All entries for Saturday 25 November 2006
November 25, 2006
I have been looking at the implementation and starting to get some ideas on what to write.
I will have these completed for early tomorrow and then get the code together in the evening.
Here is a flow chart that I have made to go with the implementation and is the idealised version that if we had time the program would follow.
Here’s a more concise version of the colour, histogram and blob and contour based tracking parts that Nick wrote. You may need the Microsoft Equation 3.0 component of office installed if it isn’t already, it’s part of the office tools.
Also an update on the bibliography with 2 more sources and numbered, incase anyone wants to reference them in the report.
The word count for the colour + histogram + contour is around 530 words, and the word limit is 1500 so there’s a fair bit to fill up. I may be changing these 3 depending on whether we need to fill up the word count, so they’re here just to give you an idea of what I’ve been doing.Things that need to be finished are: -
- Introduction to the problem
Combining the 3 methods Our implementation (Ant’s doing this)
- Appendix (code needs more commenting)
- Putting together all the parts and printing (do we need colour prints?)
- Creating the CD with all the code, pictures, videos and presentation (should we include the robot resource cd?)
- Signing the contribution sheet
Deadline is Monday, so all of this needs to be done tomorrow!
Here is my contribution. Keep everyone posted on progress!
Description of Further possible Developments to the Project:
Four main ideas were identified to expand and/or enhance the project. The following ideas were decided upon based on what would improve the use of the system for the user, by making it easier to use or improving its accuracy or functionality.
The first suggestion is to improve the speed of the algorithm used, so that the system will be capable of tracking a person who is travelling at a much higher speed than currently catered for. This will involve speeding up the running time of the algorithm in question, possibly by removing more of the redundant data being processed, or by finding a better method of comparing models.
The second suggestion is to introduce functionality capable of tracking multiple people within the same image. This would mean that the two people would have to wear different coloured clothing, or some method should be implemented to distinguish between them.
The third suggestion for improvement would be to explore the problems in tracking a person wearing a multi-coloured shirt further. The aim of this would be to tweak the histogram algorithm already in place to remove redundant background information picked up as part of the blob, thus making the method more accurate.
The fourth suggestion is to effectively remove the initial start-up phase which is currently required to create the blob mask. To achieve this, the system would need to be capable of dynamic person recognition, whereby it could recognise and create a mask of a person anywhere within an image, and continue to track that person based on the model created from this mask.