I'd like a piece of software "view" an image, find objects on that image, then count the number of objects in predefined "zones".
In short I'd like to compare a photo with one of several templates:
Template = High contrast silhouette (there would be several templates of different shapes. Generally 2 color, but sometimes 3 colors)
Image = camera phone image of silhouette with objects added.
(camera phone = iPhone / Android or point and shoot. Not 5 year old nokia blurry photo)
Steps as I see them:
1. Match the image silhouette with on of several stored templates
2. Image may be taken from a slightly different angle than the template silhouette. The photo of the silhouette may need to be overlaid on top of the silhouette template to get the "zones" of where the objects are.
3. Find objects on an image. (objects would be blobs or differences between the photo and the template)
4. Count the number of objects in different "zones" of the image. (Zones would be defined by the silhouette template)
In the end I'd like a server side application where images could be uploaded and the results could be spit back out. (the results being the number of objects in each zone)
I'm new to image processing so I may be over or under-complicating the task.
We could separate this out into two projects.
The first would be a proof of concept and wouldn't require as much automation.
The second step would be to build an automated server application that would accept photos and spit out the results.
The bidding for this project will be for the proof of concept only. Automation is not necessary to complete this task
However I'd like the person who does this project to have the capability to build the server application that automates this task.
Hi, I have experience with image processing and worked in similar projects, extracting and matching image features. Sample images of both photos and templates are appreciated.
Hi, please send some test images for evaluation. And what's your server OS? Linux or windows? Please see my previous projects demo attached in PM. Thanks, Derick.