System and method for visual searching of objects using lines
AssigneeRecognition Robotics, Inc.
InventorSimon Melikian
Life of this patent
Estimating the term…
Abstract
Disclosed is method of visual search for objects that include straight lines. A two-step process is used, which includes detecting straight line segments in an image. The lines are generally characterized by their length, midpoint location, and orientation. Hypotheses that a particular straight line segment belongs to a known object are generated and tested. The set of hypotheses is constrained by spatial relationships in the known objects. The speed and robustness of the method and apparatus disclosed makes it immediately applicable to many computer vision applications.
Claims19
A method for matching a learned object with a target object, the method comprising using a computer to perform the steps of:
- providing at least one learned object and at least one target object, wherein the learned object and the target object include at least one line segment;
- selecting the at least one line segment from the at least one learned object;
- determining an amount of translation, rotation, and scaling needed to transform the selected at least one line segment of the learned object to have one or more lines substantially the same size as the at least one line segment in the target object;
- and
- determining if the learned object matches the target object based at least in part on the step of determining the amount of translation, rotation and scaling needed to transform the selected at least one line segment of the learned object to have one or more line segments substantially the same size as the at least one line segment in the target object.
A method for matching a learned object with a target object, the method comprising using a computer to perform the steps of:
- providing at least one learned object and at least one target object, wherein the learned object and the target object have a plurality of contour points, wherein contour points having a curvature below a certain threshold value are grouped together to form at least one line segment;
- extracting at least one line segment from the learned image;
- determining the amount of translation, rotation, and scaling needed to transform the at least one line segment of the learned object to have one or more line segments substantially the same size as at least one line segment in the target object;
- determining if the learned object matches the target object based at least in part on the step of determining the amount of translation, rotation and scaling needed to transform the at least one line segment of the learned object to have one or more line segments substantially the same size as at least one line segment in the target object.
A method for matching a learned object with a target object, the method comprising using a computer to perform the steps of:
- a) providing at least one learned object and at least one target object, wherein the learned object and the target object have a plurality of contour points, wherein contour points having a curvature below a certain threshold value are grouped together to form at least one line segment;
- b) extracting at least one line segment from the target image, wherein the extracted line segment corresponds to a longest line segment of the target image;
- c) extracting at least one line segment from the learned image, wherein the extracted line segment corresponds to a longest line segment of learned image;
- d) determining a transformation hypothesis that maps the learned image to the target image;
- e) selecting a next longest line segment from the learned image and the target image;
- f) determining if the learned object matches the target image based at least in part on the step of determining the transformation hypothesis that maps the learned image to the target image.
A program stored on a non-transitory computer readable medium, the program being suitable for use in matching a learned object with a target object, wherein
- when the program is loaded in memory of an associated computer and executed, causes extracting a line segment from a learned image and a target image, wherein the extracted line segments correspond to a longest line segment of the learned image and the target image;
- determining a transformation hypothesis that maps the learned image to the target image;
- selecting a next longest line segment from the learned image and the target image;
- and
- determining if the learned object matches the target image based at least in part on the step of determining a transformation hypothesis that maps the learned image to the target image.
A method for matching a learned object with a target object, the method comprising using a computer to perform the steps of:
- providing at least one learned icon in electronic form, wherein the learned icon is associated with a learned object, wherein the learned icon includes at least a learned icon length and an angle;
- providing a target object in electronic form;
- selecting a first learned icon;
- extracting a first target icon from the target object;
- determining the amount of translation, rotation, and scaling needed to transform the first learned icon to have one or more lines substantially the same size as the first target icon;
- and
- determining if the learned object matches the target object based at least in part on the step of determining the amount of translation, rotation and scaling needed to transform the first learned icon to have one or more lines substantially the same size as the first target icon.
A method for matching a learned object with a target object, the method comprising using a computer to perform the steps of:
- providing at least one learned object, wherein the at least one learned object has a plurality of contour points, wherein contour points having a curvature below a certain threshold value are grouped together to form a learned icon, wherein the learned icon has a first end point and a second end point;
- providing a target object in electronic form;
- selecting the learned icon;
- extracting a first target icon from the target object;
- determining the amount of translation, rotation, and scaling needed to transform the learned icon to have a size and shape that corresponds to the target icon;
- and
- determining if the learned object matches the target object based at least in part on determining the amount of translation, rotation, and scaling needed to transform the learned icon to have the size and shape that corresponds to the target icon.
Dependencies are inferred from the claim text.