You can use SURF features from the image object and check each frame of the video for those features, if you have enough features you have found your object. This method also works if the object is turned or made smaller/bigger to a certain degree.
You can also use a object detector like a cascade object detector to find the object you want in a video or a live stream. This method is very fast, but only works if the object you have teached is not turned. ( e. g. if you teach the detector to find the side of a car it will never detect the front, even if its the same car)
Best regards
