WO2003105087A1 - Detection de bords a l'aide de la transformation de hough - Google Patents

Detection de bords a l'aide de la transformation de hough Download PDF

Info

Publication number
WO2003105087A1
WO2003105087A1 PCT/US2002/018337 US0218337W WO03105087A1 WO 2003105087 A1 WO2003105087 A1 WO 2003105087A1 US 0218337 W US0218337 W US 0218337W WO 03105087 A1 WO03105087 A1 WO 03105087A1
Authority
WO
WIPO (PCT)
Prior art keywords
row
image
cells
cell
edge
Prior art date
Application number
PCT/US2002/018337
Other languages
English (en)
Inventor
Daniel R. Paquette
Original Assignee
Lockeed Martin Corporation
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Lockeed Martin Corporation filed Critical Lockeed Martin Corporation
Priority to AU2002315022A priority Critical patent/AU2002315022A1/en
Priority to PCT/US2002/018337 priority patent/WO2003105087A1/fr
Publication of WO2003105087A1 publication Critical patent/WO2003105087A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/12—Edge-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/48—Extraction of image or video features by mapping characteristic values of the pattern into a parameter space, e.g. Hough transformation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G06T2207/10008—Still image; Photographic image from scanner, fax or copier
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20048—Transform domain processing
    • G06T2207/20061—Hough transform
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30176—Document

Definitions

  • the present invention relates to a system and method for identifying edges of an object in an image via analysis of a Hough transformation of the image.
  • OCR optical character recognition
  • Another application in which it may be desirable to determine the location of objects in an image relates to the automated processing (e.g., sorting, routing, postage checking, etc.) of parcels of mail.
  • an image of a mail parcel is generated and analyzed in order to determine information, such as the destination address, return address, and indicia (i.e., postage mark, stamp, etc.) on the parcel. This information can then be used to determine how to process the mail parcel.
  • mail parcels may have a variety of shapes and sizes.
  • the area scanned in order to generate the image of the parcel must thus be large enough to accommodate these varying shapes and sizes.
  • the parcel may comprise only a portion of the image.
  • the orientation of the parcel in the image may also vary. Thus, it is desirable to distinguish the portion of the image including the parcel from other areas in the image.
  • the present invention is directed to a system and method for identifying edges of an object in an image.
  • the invention relates to a method for identifying edges of an object in an image by analyzing an accumulator array produced via a Hough transformation of the image.
  • the array includes a plurality of rows each including a plurality of cells.
  • the method includes the step of identifying a first cell in a first row of the array associated with a first edge of the object.
  • the method also includes the step of searching the first row for a second cell in the array associated with a second edge of the object, the second edge being parallel to the first edge.
  • the method also includes the step of searching a second row of the array shifted ninety degrees from the first row for a third cell in the array associated with a third edge of the object, the third edge being perpendicular to the first edge.
  • the method further includes the step of searching the second row for a fourth cell in the array associated with a fourth edge in the image, the fourth edge being parallel to the third edge.
  • the invention relates to a computer product for identifying edges of an object in an image by analyzing an accumulator array produced via a Hough transformation of the image, the array including a plurality of rows each including a plurality of cells.
  • the computer product includes a portion for identifying a first cell in a first row of the array associated with a first edge of the object.
  • the computer product also includes a portion for searching said first row for a second cell in the array associated with a second edge of the object, the second edge being parallel to the first edge.
  • the computer product also includes a portion for searching a second row of the array shifted ninety degrees from the first row for a third cell in the array associated with a third edge of the object, the third edge being perpendicular to the first line.
  • the computer product further includes a portion for searching the second row for a fourth cell in the array associated with a fourth edge in the image, the fourth edge being parallel to the third edge.
  • the invention relates to a method for identifying edges of an object in a graphical image by analyzing a Hough image of the graphical image.
  • the method includes the step of searching the Hough image from a first edge of the Hough image towards the center of the Hough image for a first cell in the Hough image indicative of a first potential edge line in the graphical image.
  • the method also includes the step of searching the Hough image from a second edge of the Hough image opposite the first edge towards the center of the Hough image for a second cell in the Hough image indicative of a second potential edge line in the graphical image extending parallel to the first potential edge line.
  • the method also includes the step of searching the Hough image from the first edge towards the center of the Hough image for a third cell in the Hough image indicative of a third potential edge line in the graphical image extending perpendicular to the first potential edge line.
  • the method further includes the step of searching the Hough image from the second edge towards the center of the Hough image for a fourth cell in the Hough image indicative of a fourth potential edge line in the graphical image extending parallel to the third potential edge line.
  • Fig. 1 is a functional block diagram illustrating a system for identifying a edges of an object in an image using a Hough transformation, according to an illustrated embodiment of the present invention
  • Fig. 2 is a schematic representation of an image generated and evaluated by the system of Fig. 1;
  • Fig. 3 is a functional block diagram illustrating a process performed in a portion of the system of Fig. 1;
  • Figs. 4A and 4B are graphs illustrating a Hough transformation process performed during a step in the process of Fig. 3;
  • Fig. 5 is a functional block diagram illustrating a process performed during a step in the process of Fig. 3;
  • Figs. 6 and 7 are a functional block diagrams illustrating a process performed during steps in the process of Fig. 3.
  • the field of the present invention relates to identifying lines in an image via analysis of a Hough transformation of the image.
  • the present invention relates to a system and method for detecting edges of an object in an image using a Hough transformation of the image.
  • the object is a parcel of mail.
  • Fig. 1 illustrates a system 10 in which the method of the present invention may be implemented to identify a mail parcel 12 in an image.
  • the system 10 includes an image processor 14 comprising an image capturing portion 16 and an image processing portion 18.
  • the image capturing portion 16 comprises means, such as a digital camera, for capturing an image, indicated at 17.
  • the mail parcel 12 is passed by the image capturing portion 16 by known means (not shown) , such as a conveyor or other mail handling equipment.
  • the image 17, which includes the mail parcel 12 is captured by the image capturing portion 16 as the parcel passes by.
  • the image 17 is preferably a grayscale image, such as an eight-bit, 256 color digital grayscale image.
  • the image 17 is provided to the image processing portion 18.
  • the image processing portion 18 performs a known binarization process on the image 17 to convert the image to a binary form.
  • the image processor 14 thus produces a binary image, indicated at 20, representative of the image 17.
  • the binary image 20 comprises a digital array of black and white pixels representative of the image 17 from which the binary image was generated.
  • the image processing portion 18 may also perform pre-processing on the image 17 before converting the image to the binary image 20. This may be done in order to reduce the amount of data in the image and thus increase the speed at which the image can be processed.
  • This pre-processing may include sub- sampling of the image 17 to reduce the size of the image. In the sub-sampling process, a group of pixels in the grayscale image 17 would be analyzed to determine the average grayscale intensity of the pixels in the group. ' This group of pixels would then be converted to a single pixel in a reduced image, the single pixel having the average grayscale intensity.
  • a 4096 x 4096 pixel image may have 16 x 16 pixel areas sub-sampled to provide a reduced image of
  • the sub-sampling should be of a degree such that the reduced image still contains the relevant information.
  • the sub-sampled image would then be binarized to produce the binary image 20.
  • the binary image 20 is provided to a parcel image identification portion 32 of the system 10.
  • the parcel image identification portion 32 is operative to identify the region in the image that includes the mail parcel 12 by analyzing the data contained in the binary images 20.
  • Fig. 2 illustrates an example of the binary image 20 produced by the image processor 14. As shown in Fig. 2, the mail parcel 12 occupies a portion of the binary image 20.
  • the mail parcel 12 illustrated in Fig. 2 is that of a typical package but could, however, be any type of parcel, such as an envelope, periodical, flyer, etc.
  • the parcel 12 may have an angled orientation with respect to the axes (edges) of the image 20.
  • the parcel 12 includes regions of interest that may include information relevant to sorting and/or routing the parcel. Such information may include the destination address 22, return address 24, and indicia 26, such as postage stamps, meter marks, bar codes, or any other type of marking used to identify postage affixed to the parcel 12.
  • the parcel image identification portion 32 (Fig. 1) is operative to identify a parcel image 34, that is, the portion of the binary image 20 that includes the parcel 12.
  • the parcel image identification portion 32 may be embodied as a computer program compilable to provide a computer product (i.e., program) executable to determine the portion of the binary image 20 that includes the parcel 12. It will thus be appreciated that the system 10 may comprise any computer means suited to provide a platform upon which to execute the computer product of the parcel image identification portion 32.
  • the parcel image 34 is identified by identifying the edges 28 of the parcel 12 in the binary image 20.
  • the edges 28 of the parcel 12 are identified in the binary image 20 via analysis of a Hough transformation of the binary image.
  • the Hough transformation is a method known in the art for identifying lines (straight lines, curves, circles, etc.) in an image.
  • the Hough transformation is used to detect straight lines in the binary image 20 that are representative of the generally rectangular edges 28 of the parcel 12.
  • the Hough transformation method relies on a parametric representation of points in the binary image 20. This representation is based on the principal that a straight line in a Cartesian (X-Y) coordinate plane may be represented by a single point in a Hough
  • a line 50 is represented in a Cartesian coordinate plane 52.
  • any straight line in a Cartesian coordinate system can be parameterized in terms of an
  • the angle ⁇ is the angle with respect to the X-axis of a line extending through the origin and
  • the distance p is the distance along this normal line from the line 50 to the origin of the coordinate plane 52.
  • p x(cos ⁇ ) + y(sin ⁇ ) (1)
  • x and y are the Cartesian (x,y) coordinates of any point on the line.
  • a point on the straight line 50 in the Cartesian coordinate plane 52 may be parameterized (i.e., transformed) in accordance with Equation (1) to a sinusoidal curve in a parametric plane.
  • This plane referred to herein as a Hough parameter plane, is illustrated at 60 in Fig. 4B.
  • the Hough parameter plane 60 plots the distance p
  • a point in the Cartesian coordinate plane 52 corresponds to a sinusoidal curve in the Hough parameter plane 60.
  • a point in the Hough parameter plane 60 corresponds to a straight line in the Cartesian coordinate plane 52. From these properties, it will thus be appreciated that points lying on the same straight line in the Cartesian coordinate plane 52 correspond to sinusoidal curves through a common point in the Hough parameter plane 60. Further, points lying on the same curve in the Hough parameter plane 60 correspond to straight lines through the same point in the Cartesian coordinate plane 52.
  • the parcel image identification portion 32 includes a portion 40 for performing a Hough transformation on the binary image 20.
  • the Hough transformation portion 40 is operative to generate a Hough image in the form of an accumulator array 42 in a manner known in the art. Basically, the pixels in the binary image 20 are mapped to a Cartesian coordinate system.
  • the Hough transformation portion 40 steps sequentially through each pixel in the binary image 20 and extracts the coordinates (x,y) for each black pixel in the binary image 20.
  • the Hough transformation portion 40 constructs a sinusoidal curve in the Hough image for each black pixel in accordance with Equation (1) .
  • the Hough image is constructed as an accumulator array 42 comprising a plurality of rows each including a plurality of cells.
  • the accumulator array 42 is a two-dimensional data array wherein one dimension
  • the rows in the array 42 correspond to the distance p.
  • the rows in the array 42 correspond to the
  • ⁇ and distance p may be represented in the accumulator array 42 at any desired resolution.
  • the ⁇ and distance p may be represented in the accumulator array 42 at any desired resolution.
  • resolution of the angle ⁇ may be 1 degree and the
  • each cell in the accumulator array 42 is initialized with an integer value of zero. As the sinusoidal curves are plotted in the accumulator array 42, the values of the cells are incremented by one for each curve that passes through the cell. As stated above, points lying on the same straight line in a Cartesian coordinate plane (i.e., the binary image 20) correspond to sinusoidal curves through a common point in the Hough parameter plane (i.e., the accumulator array 42) . Thus, the cells in the accumulator array 42 with high values relative to the other cells may be associated with intersecting sinusoidal curves indicative of straight lines in the binary image 20. Each cell in the accumulator array 42 is thus
  • the angle ⁇ and distance p associated with these cells can therefore be used to help identify straight lines, and thus the edges 28 (Fig. 2), of the parcel 12 in the binary image 20.
  • the magnitude of the cell may be used to approximate the length of the straight line represented by the cell.
  • the binary image 20 may include straight lines other than those indicative of the edges 28 of the parcel 12.
  • the data in the accumulator array 42 is analyzed in a manner wherein all of the cells in the accumulator array 42 are analyzed to find every possible straight line in the image. All of these possible straight lines are then analyzed to determine which of the lines are indicative of the edges 28 of the parcel 12 in the image 20.
  • an accumulator array analysis portion 44 of the parcel image identification portion 32 helps avoid such unnecessary analysis.
  • the accumulator array analysis portion 44 is operative to analyze the data in the accumulator array 42 in a systematic and efficient manner to identify the parcel image 34 by finding those straight lines in the binary image 20 that are most likely to be indicative of the edges 28 of the parcel 12.
  • the accumulator array analysis portion 44 is illustrated in Fig. 5. In this description of the present invention, the use of the word "step” is used to describe functions performed during various processes performed by the accumulator array analysis portion 44.
  • cells in the accumulator array 42 are identified as potential edge lines. The potential edge lines are determined by systematically stepping through the cells in the accumulator array 42 to determine which cells have a magnitude above a predetermined threshold indicative of a line in the binary image 20.
  • the predetermined threshold magnitude required to indicate a potential edge line may be determined based upon a variety of factors, such as the resolution and/or size of the graphical image. For example, it will be appreciated that the maximum magnitude of a cell indicative of a single line would be about equal to the length (i.e., number of pixels) in the longest line possible in the binary image 20. Thus, the predetermined magnitude required to indicate a potential edge line may be determined as a percentage of this maximum magnitude.
  • the process performed at step 70 is illustrated in
  • the potential edge line determination process of step 70 identifies the potential edge lines by analyzing rows in the accumulator array 42. Each row of the accumulator array 42 corresponds to a particular angle ⁇ and includes a cell for each increment of the
  • a first row in the accumulator array 42 is selected for analysis.
  • the first row may be selected in any order, preferably in a bottom-to-top or top-to-bottom manner.
  • the cells in the selected row are examined by stepping through the cells from opposite ends of the row inward toward the center of the row. This is done in an alternating manner in which the outermost cell at a first end of the row is examined, then the outermost cell at a second end opposite the first end is examined, then the second outermost cell at the first end is examined, and so on. This is done so that the Hough image is analyzed from the edges of the image towards the center of the image
  • the process searches the selected row for cells having a magnitude above the predetermined threshold. Once a predetermined number of cells having a magnitude above the threshold are identified, the process skips to the next row regardless of whether the entire row has been searched. This is done to help prevent the unnecessary examination of cells for lines that lie within the edges 28 of the parcel 12. For example, at step 102, the process may search the selected row from opposite ends of the row until two cells having a magnitude above the threshold are identified. Also, once a cell having the requisite magnitude is identified while stepping in from one end of the row, the search may continue from the opposite end of the row only. This may further help to identify only the outermost lines in the image.
  • step 104 the parametric
  • step 106 a determination is made as to whether the selected row is the last row in the accumulator array 42. If the selected row is the last row, the process ends at step 108. If the selected row is not the last row, the process proceeds to step 110 where the next row in the accumulator array 42 is selected. The process then proceeds back to step 102, where the selected row is searched for cells that may be identified as potential edge lines and the process continues as described above.
  • the process performed at step 70 is thus operative to analyze each row of the accumulator array 42 in a systematic and efficient manner to determine a set of potential edge lines in the accumulator array 42.
  • edges 28 of the parcel 12 in the binary image 20 may not appear as straight lines. This may be because the edges 28 are, in fact, not purely straight, as may be the case with padded envelope type parcels. Also, noise may be introduced during the image capturing and processing portions of the system 10. Thus, it may be desirable to introduce error compensation factors into the potential edge line determination portion 70 of the accumulator array analysis 44. This may be done, for example, by
  • the location of the cell indicative of the potential edge line in the accumulator array 42 may be determined at the geometric center of the region or by a center of mass type calculation for the region.
  • step 70 the process of the accumulator array analysis 44 proceeds to step 72, where a potential edge line is selected from the set.
  • the potential edge line may be selected in any manner, such as in the order in which the potential edge lines were identified at step 70. Once the potential edge line is selected, the process proceeds to step 74.
  • the accumulator array 42 is searched for a cell indicative of a parallel partner line for the selected potential edge line. It will be appreciated that the cells in any given row of the accumulator array 42, i.e., cells having the same angle
  • ⁇ represent parallel lines in the binary image 20 (Fig. 1).
  • the row in the accumulator array 42 that includes the potential edge line is searched for the parallel partner line.
  • This search is preferably done from the end of the row inwards toward the potential edge line in case there are multiple parallel partners to the potential edge line. For this same reason, this search may also be done from the end of the row opposite the end from which the selected potential edge line was found. This helps ensure that the parallel partner spaced farthest from the potential edge line is identified at step 74.
  • the parametric data (angle ⁇ , distance p, and magnitude) for the parallel partner line is recorded.
  • searching for the parallel partner line the row is searched for cells having a magnitude above a predetermined threshold indicative of a line in the image. This threshold may be similar to or identical to the threshold used to identify the potential edge lines at step 70. It will also be appreciated that, since the magnitude of any given cell is indicative of the pixel length of the line represented by that cell, the row may also be searched for cells having a magnitude equal to or within a predetermined margin of the magnitude of the potential edge line. If a parallel partner line is identified, the process proceeds to step 76 regardless of whether all of the cells in the row have been analyzed. Once the search for the parallel partner line is completed, the process proceeds to step 76 regardless of whether a parallel partner line is found.
  • the accumulator array 42 is searched for a cell indicative of a first line perpendicular to the line indicated by the selected potential edge line. It will be appreciated that the cells in row of the
  • the rows in the accumulator array 42 having an angle ⁇ offset or shifted by ninety degrees, or within some predetermined margin thereof, represent lines extending perpendicular to the line indicated by the potential edge line.
  • This threshold may be similar to or identical to the threshold used to identify the potential edge lines at step 70.
  • the search for the first perpendicular line is done from an end or from opposite ends of the row(s) to help ensure that the outermost perpendicular line is identified first.
  • the search may begin at the end of the row that coincides with the end from which the selected potential edge line. If found, the
  • step 80 the process proceeds to step 80 regardless of whether all of the cells in the row have been analyzed. Once the search for the first perpendicular line is completed, the process proceeds to step 80 regardless of whether a first perpendicular line is found.
  • the accumulator array 42 is searched for a cell indicative of a parallel partner line for the first perpendicular line.
  • the row in the accumulator array 42 that includes the first perpendicular line is searched for cells having a magnitude above a predetermined threshold indicative of a line in the image. This threshold may be similar to or identical to the threshold used to identify the potential edge lines at step 70. Since the parallel edges of the parcel should be about equal in length, the row could be searched for cells having a magnitude equal to or within a predetermined margin of the magnitude of the first perpendicular point.
  • the search for the parallel partner for the first perpendicular line is preferably done from the end of the row opposite the end from which the first perpendicular line was identified. This helps ensure that the parallel partner line spaced farthest from the first perpendicular line is identified. If found, the
  • step 82 the process proceeds to step 82 regardless of whether all of the cells in the row have been analyzed. Once the search for a parallel partner line for the first perpendicular line is completed, the process proceeds to step 82 regardless of whether a parallel partner line is found.
  • a line may not be identified.
  • a determination is made as to whether at least three lines, including the potential edge line, have been identified.
  • a determination is made as to whether a rectangle can be constructed using the lines identified at steps 72, 74, 76, and 80 as sides. If at least three lines have not been identified, a rectangle cannot be constructed and the process proceeds to step 90, where a determination is made as to whether there are any more potential edge lines. If there are more potential edge lines, the process proceeds to step 92, where the next potential edge line is selected. The process then proceeds to step 74, and the process continues as described above. If, at step 82, at least three lines are identified, the process proceeds to step 84, where a rectangle is constructed using the recorded data associated with the identified lines.
  • Fig. 7 The process performed by the rectangle construction step 84 is illustrated in Fig. 7.
  • the edge lines 120 identified at steps 72, 74, 76, and 80 are evaluated at step 122 (Fig. 7) to determine how many lines have been identified. If four lines are identified, the process proceeds to step 124, where a rectangle is constructed using the identified edge lines 120.
  • the rectangle is constructed in the Cartesian coordinate system of the binary image 20 (Fig. 1) using the angle
  • step 122 if three edge lines 120 are identified, the process proceeds to step 126, where three sides of a rectangle are constructed using the identified edge lines. At step 126, the three sides are constructed in the Cartesian coordinate system of
  • the edge lines 120 are constructed as having an infinite length and the corners of the rectangles are determined at the intersection of the edge lines.
  • the process then proceeds to step 130.
  • the fourth side of the rectangle is estimated based on the parametric data associated with the cells representative of the other three sides. This estimation is done by constructing a side parallel to the opposite side and extending between the free ends of the other two known sides. The position of the free ends is determined based on the length of the sides determined by the magnitude of the cells in the accumulator array 42 associated with the sides.
  • step 86 Figs. 5 and 7
  • the identified edge lines used to construct the rectangle at step 84 thus represent a potential solution set indicative of a rectangle in the image that represents the edges of the parcel in the image.
  • a confidence value is assigned to the rectangle (solution set) constructed at step 84.
  • the confidence value provides a means by which to indicate of a degree of certainty as to whether the rectangle constructed at step 84 is representative of the edges 28 of the parcel 12 (Fig. 1) in the binary image 20.
  • the confidence value may incorporate any suitable system or scale (e.g., numerical, alphabetical, alphanumeric, etc.) by which this degree of certainty can be evaluated.
  • the confidence value may be a numerical value that ranges from zero to one (0.0-1.0).
  • a confidence value of 1.0 indicates a high confidence that the rectangle is indicative of the edges of the parcel. Conversely, a confidence value of
  • the confidence value 86 is initialized to a high confidence value. The confidence value is then adjusted based on calculations using the parametric
  • step 74 If, at step 74, no parallel partner line for the selected edge line was found, subtract 0.20 from the confidence value. This reduces the confidence value in the event that only three sides of a rectangle are found.
  • Valc onf New New confidence value
  • Valc on oid Old confidence value
  • agp otE dgeLin e Magnitude of the potential edge line
  • Magp ara iieip artn er Magnitude of the parallel partner for the potential edge line.
  • ⁇ otEdgeiine the angle ⁇ associated with the potential edge line
  • step 80 If, at step 80, no perpendicular point was found, subtract 0.20 from the confidence value. This reduces the confidence value in the event that only three sides of a rectangle are found.
  • ⁇ FirsPerpLine the angle ⁇ associated with the first line perpendicular to the potential edge line. This reduces the confidence value by a factor that is
  • ⁇ perpPa nerLine the angle ⁇ associated with the parallel partner for the first perpendicular line. This reduces the confidence value by a factor that is
  • MagerpPartnerine Magnitude of the parallel partner for the first perpendicular line. This reduces the confidence value by a factor related to the difference in length (magnitude) between the first line perpendicular to the potential edge line (Mag F i rs t P erLi n e) and the parallel partner for the first » perpendicular line (Magp erp partnerLine) •
  • step 86 If confidence value is adjusted to a velue less than zero, set the confidence value equal to zero. This ensures that the confidence value associated with any given rectangle is not negative.
  • step 90 a determination is made as to whether there are any more potential edge lines. If there are more potential edge lines, the process proceeds to step 92, where the next potential edge line is selected and the process proceeds as described above. If there are no more remaining potential edge lines to evaluate, the process proceeds to step 94.
  • step 94 the rectangle with the highest assigned confidence value is selected as being indicative of the edges 28 (Fig. 1) of the parcel 12 in the binary image 20.
  • the parcel image 34 may thus be extracted from the binary image 20 and, thus, the image 17.
  • the parcel image 34 is then provided to a parcel image processing portion 36 of the system 10, where the information in the parcel image 34 is processed. This information may then be provided to parcel sorting and/or routing equipment 38 of the system 10.
  • a confidence value is assigned for each rectangle constructed at step 84 and the rectangle with the highest confidence value is selected as being representative of the parcel image 34.
  • the accumulator array analysis 44 could be adapted to evaluate the rectangles as each is assigned a confidence value.
  • a rectangle having a confidence value at or above a predetermined threshold may be selected as being representative of the parcel image 34, in which case further analysis of the accumulator array 42 may cease.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)

Abstract

Procédé permettant d'identifier des bords d'un objet dans une image par analyse d'un tableau accumulateur produit à l'aide d'une transformation de Hough de l'image. Ledit procédé consiste à identifier une première cellule dans une première rangée du tableau associée à un premier bord de l'objet. Une deuxième cellule du tableau associée à un deuxième bord de l'objet est ensuite recherchée dans la première rangée, le deuxième bord étant parallèle au premier bord. Une troisième cellule du tableau associée à un troisième bord de l'objet est recherchée dans une deuxième rangée décalée de quatre-vingt-dix degrés par rapport à la première rangée, le troisième bord étant perpendiculaire au premier bord. Une quatrième cellule du tableau associée à un quatrième bord dans l'image est recherchée dans la deuxième rangée, le quatrième bord étant parallèle au troisième bord.
PCT/US2002/018337 2002-06-10 2002-06-10 Detection de bords a l'aide de la transformation de hough WO2003105087A1 (fr)

Priority Applications (2)

Application Number Priority Date Filing Date Title
AU2002315022A AU2002315022A1 (en) 2002-06-10 2002-06-10 Edge detection using hough transformation
PCT/US2002/018337 WO2003105087A1 (fr) 2002-06-10 2002-06-10 Detection de bords a l'aide de la transformation de hough

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/US2002/018337 WO2003105087A1 (fr) 2002-06-10 2002-06-10 Detection de bords a l'aide de la transformation de hough

Publications (1)

Publication Number Publication Date
WO2003105087A1 true WO2003105087A1 (fr) 2003-12-18

Family

ID=29731321

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2002/018337 WO2003105087A1 (fr) 2002-06-10 2002-06-10 Detection de bords a l'aide de la transformation de hough

Country Status (2)

Country Link
AU (1) AU2002315022A1 (fr)
WO (1) WO2003105087A1 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017058252A1 (fr) * 2015-10-02 2017-04-06 Hewlett-Packard Development Company, L.P. Détection d'objets de document
CN110264488A (zh) * 2019-06-20 2019-09-20 合肥工业大学 一种二值图像边缘提取装置

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
LLADOS J ET AL: "A system to understand hand-drawn floor plans using subgraph isomorphism and Hough transform", MACHINE VISION AND APPLICATIONS, 1997, SPRINGER-VERLAG, USA, vol. 10, no. 3, pages 150 - 158, XP002230559, ISSN: 0932-8092 *
LOPEZ KRAHE J ET AL: "THE APPLICATION OF THE HOUGH TRANSFORM TO THE DETECTION OF ROADS", SIGNAL PROCESSING: THEORIES AND APPLICATIONS. GRENOBLE, SEPT. 5 - 8, 1988, PROCEEDINGS OF THE EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO), AMSTERDAM, NORTH HOLLAND, NL, vol. 3 CONF. 4, 5 September 1988 (1988-09-05), pages 1673 - 1676, XP000092552 *
YAN PING ZHOU ET AL: "Hough technique for bar charts detection and recognition in document images", PROCEEDINGS OF 7TH IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING, VANCOUVER, BC, CANADA, 10-13 SEPT. 2000, 10 September 2000 (2000-09-10) - 13 September 2000 (2000-09-13), pages 605 - 608, XP010530057 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017058252A1 (fr) * 2015-10-02 2017-04-06 Hewlett-Packard Development Company, L.P. Détection d'objets de document
US20180225508A1 (en) * 2015-10-02 2018-08-09 Hewlett-Packard Development Company, L.P. Detecting document objects
US10997407B2 (en) 2015-10-02 2021-05-04 Hewlett-Packard Development Company, L.P. Detecting document objects
CN110264488A (zh) * 2019-06-20 2019-09-20 合肥工业大学 一种二值图像边缘提取装置
CN110264488B (zh) * 2019-06-20 2021-03-16 合肥工业大学 一种二值图像边缘提取装置

Also Published As

Publication number Publication date
AU2002315022A1 (en) 2003-12-22

Similar Documents

Publication Publication Date Title
US7085432B2 (en) Edge detection using Hough transformation
US5642442A (en) Method for locating the position and orientation of a fiduciary mark
EP1234273B1 (fr) Localisateur d'etiquettes, a resolutions multiples
US8326037B1 (en) Methods and apparatus for locating an object in an image
US5048099A (en) Polygon-based method for automatic extraction of selected text in a digitized document
US6729544B2 (en) Fast barcode search
US6708884B1 (en) Method and apparatus for rapid and precision detection of omnidirectional postnet barcode location
EP0669593A2 (fr) Procédé de reconnaissance de code bidimensionnel
US6965703B1 (en) Optical scanner and software for correcting distorted images
EP0784827A1 (fr) Procede et appareil de determination de l'orientation angulaire precise de symboles de code a barres dans des images ccd bidimensionnelles
US5841905A (en) Business form image identification using projected profiles of graphical lines and text string lines
JPH0420226B2 (fr)
US6934413B2 (en) Segmentation of text lines in digitized images
RU2251738C2 (ru) Способ приведения в соответствие заполненной машиночитаемой формы и ее шаблона при наличии искажений (варианты)
US5438636A (en) Apparatus for simultaneously convolving multiple digital binary images using a single convolver with a binary mask to determine pixel densities
EP0651337A1 (fr) Procede et appareil de reconnaissance d'objets, et procede et appareil de traitement d'images
KR100498763B1 (ko) 바코드 고속 관심 영역 위치 판독 시스템 및 제어 방법
WO2003105087A1 (fr) Detection de bords a l'aide de la transformation de hough
US7058203B2 (en) Region of interest identification using region of adjacent pixels analysis
JP4518141B2 (ja) 画像照合方法及び画像照合装置並びに画像照合プログラム
US6556701B1 (en) Method of and apparatus for extracting dotted line, and storage medium thereof
US20040024716A1 (en) Mail sorting processes and systems
CN113837119A (zh) 一种基于灰度图像识别易混淆字符的方法及设备
EP0702320A1 (fr) Détection d'inclinaison
EP0601730A1 (fr) Procédé de traitement d'images pour la correction de l'inclinaison de caractères

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT TZ UA UG UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
122 Ep: pct application non-entry in european phase
NENP Non-entry into the national phase

Ref country code: JP

WWW Wipo information: withdrawn in national office

Country of ref document: JP