The constraint here might be availability of usable high resolution 
imagery. I believe Google prohibits use of its imagery and the imagery that 
these tutorials use are not available in India. Some people use night time 
lights overlayed on street networks as a proxy for vehicle density but I 
think GPS based google data would be more reliable.

Regards
Anupam

On Wednesday, June 18, 2025 at 12:18:08 PM UTC+5:30 Sarath Guttikunda wrote:

> In our group, we frequently perform extensive numerical analysis, 
> particularly to understand emission intensities, both spatially and 
> temporally. The spatial understanding of emissions is a significant 
> component of our work, as it is crucial for accurately placing emissions 
> before they are modeled and concentration maps are generated.
>
> This process is technically known as the "gridding of emissions." For 
> example, if we know there are one thousand trucks operating in a city, each 
> traveling one hundred kilometers a day, we can multiply these figures by an 
> emission factor for a specific pollutant to determine the total emission 
> intensity of trucks moving within the city's airshed. The question then 
> becomes: how do we distribute these emissions into various grids for a 
> city? We typically work with one-square-kilometer grids on average, and you 
> can see some examples below.
> https://urbanemissions.info/india-air-quality/india-ncap-city-airsheds/
>
> One of the proxies we use for trucks is highways. The assumption is that 
> most trucks will travel on highways and spend the majority of their time 
> there. Therefore, we assign a higher weight to the grids that intersect 
> with highways. We also incorporate other layers of information with 
> additional weights. For instance, industrial hubs, commercial hubs, malls, 
> and markets are places where these vehicles are likely to go and spend some 
> time. This methodical approach generates various weighting functions, and 
> once we have the emission intensities, it produces a gridded emission file. 
> So far this method of madness works and we have a good understanding of how 
> the layers are behaving with some plus minus. We have an example tool to 
> play with this method -- https://urbanemissions.info/tools/
>
> We aim to improve this process. One of the layers we introduced in the 
> past was speed information from the Google Maps API. We can download speed 
> data, which also indicates congestion times. We utilized this as another 
> proxy to understand where and for how long vehicles spend time, and 
> accordingly, assign weights. See example image for Mumbai here - 
> https://urbanemissions.info/india-apna/mumbai-india/
>
> A new approach we want to explore, given some recently available 
> information (and algorithms), is vehicle density. This would again be a 
> static input. For example, if you take a satellite image and apply an 
> algorithm, you could determine how many vehicles are visible within each 
> grid. Because this is a static image for a specific time, we cannot use it 
> as a layer for all-purpose gridding. However, it would serve as an 
> additional layer of information that accurately reflects what is happening 
> on the ground. It could also be used to extract information about official 
> and unofficial parking lots where vehicles spend a significant amount of 
> time on a given day. This would allow us to extract valuable insights.
>
> There are many online examples of this being done using geostationary 
> images in Europe and the United States and most of them require an image 
> and rest seem to work (take it with a pinch of salt -- non-it-person 
> speaking).
>
> https://www.linkedin.com/posts/giswqs_geoai-geospatial-ai-activity-7309216621997281280-46yC/
>
> https://up42.com/marketplace/analytics/detection-cars-oi
> A commercial portal -- but seems to do exactly what we want at a price
>
> So, the question to the group today is this: If there is a grid file, let 
> us say for Bangalore -- has anyone done anything similar to create a 
> vehicle density map, regardless of the vehicle type? or have any ideas on 
> how to approach this for Indian cities?
>
> Please keep in mind that the ultimate goal is not to identify individual 
> vehicles or count vehicles from traffic cameras. The focus is on a static 
> image: if we have one, can we, or has anyone, worked on creating a vehicle 
> density map from it?
>
> Any sights into making an example and hopefully scaling it up is 
> appreciated.
>
> Looking forward to the follow ups.
>
> With best wishes,
> Sarath
>
> --
> *Dr. Sarath Guttikunda*
>
> *https://urbanemissions.info <https://urbanemissions.info/about-ueinfo/>*
>

-- 
Datameet is a community of Data Science enthusiasts in India. Know more about 
us by visiting http://datameet.org
--- 
You received this message because you are subscribed to the Google Groups 
"datameet" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/d/msgid/datameet/c0afb013-c9b2-4f13-8b92-570b3d79b916n%40googlegroups.com.

Reply via email to