# US 8,787,678 B2 · System and method for visual searching of objects using lines > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 8,787,678 B2 - Document ID: US8787678B2 - Status: Granted - Kind: B2 - Application number: 12/941,706 - Filed: 2010-11-08 - Published: 2011-03-03 - Granted: 2014-07-22 - Earliest claimed: 2006-11-07 - Assignees: Reognition Robotics - Applicants: Simon Melikian - Inventors: Simon Melikian - Web page: https://patentgrep.com/patent/US8787678B2 ## CPC - G06V 20/582: {of traffic signs} - G06V 10/243: {by compensating for image skew or non-uniform image deformations} - G06V 10/255: {Detecting or recognising potential candidate objects based on visual cues, e.g. shapes} - G06V 10/46: Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features (colour feature extraction G06V10/56) - G06V 10/752: {Contour matching} - G06V 20/588: {Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road} - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## 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. ## Claims ### Claim 1 A method for training a computer vision system to recognize a reference object provided in an electronic form, the method comprising using a computer to perform the steps of: extracting line segment information from one or more contour points in the reference object by grouping contour points, wherein the line segment information is selected based on the one or more contour points having a curvature at or near zero as a line; storing the line segment information in a computer readable form; processing the line segment information to extract at least one image patch from the reference object, wherein the at least one image patch comprises a sub-image of the reference object that is scale and rotation invariant and includes at least one linear feature formed by the one or more contour points, the at least image patch including at least one end point associated with the linear feature and the at least image patch having a size determined by a distance between the least one end and an end contour point; and storing the at least one image patch in a computer readable form, wherein the at least one image patch is associated with the reference object. ### Claim 2 (depends on 1) The method of claim 1, wherein the line segment information includes at least one from the group consisting of: an endpoint, a midpoint, a line angle or a line length for the one or more line segments. ### Claim 3 (depends on 2) The method of claim 2, wherein the line segment information includes at least one contour point. ### Claim 4 (depends on 3) The method of claim 3, wherein the line segment information includes a vector from the midpoint for each of the one or more line segments to a reference point. ### Claim 5 (depends on 1) The method of claim 1, wherein the computer readable form is a database. ### Claim 6 (depends on 1) The method of claim 1, wherein the reference object is provided from an electronic computer aided design file. ### Claim 7 (depends on 1) The method of claim 1, wherein the at least one image patch is extracted for all linear features of the object having a segment length above a length threshold value. ### Claim 8 (depends on 1) The method of claim 1 further including fitting an analytic line to the line segment using linear regression for representation of the at least one image patch. ### Claim 9 (depends on 8) The method of claim 8, wherein the analytic line utilizes the end point associated with the linear feature and the end contour point. ### Claim 10 (depends on 1) The method of claim 1 further including storing information related to the at least one image patch in a database of image patches. ### Claim 11 (depends on 10) The method of claim 10, wherein the at least one image patch is stored for all linear features of the object having a segment length above a length threshold value. ### Claim 12 (depends on 10) The method of claim 10, wherein the information includes a length associated with at least one image patch. ### Claim 13 (depends on 10) The method of claim 10, wherein the information includes a scale associated with at least one image patch. ### Claim 14 (depends on 10) The method of claim 10, wherein the information includes an image patch angle, wherein the image patch angle is the relation between the image patch and a reference point in the object. ### Claim 15 (depends on 1) The method according to claim 1, wherein the line segment information includes information corresponding to an endpoint of the line segment, and wherein the image patch is centered at the endpoint. ### Claim 16 (depends on 1) The method according to claim 1, wherein the line segment information includes information corresponding to length of the line segment, and wherein a size of the image patch is proportional to a the length of the line segment. ### Claim 17 (depends on 1) The method according to claim 1, wherein the line segment information includes information corresponding to an angle of the line segment, and wherein an orientation of the image patch corresponds to the angle of the line segment.