Dog Tracking

I recently spent some of my free time looking for an acceptable tool set to develop software to track Dexter. He is kind enough to use a litter box most of the time, and my goal is to use computer vision to give him positive (and maybe negative) feedback to help him reduce his error rate.

I spent some time looking at IM but eventually settled into openCV. It supports rudimentary firewire capture and includes a "highgui" toolset to help gui novices, such as myself, display results quickly. Installation required building the libraries and I wouldn't have been able to decipher the error messages had it not been for Surgey Ten's blog.

Now I'm working on identifying the dog object. Once it's accurately recognized, upon entering and exiting the litter box a positive reinforcement statement will be played.

Later on, and this I think is the trickier part, if the dog-object sub-divides then a negative reinforcement statement can be played. I think the tricky part has something to do with differentiating the head and rear of the object -- so as not to scold a sick puppy. But, I guess, many would argue that negative reinforcement directed at a Chihuahua is ineffective.