search

mWater / mWater-Android-App

An Android App that snaps a picture of a petrifilm, counts the bacteria and posts the counts on a server for the mWater project
https://github.com/AndroidImageProcessing
GNU General Public License v3.0
24 stars 16 forks source link

Questions about using the app in the field: #34

Closed peristeri closed 12 years ago

peristeri commented 12 years ago

@rocketboy76 @cesine

I have a few questions that might help the usability of the client application:

  1. How often are samples taken from a water source?
  2. Can we label the petri gells with a bar code of sorts?
  3. What are the distances between water sources? The GPS have an accuracy of +- 30 meters
  4. Can the many devices capture images of multiple water source dishes? This link to the idea of id(ing) the dishes.
cesine commented 12 years ago

Here is my shot at the answers:

  1. maybe once, maybe many times.. why?
  2. its true, we can't reuse the petri gell, so yeah, they could have unique barcodes printed in them, if that is possible by the manufacturer? But it might make them too expensive, printers dont like printing unique things... and that would go against the point of having a low cost water kit. But yeah, that could be a cool addition for the fieldworker, then they would have nothing to write. Just scan the barcode with the Android to know which source it is. On the other hand, its often faster for people to just read what is written on the sample then to scan it with an Android to find out where it's from.
  3. true, true... On the other hand the GPS in my Android seems to be able to track which side of the street I'm biking on and when I swerve to avoid puddles. So, +/- 1 meter accuracy when used outdoors in suburban areas with lots of satellites. I looked up rural Bangladesh and they seemed to get it up to 5 meter accuracy when using "a network of 144 submeter-accurate ground control points (GCPs), so [they] could easily tie into those for [their] differential postprocessing." So yeah, getting acceptable GPS accuracy might require some extra work and some extra data http://findarticles.com/p/articles/mi_m0BPW/is_3_11/ai_n27557514/
  4. Is this the question: what if the android that took a picture of the source (and geo taged it), isn't the same android that takes a picture of the petri gel (and geo tagged it).. thats an interesting question. its a possibility that we should plan for in v2.

We have to demo what we have for rocketboy76 so he can get an idea of how we are associating sources and petri gells. @peristeri do you have time to make him a YouTube demo of the 3 user stories? If not now, I think we can wait until the holidays when we'll have more time.

Once he sees a demo then he will be able to give us better feedback on whether our user stories will make sense in the field.

rocketboy76 commented 12 years ago

@peristeri @cesine These are good questions. Here are my thoughts:

  1. Samples might be taken once per week or less often during normal operations. However, during intense periods such as startup of a new system or disease outbreaks, one could imagine taking a daily sample.
  2. I am not in favor of the bar code because it would add cost and complexity as Cesine points out. Also, it requires that all the components be purchased from a central supplier who would control the numbering scheme. We would prefer to keep the ecosystem for the supplies open. This helps particularly with launching the kit in a new country, since they can use local supply chains who are familiar with imports and customs procedures.
  3. Each water source should have a unique ID number or string. This is how we will identify samples from the same source taken at different time points. Geographic data is not sufficient, particularly in urban areas. Even in my work in villages in Bangladesh, it was very difficult to figure out if two geocoded sample locations were the same after the fact. For this reason we used an ID number that was printed on a metal tag that was then attached to the well, which makes database management very easy. Also, there will often be situations where we sample the inlet to a storage tank and the outlet, which would be essentially the same geographic location but very different water quality.
  4. In lieu of a physical ID tagging approach, which makes sense only for a limited project, we would need other information such as household ID, physical description, owner, etc. to ensure correct identification. The workers who would take the samples on a regular basis would come to know which ID number goes with which source. I'm sure this an issue that the ODK Collect developers have wrestled with, so hopefully we can continue the discussion with them and get some new insights.