System and method for image analysis using a chaincode
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G06K-009/48
G06K-009/30
G06K-009/36
G06K-009/00
출원번호
US-0193544
(2002-07-10)
발명자
/ 주소
Clark,Walter D.
Hoffman,Richard L.
출원인 / 주소
Northrop Grumman Corporation
대리인 / 주소
Fulbright &
인용정보
피인용 횟수 :
6인용 특허 :
35
초록▼
The present invention includes a method for analyzing an image to screen for target candidates wherein an edge-follower detects straight segments within a plurality of longer edges and detects also from the longer edges, contours that go around areas that are the right size. The invention also conte
The present invention includes a method for analyzing an image to screen for target candidates wherein an edge-follower detects straight segments within a plurality of longer edges and detects also from the longer edges, contours that go around areas that are the right size. The invention also contemplates starting the edge-follower from seeds on the image and terminating the edge-following process if the edge meanders too much to ever become an acceptable contour. In accordance with another embodiment, the straight segments can be further screened by measuring the gradient direction and then noting the deviation from the mean of those direction values. The highly straight edges thus found, in confluence with each other, or near a correct size contour define likely candidates.
대표청구항▼
What is claimed is: 1. A method for analyzing an image, comprising: selecting an initial location within an image to initiate a chaincode; initiating the chaincode at the initial location, the chaincode being operable to follow a path defined, at least in part, by a plurality of adjacent pixels hav
What is claimed is: 1. A method for analyzing an image, comprising: selecting an initial location within an image to initiate a chaincode; initiating the chaincode at the initial location, the chaincode being operable to follow a path defined, at least in part, by a plurality of adjacent pixels having approximately equal changes in brightness therebetween; measuring direction changes of the path; and calculating meander values corresponding to successive positions of the chaincode as a chaincode follower progresses along the path, wherein the meander values are related to the proportion of traversed pixels with respect to a straight-line distance traveled by the chaincode from an initial point. 2. The method of claim 1, further comprising storing a starting point and a history of the direction changes of the path. 3. The method of claim 1, further comprising: terminating the chaincode if one of the meander values exceeds a predetermined threshold meander value. 4. The method of claim 1, wherein the path includes at least a first limit of extent which is approximately parallel to a reference axis, wherein a length of the first limit of extent is determined by a first point at which the chaincode follower is no longer extending in a direction of the first limit of extent. 5. The method of claim 4, wherein: the path includes at least a second limit of extent which extends approximately perpendicular to the reference axis; a length of the second limit of extent is determined by a second point at which the chaincode follower is no longer extending in a direction of the first limit of extent; and the first and second limits of extent at least partially defining a contour. 6. The method of claim 1, further comprising: identifying portions of the chaincode that include contiguous points that do not deviate by more than one least significant bit of a chaincode range; and recording straight segments for further analysis of straightness. 7. The method of claim 5, wherein the contour is an edge with at least first and second limits of extent, and further comprising: calculating an approximate area of the contour defined, at least in part, by the first and second limits of extent, using x, y locations of the first and second limits of extent. 8. The method of claim 4, further comprising identifying generally straight edges which occur within a region defined by the contour as being a location of a target candidate. 9. The method of claim 1, wherein selecting the location to initiate the chaincode comprises: selecting a first seed point within the image; analyzing a first linear segment which extends from the first seed point, the first linear segment including a plurality of pixels of the image; detecting changes in brightness between successive ones of the plurality of pixels of the image; identifying first and second adjacent pixels, of the plurality of pixels, having a change in brightness therebetween which exceeds a predetermined threshold change in brightness; and wherein the location is selected proximate the first and second pixels. 10. The method of claim 9, wherein selecting the location to initiate the chaincode further comprises: identifying two of the successive pixels having the greatest change in brightness therebetween; and wherein the first and second adjacent pixels comprise the two of the successive pixels. 11. The method of claim 9, wherein selecting the location to initiate the chaincode further comprises: selecting a grid of seed points within the image, the grid of seed points including the first seed point; analyzing a plurality of linear segments along the image, each linear segment initiating at, and extending away from, a respective one of the grid of seed points; wherein the plurality of linear segments includes the first linear segment; for each of the plurality of linear segments, identifying a pair of adjacent pixels having the greatest change in brightness; wherein the first and second adjacent pixels comprise one of the pairs of adjacent pixels; and initiating chaincodes proximate each of the pairs of adjacent pixels. 12. The method of claim 1, wherein the path of the chaincode at least partially defines a contour and further comprising: measuring an approximate area of the contour; comparing the area with a range of acceptable area values; and performing a second level of evaluation of the contour, if the area is within the range of acceptable area values. 13. The method of claim 1, further comprising terminating the chaincode if the chaincode demonstrates singularity such that the chaincode returns to an identical pixel, in one step. 14. The method of claim 1, further comprising measuring a plurality of edge directions along the path, wherein the edge directions are determined by a ratio of adjacent gray values. 15. The method of claim 14, further comprising calculating, for an edge segment, a deviation from a mean of the plurality of edge directions to obtain a measure of a degree of straightness of the edge segment. 16. The method of claim 15, further comprising: detecting a plurality of generally straight edges of the path, clustered in a region of the image; calculating a centroid of a contour defined, at least in part, by the plurality of generally straight edges; and identifying the centroid as a target candidate. 17. The method of claim 15, further comprising: detecting at least one generally straight edge along the path; determining if the at least one generally straight edge is located within a contour of interest; and identifying the contour as a target candidate. 18. The method of claim 1, wherein the path at least partially defines a contour, and further comprising: identifying the contour as a target candidate if the path includes a first limit of extent which extends generally parallel to a reference axis, and which is smaller than a predetermined maximum size, and at least a second limit of extent which extends generally perpendicular to the reference axis. 19. A method for analyzing an image, comprising: initiating chaincodes at a plurality of initial points of an image; for each chaincode, monitoring a path of the chaincode; terminating each of the plurality of chaincodes that meanders excessively; and identifying portions of the chaincode that include contiguous points that do not deviate by more than one least significant bit of the chaincode range. 20. The method of claim 19, further comprising: for each chaincode, storing the initial points and directional changes until the chaincode is terminated; and for each chaincode that meets certain predefined target candidate criteria, passing the initial points and directional changes along to a second level of evaluation. 21. A method for analyzing an image, comprising: initiating chaincodes at a plurality of edge points of an image; for each chaincode, monitoring a path of the chaincode; identifying portions of the chaincode that include contiguous points that do not deviate by more than one least significant bit of the chaincode range; and terminating any of the plurality of chaincodes having a straight-line distance that exceeds a predetermined maximum length. 22. The method of claim 21, further comprising: for each chaincode, storing a starting point and a history of directional changes until the chaincode is terminated; and for each chaincode that meets certain predefined target candidate criteria, passing the starting point and history of directional changes along to a second level of evaluation. 23. The method of claim 21, further comprising: monitoring a plurality of extents associated with each chaincode; wherein the plurality of extents of each chaincode at least partially define an associated contour of each chaincode; for each chaincode, continuously calculating an approximate area of the associated contour as the chaincode extends along the path; and terminating any of the plurality of chaincodes having an associated contour that exceeds a predetermined maximum area. 24. The method of claim 21, further comprising recording the portions for further analysis of straightness. 25. The method of claim 24, further comprising: identifying generally straight edges which occur within a region defined by an appropriate size contour as being a location of a target candidate. 26. The method of claim 21, further comprising measuring a plurality of edge directions along the path, wherein the directions are determined by a ratio of adjacent gray values. 27. The method of claim 26, further comprising calculating, for an edge segment, a deviation from a mean of the plurality of edge directions to obtain a measure of degree of straightness of the edge segment.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (35)
Stone David H. (West Hartford CT) Pintsov Leon A. (Manchester CT) D\Amato Donald P. (Avon CT), Alphanumeric handprint recognition.
Grasmueller Hans (Mammendorf DEX) Beutel Wilhelm (Munich DEX) Hohberger Karl (Feldkirchen-Westerham DEX) Koellensperger Paul (Munich DEX) Reischl Anna (Munich DEX) Rummel Peter (Festenbach DEX), Arrangement for the segmentation of lines.
Neff Michael G. ; Johnson Ted L. ; Wright Steven A. ; Ratkowski ; Jr. Victor E., Image processing method and apparatus for correlating a test image with a template.
Neff Michael G. ; Johnson Ted L. ; Wright Steven A. ; Ratkowski ; Jr. Victor E., Image processing method and apparatus for correlating a test image with a template.
Pearson William A. (Orlando FL) Benton Richard W. (Altamonte Springs FL) Lofgren John D. (Orlando FL), Methods and apparatus for centroid based object segmentation in object recognition-type image processing system.
Pearson William A. ; Benton Richard W. ; Lofgren John D., Methods and apparatus for centroid based object segmentation in object recognition-type image processing system.
Peurach Thomas M. ; Whalen Peggy ; Haanpaa Douglas ; Whalen Mike ; Jacobus Charles J., Pose determination and tracking by matching 3D objects to a 2D sensor.
Eppler Walter G., System and method for automatically determining the position of landmarks in digitized images derived from a satellite-based imaging system.
Hsu Shin-Yi (2312 Hemlock La. Vestal NY 13850), System for texture-based automatic detection of man-made objects in representations of sensed natural environmental scen.
Jenkins Gary Kim (Arlington TX) Evans Bruno Jack (Euless TX) Williams ; Jr. David Collis (Burleson TX) Bornowski Arthur Steven (Garland TX), Visual recognition system for LADAR sensors.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.