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.
