Object/Human tracking in a video stream has been an active research area in the field of
Computer Vision and Image processing and have been explored by many researchers and today number of practical applications are in use for security and other purposes. The first task in the tracking problem is to identify the object of interest in the video. A very basic technique to accomplish this objective would be a supervised learning system that would be trained over a number of images in which the region of interest is identified the user and then in the next frames of the video the system will itself keep track of this object of interest. Another alternative technique would be to use the simple way of background subtraction i.e. the system is trained over the background image that remains the same over the period of time and then an intruder or object appears in the video stream and is then detected by the system and later is tracked throughout the video sequence
As this is a basic level Digital Image Processing course we will constrain to ourselves only to the background subtraction system which can then be extended with the use of Histogram Equalization. We will use the basic concepts from statistics e.g. mean, covariance etc. to study the relation among different pixels of the image and hence find sort of clustering or classification of these pixels on the measure of some distance.
The algorithm first would be trained over a number of images. Once the system is trained their nput to the system would be a gray scale or colored image and the output would be either in the form of a binary image separating the object of interest from the background or a colored image in which the object of interest would be outlined by some lines etc. to distinguish it from the background. The different major steps of the algorithm can be stated as,
1. Training over a number of images.
2. Input to the trained system i.e. an image.
3. Object Identification in the given frame using statistical measures of mean, covariance
and Mahalanobis distance. K means clustering or any standard and acceptable modeling method.
4. Output as a binary image separating object of interest from the background.
5. Saving of the tracked object throughout the video stream i.e. frame sequence.
I want this project to be done in 15 days in MATLAB. only code.
Décerné à :
6 freelance ont fait une offre moyenne de 187 $ pour ce travail
Dear Sir, Thanks for give me the opportunity bid on your projects. I can slove this work in matlap. I like to talk abt whole requiremt pl contact gtalk:matlapguru. Regards matlapguru