Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Traffic AI insights #2

Open
hossainalmahdi opened this issue Mar 29, 2024 · 0 comments
Open

Traffic AI insights #2

hossainalmahdi opened this issue Mar 29, 2024 · 0 comments
Labels
question Further information is requested

Comments

@hossainalmahdi
Copy link

Problems detected while implementing Traffic-AI

  1. Detecting traffic Collision from 3rd perspective view is tough because of camera angle. After detecting vehicle if we try to detect traffic collision based on the intersection of two bounding box; most of the time it will create false alarm as per the camera angle & overlapping two detected vehicle. Also, if there is any large vehicle eg: truck/ covered van it will create destruction. Other vehicle might not even detected by camera because of it.

1.1. However, as we have already studied few research papers/ published journals; we have seen that, the authors are trying to detect traffic collision from the car dashboard. mostly they're measuring the distance, predicting front vehicle next pixel position & based on that they are passing the signals. In my opinion, sensors are more powerful in the situation like this and all the vehicle company implemented auto breaking system/ emergency breaking system based on high quality sensors; eg: sonar sensor. Also I have a question that what if the collision is coming from the backside of a vehicle?

  1. Accurate speed detection is challenging for many cases. as we are calculating the speed via distance of a pixel or how fast the object is moving from one pixel to another one. based on the various road condition this calculation or the algorithm might not be universal for all the situation.

  2. Camera placement during traffic analysis also plays an important role. Because computer vision or the generated model does have some lacking. If we place the camera far away from the ROI, It's very common case that the model can't even detect the vehicle/object when it's too far in the frame. in human eyes it is also tough to track a vehicle where the size of the vehicle in video feed covers only some pixels.

  3. Traffic congestion part is kind of dynamic based on region. In Bangladesh if we consider the road condition it is so tough to detect actual traffic congestion only based on some algorithm. However, it can be implemented only in highways where traffic congestion is minimal and only happens where there is real traffic problems

@bhuiyanmobasshir94 bhuiyanmobasshir94 added the question Further information is requested label Mar 30, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
question Further information is requested
Projects
None yet
Development

No branches or pull requests

2 participants