I am working on a project implementing deep learning and computer vision to estimate the traffic density of any random given road segment/roundabout or intersection.
I am given a camera mounted on the drone, which will capture the traffic footage and I aim to extract vehicles and road scene (image segmentation) out of it in real-time to calculate the density.
The problem is that the original formula for calculating the traffic density is number of vehicles/unit length of road, while with the present method & resources, I could not measure the length but only the area that the vehicles and the road/ pavement cover. Is it possible for me to calculate the traffic density as area of vehicles/area of road? If not, can anyone suggest me a method measuring the length of the road by deep learning automatically on any road segment (straight/curve).
I have read many paper mentioning the approach of estimating the traffic flow rather than the density to get the traffic status. However, I found that it is impossible to infer from the traffic flow that whether the road is congested or not (flow = 0 could mean heavy traffic or no traffic at all). Also, some of them can measure the length of the road because they mount a static camera on the light post, which makes the background static and they can easily calculate the pixel length of the road manually. Alternatively, I am using the drone the get the traffic data autonomously at anywhere, so measuring the length of the road is impossible.
Asked
Active
Viewed 139 times
0
khiem nguyen trong
- 11
- 1
- 3
-
1I'm voting to close this question as off-topic because its not about programming. – Dr. Snoopy Jul 30 '18 at 16:07
-
I am not mentioning anything relating to programming. What I am asking is the approach for the problem. Sorry for any misunderstanding. – khiem nguyen trong Jul 30 '18 at 17:07
-
That's exactly why your question is off-topic, Stack Overflow is a QA site about programming. – Dr. Snoopy Jul 30 '18 at 23:52
-
I'm sorry for this inconvenience. I'm a newbie to this site. Can you suggest me any other page that I can post my question? – khiem nguyen trong Jul 31 '18 at 01:51
1 Answers
0
For your problem, I think semantic segmentation might work better. From what I understand you want to calculate the ratio of image containing road to that of vehicles. You can use semantic segmentation to find out the number of pixels belonging to each of the classes i.e. road and vehicles and then use the ratio of those pixels as a representative of the metric that you want.
