Labeling component parts of objects and detecting component properties in imaging data
Techniques related to labeling component parts and detecting component properties in imaging data are discussed. Such techniques may include generating a feature vector including invariant features associated with an area of interest within an image of an object such as an image of a hand and providing a component label such as a hand part label for the area of interest based on an application of a machine learning classifier to the feature vector.
1. A method for labeling areas of an input image comprising:
determining an area of interest within the input image;
providing a normalized grid of cells over the area of interest of the input image;
determining a plurality of properties corresponding to the normalized grid of cells, the plurality of properties including multiple cell properties for each cell of the normalized grid of cells, the multiple cell properties comprising a dominant orientation of gradients within each cell, an average magnitude of gradients within each cell, and a depth difference between a mean depth of each cell and a mean depth of an origin cell of the grid of cells;
generating a feature vector comprising a plurality of invariant features associated with the area of interest, the invariant features comprising a concatenation of the plurality of properties corresponding to the normalized grid of cells; and
providing a component label for the area of interest based on an application of a machine learning classifier to the feature vector.
2. The method of claim 1 , wherein the invariant features comprise scale invariant, rotation invariant, and position invariant features.
3. The method of claim 1 , wherein the area of interest comprises a representation of at least a portion of a suspected finger within an image of a hand, the representation comprising at least one of a finger axis, a finger base point, or a finger tip point, and wherein the invariant features comprise at least one of a normalized finger length, a normalized finger width, a normalized distance from a finger base point to a center of a hand, a palm radius, a number of detected fingers in the hand image, a number of fingers pointing in the same direction as the finger, a number of fingers pointing in the opposite direction of the finger, or a number of fingers perpendicular to the finger.
4. The method of claim 1 , wherein the area of interest comprises a representation of at least a portion of a suspected finger within an image of a hand, the representation comprising at least a finger axis, a finger base point, and a finger tip point, and wherein the invariant features comprise a normalized finger length, a normalized finger width, a normalized distance from a finger base point to a center of a hand, a palm radius, a number of detected fingers in the hand image, a number of fingers pointing in the same direction as the finger, a number of fingers pointing in the opposite direction of the finger, and a number of fingers perpendicular to the finger.
5. The method of claim 1 , wherein the input image comprises at least one of a 2D image, a 3D image, a depth image, or an infrared image.
6. The method of claim 1 , wherein the area of interest comprises a representation of a suspected finger within an image of a hand, a portion of a suspected finger within the image of the hand, a suspected body part within an image of a body, or a suspected face part within an image of a face.
7. The method of claim 1 , wherein the machine learning classifier comprises at least one of a K-nearest neighbor classifier, a classification and regression tree classifier, a decision tree learning classifier, a random forest learning classifier, an adaptive boosting classifier, a cascading classifier, a gradient boosting tree classifier, a support vector machine, a linear classifier, a logistic regression classifier, a naive Bayes classifier, a generalized likelihood ratio test classifier, a radial base function classifier, a multilayer perceptron classifier, a neural network classifier, a Bayesian network classifier, or a combination classifier.
8. A system for labeling areas of an input image comprising:
a memory configured to store image data; and
a processor coupled to the memory, the processor to determine an area of interest within the input image, to provide a normalized grid of cells over the area of interest of the input image, to determine a plurality of properties corresponding to the normalized grid of cells, the plurality of properties including multiple cell properties for each cell of the normalized grid of cells, the multiple cell properties comprising a dominant orientation of gradients within each cell, an average magnitude of gradients within each cell, and a depth difference between a mean depth of each cell and a mean depth of an origin cell of the grid of cells, to generate a feature vector comprising a plurality of invariant features associated with the area of interest, the invariant features comprising a concatenation of the plurality of properties corresponding to the normalized grid of cells, and to provide a component label for the area of interest based on an application of a machine learning classifier to the feature vector.
9. The system of claim 8 , wherein the invariant features comprise scale invariant, rotation invariant, and position invariant features.
10. The system of claim 8 , wherein the area of interest comprises a representation of at least a portion of a suspected finger within an image of a hand, the representation comprising at least one of a finger axis, a finger base point, or a finger tip point, and wherein the invariant features comprise at least one of a normalized finger length, a normalized finger width, a normalized distance from a finger base point to a center of a hand, a palm radius, a number of detected fingers in the hand image, a number of fingers pointing in the same direction as the finger, a number of fingers pointing in the opposite direction of the finger, or a number of fingers perpendicular to the finger.
11. The system of claim 8 , wherein the area of interest comprises a representation of at least a portion of a suspected finger within an image of a hand, the representation comprising at least a finger axis, a finger base point, and a finger tip point, and wherein the invariant features comprise a normalized finger length, a normalized finger width, a normalized distance from a finger base point to a center of a hand, a palm radius, a number of detected fingers in the hand image, a number of fingers pointing in the same direction as the finger, a number of fingers pointing in the opposite direction of the finger, and a number of fingers perpendicular to the finger.
12. The system of claim 8 , wherein the input image comprises at least one of a 2D image, a 3D image, a depth image, or an infrared image.
13. The system of claim 8 , wherein the area of interest comprises a representation of a suspected finger within an image of a hand, a portion of a suspected finger within the image of the hand, a suspected body part within an image of a body, or a suspected face part within an image of a face.
14. The system of claim 8 , wherein the machine learning classifier comprises at least one of a K-nearest neighbor classifier, a classification and regression tree classifier, a decision tree learning classifier, a random forest learning classifier, an adaptive boosting classifier, a cascading classifier, a gradient boosting tree classifier, a support vector machine, a linear classifier, a logistic regression classifier, a naive Bayes classifier, a generalized likelihood ratio test classifier, a radial base function classifier, a multilayer perceptron classifier, a neural network classifier, a Bayesian network classifier, or a combination classifier.
15. At least one non-transitory machine readable medium comprising a plurality of instructions that, in response to being executed on a device, cause the device to label areas of an input image by:
determining an area of interest within the input image;
providing a normalized grid of cells over the area of interest of the input image;
determining a plurality of properties corresponding to the normalized grid of cells, the plurality of properties including multiple cell properties for each cell of the normalized grid of cells, the multiple cell properties comprising a dominant orientation of gradients within each cell, an average magnitude of gradients within each cell, and a depth difference between a mean depth of each cell and a mean depth of an origin cell of the grid of cells;
generating a feature vector comprising a plurality of invariant features associated with the area of interest, the invariant features comprising a concatenation of the plurality of properties corresponding to the normalized grid of cells; and
providing a component label for the area of interest based on an application of a machine learning classifier to the feature vector.
16. The machine readable medium of claim 15 , wherein the invariant features comprise scale invariant, rotation invariant, and position invariant features.
17. The machine readable medium of claim 15 , wherein the area of interest comprises a representation of at least a portion of a suspected finger within an image of a hand, the representation comprising at least one of a finger axis, a finger base point, or a finger tip point, and wherein the invariant features comprise at least one of a normalized finger length, a normalized finger width, a normalized distance from a finger base point to a center of a hand, a palm radius, a number of detected fingers in the hand image, a number of fingers pointing in the same direction as the finger, a number of fingers pointing in the opposite direction of the finger, or a number of fingers perpendicular to the finger.
18. The machine readable medium of claim 15 , wherein the area of interest comprises a representation of at least a portion of a suspected finger within an image of a hand, the representation comprising at least a finger axis a finger base point, and a finger tip point, and wherein the invariant features comprise a normalized finger length, a normalized finger width, a normalized distance from a finger base point to a center of a hand, a palm radius, a number of detected fingers in the hand image, a number of fingers pointing in the same direction as the finger, a number of fingers pointing in the opposite direction of the finger, and a number of fingers perpendicular to the finger.
19. The machine readable medium of claim 18 , wherein the input image comprises at least one of a 2D image, a 3D image, a depth image, or an infrared image.
20. The machine readable medium of claim 15 , wherein the area of interest comprises a representation of a suspected finger within an image of a hand, a portion of a suspected finger within the image of the hand, a suspected body part within an image of a body, or a suspected face part within an image of a face.