I have a project of video monitoring, which uses difference algorithm to locate moving objects inside video, draw tracking line for each moving objects.it can find out any moving things inside video, like human, animal, cars, snow flakes (depending on resolution of camera).
the problem is the difference algorithm can not recognize the charactristics of each object, so it needs physical assistant, such as "signs" on the objects, if you want to follow specific objects among all moving things.
For example, you need put coat on dogs with numbers, 1,2,3...etc.on the coat, if you want to follow each dogs from real time video monitoring.
we need neural network to work out and this is my thinking:
1, to make a video monitoring into intelligent system, which means it can recognize and make decision alone;
the raw data is the video (30 frames per seconds ) which will store in the PC (video captured by cable-connected dome camera) and difference algorithm already marks all moving objects.
2, start from dogs, the recognition should has 2 levels:
2-1 recognize dogs from any other moving objects;
2-2 recognize dog from dog
3, make the intelligence into a PC software (DLL? or plugin?)
which will tell the total monitoring software who's who
If you think my idea is workable and you have interest, I can provide difference algorithm video monitoring software and video (all moving objects marked) to you for developing purpose,