I am trying to write an algorithm for surveillance camera where it sends only changed frames(any deviations to base frame) to control centre than just simply always sending live stream. I am implementing this to save bandwidth allocated to wireless surveillance cameras.Here I am using Color histogram and Daubechies wavelet transformation. Can any one suggest any new ideas in doing this.
Not sure why you are using a histogram and a wavelet. In general, you can subtract any coming frame from the median of the last several un-changed frames, and compare to a threshold and sum the values above the threshold. Then you send only frames with a result above the threshold.
actually i am not only comparing for variation, i want to classify whether deviation from base image is by means of any object.currently i am considering two types of objects humans and vehicles.if there is any persons/vehicles entered in surveillance unit then only notification will be raised.for others like any birds or animals it will not generate any notifications.
The task you described in the last post actually consists of two parts: first, you need to detect if there are any new object in your frame, second you need to classify your object to understand whether this is human, vehicle or something else. For the first part alone any decent background subtraction method would fit, yoг can try something from here: https://github.com/andrewssobral/bgslibrary
The object recognition part is much more sophisticated, there is no universal recipe. Anyway, you will need some example data, extract features from detected objects and classify them using machine learning techniques. For standard tasks, like human body detection, you may try Haar Cascade based detector http://docs.opencv.org/modules/objdetect/doc/cascade_classification.html
You may find pre-trained cascades in OpenCV distribution, also some are available on the Internet, like here http://alereimondo.no-ip.org/OpenCV/34
Hello Dhanunjaya,
We have developed a video-surveillance system where we implement some of the methods that you mention. We only work with human detection, but you probably find our work interesting. It is recently published in the prestigious journal "Expert Systems With Applications (ESWA)". I attach you the paper if you want to check it out.
Good luck with your research!
Article Expert Video-Surveillance System for Real-Time Detection of ...
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication