SYSTEMS, METHODS AND COMPUTER PROGRAM PRODUCTS FOR IDENTIFYING OBJECTS IN VIDEO DATA
First Claim
Patent Images
1. A method of identifying an object in a feed of images, the method comprising the steps of:
- a. analyzing a plurality of sequential frames from a feed of frames;
b. selecting in each frame, one or more unit areas of interest;
c. applying to each unit area a plurality of detectors, wherein each detector provides a value expressing one aspect of a target object;
d. assembling data in a sample vector, wherein the data comprises a plurality of components and each component is the individual value outcome of a given detector applied on a given unit area from a given frame; and
e. evaluating the performance of the sample vector by counting in the vicinity of the sample vector, all vectors known with the target object to form a first count, and counting all vectors known with or without the target object, as acquired during a training phase to form a second count, wherein the ratio of the first count to the second count estimates the probability of the presence of the target object.
0 Assignments
0 Petitions
Accused Products
Abstract
Image based operating systems and methods are provided that identify objects in video data and then take appropriate action in a wide variety of environments. In some embodiments, the image based operating systems and methods allow a user to activate other devices and systems by making a gesture.
-
Citations
25 Claims
-
1. A method of identifying an object in a feed of images, the method comprising the steps of:
-
a. analyzing a plurality of sequential frames from a feed of frames; b. selecting in each frame, one or more unit areas of interest; c. applying to each unit area a plurality of detectors, wherein each detector provides a value expressing one aspect of a target object; d. assembling data in a sample vector, wherein the data comprises a plurality of components and each component is the individual value outcome of a given detector applied on a given unit area from a given frame; and e. evaluating the performance of the sample vector by counting in the vicinity of the sample vector, all vectors known with the target object to form a first count, and counting all vectors known with or without the target object, as acquired during a training phase to form a second count, wherein the ratio of the first count to the second count estimates the probability of the presence of the target object. - View Dependent Claims (2, 3, 4, 5, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22)
-
-
6. A method of identifying an object in a feed of images, the method comprising the steps of:
-
receiving video data, wherein said video data corresponds to a plurality of frames of sequential images from said feed of images; identifying a set of candidate objects from within said feed of images, wherein said identifying comprises applying a recognition operation to data corresponding to a subset of images from each of at least two of said plurality of frames to generate a individual recognition score for a plurality of candidate objects for one or more of the subset of images in which a candidate object is present; aggregating the recognition scores of a plurality of candidate objects that appear in a plurality of subsets of images across said at least to of said plurality of frames to generate an aggregate score for each object, wherein said aggregate score indicates a relative likelihood of presence of an object in the video data; obtaining a subset of candidate objects based on said aggregate scores; and comparing a change in aspect of at least one of said subset of candidate objects across said at least two of said plurality of frames to an established scenario of transformation of aspect, thereby determining an object probability score for each of one or more candidate objects from said subset, wherein an object probability score reflects the probability of the presence of an object in said video data. - View Dependent Claims (7, 8, 9, 10, 11)
-
-
23. A computer readable non-transitory storage medium storing instructions that, when executed by a computer, causes the computer:
-
to receive video data, wherein said video data corresponds to a plurality of frames of sequential images; to apply a recognition operation to data corresponding to a subset of images from at least two of said plurality of frames to generate a recognition score for a plurality of candidate objects for one or more of the subset of images in which a candidate object is present; to aggregate the recognition scores of a plurality of candidate objects that appear in a plurality of subsets of images across said at least two of said plurality of frames to generate an aggregate score for each object, wherein said aggregate score indicates a relative likelihood of presence of an object in the video data; to obtain a subset of candidate objects based on said aggregate scores; and to compare a change in aspect of at least one of said subset of candidate objects across said at least to of said plurality of frames to an established scenario of transformation of aspect, thereby determining an object probability score for each of one or more candidate objects from said subset, wherein an object probability score reflects the probability of the presence of an object in said video data. - View Dependent Claims (24, 25)
-
Specification