 s)


Available, to be specified in the processing options

Camera/Hardware producers


Supported Video


H.264 recommended (for other supported codecs see the version for download)

Video length

unlimited, min. duration 1 minute

Video file size

< 50 GB per single video file

200 GB per multiupload processing

(please contact us in case of a bigger video file)

Criterion 1: Camera View

1. Camera height

Camera placement is naturally an important aspect in video analytics. The camera must not be placed too low so that the objects in the front do not cover the objects behind. The vehicles further from the camera would be hidden and thus the data not collected. On the other hand, it cannot be placed too high either as the detection ability is limited in high heights, especially of small objects like pedestrians, cyclists and motorcyclists.  

We recommend camera placement ideally between 8-12 metres above the ground with 5-30m as the minimal and maximal height boundaries, depending on how broad space you want to cover, which traffic attendants you want to monitor (e.g. pedestrians) and of course, how is it possible to fix the camera on the spot.

2. Distance from monitored objects

Distance from the monitored objects is closely related to camera height. For reliable detection and tracking of the traffic objects, it is recommended that the dimensions of vehicles are at least 4-5% of the scene size. Also make sure that the objects are not covering a substantial portion (more than 30%) of the scene.

Camera tilt doesn’t affect the detection ability of GoodVision Video Insights. It is trained and versatile for various conditions. You can tilt the camera from almost horizontal up to a straight-down bird-view if needed. Anyway, when choosing the camera view angle, always keep in mind that the monitored distance should not exceed 70 metres for all vehicle classes to be tracked accurately. In bigger distances, small objects like pedestrians, bikes and motorbikes might not be reliable (this is affected also by other factors  -  lens, blur, etc.).

3. Obstacles

Obstacles are tricky. Smaller objects like thin poles, standard traffic signs, or wires do not affect the detection ability but larger obstacles do (trees, statues, buildings, bridges). When a vehicle is covered by an obstacle on its path through the scene, the system often considers it as a new object after it appears again. Thus the vehicle trajectory gets splitted which can affect some metrics during the traffic analysis. We have developed a Smart Corridor feature that aims to reconnect the splitted trajectories on the movement streams to improve the traffic volume counts in these cases.

Still, always try to place the camera in a way that the main traffic streams of your interest are fully visible and uninterrupted. If it is impossible for some reason to avoid any obstacles in the scene view, you’ll want to take it into consideration when analysing the scene (e.g. when placing lines and zones around the obstacles to define the traffic movements or when creating analytical widgets).

Criterion 2: Video parameters

1. Resolution

Camera resolution heavily influences the quality of the video and the quality of the computer vision processing. The more image data (pixels) you provide to the system, the better it recognizes the objects in it. General rule is that the vehicles should be at least 20 x 20 px large on the scene, ideally around 50 px.

GoodVision Video Insights is trained to deliver results from resolutions as low as 320 x 240 px but resolutions 1280px x 720px, 1920px x 1080px (FULL HD), or higher will guarantee the best accuracy. Change of resolution within a single video file or a data source is not supported - i.e. all frames of video files uploaded to one single data source need to share the same resolution.

Lower resolutions go hand in hand with low-quality optics and low bitrate, causing the object contours are blurry or do not resemble the object from the real world. Set the resolution which displays the object's contours clearly.

Examples of blurry traffic objects on a low resolution scene

2. Frame rate (FPS)

Frame rate of the video defines the fluency of the object’s motion in it and affects the tracking ability of the video analytic system. Tracking means preserving the identity of the object between frames on which it was detected. It has a crucial impact on having solid object trajectories i.e. for the origin-destination counting of traffic. Low FPS causes tracking problems, especially on crowded scenes and with fast moving objects, which are literally “jumping” from place to place over the scene.

Ideal FPS for GoodVision Video Insights is between 25 to 30 frames per second. The bigger the better, however FPS bigger than 60 does not have any visible impact on tracking quality. We consider FPS 10 as a minimum (but not recommended) rate for video processing.

3. Shutter speed

Camera shutter speed affects the clarity of the moving object’s contours, especially in the low-light conditions and close to the camera. Some cameras switch to longer shutter speeds in order to keep the same overall brightness of the scene during night. Try to avoid this and rather preserve the clarity of the objects. GoodVision Video Insights is trained to recognize objects in the dark, but if the objects are too smudgy and completely lack the contours, it is super-hard to detect them.


Make sure the camera doesn’t switch between multiple configurations and is stable during the while time of recording. Turn off night mode, avoid changing settings in the middle of the recording and turn off the option for switching to backup streams if possible.

Criterion 3: Environmental Conditions

1. Scene lighting

Scene lighting plays an important role in video analytics. However, GoodVision Video Insights is trained to recognize objects in the dark as well as during the day. The only condition is that the objects must be at least a bit illuminated to be visible in the image with the naked eye.

2. Weather and other conditions

There are several problematic areas regarding the weather conditions that should be avoided if possible.

  • Raindrops / wet lense - distort the image or cause light reflections. Acts like a physical obstacle.

  • Snow - similar to rain in case of heavy snow and snowflakes on the camera lense.

  • Frontal light (usually from the sun under certain angle) causing flare or reflection  -  covers the image, decreases the object clarity.

  • Dirty lenses, or with scratches or smudges  -  causes blurry image, removes object contours

  •  Barrel distortion  -  deforms and bends the objects

And if all conditions are met…

If the conditions above are fulfilled the GoodVision Video Insights system will reward you with close to 100% traffic data collection accuracy.

Did this answer your question?