CN113033592A - Shape matching and object identification method based on slope difference distribution - Google Patents

Shape matching and object identification method based on slope difference distribution Download PDF

Info

Publication number
CN113033592A
CN113033592A CN201911360618.8A CN201911360618A CN113033592A CN 113033592 A CN113033592 A CN 113033592A CN 201911360618 A CN201911360618 A CN 201911360618A CN 113033592 A CN113033592 A CN 113033592A
Authority
CN
China
Prior art keywords
dimensional
slope difference
distribution
slope
points
Prior art date
Legal status (The legal status 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 status listed.)
Withdrawn
Application number
CN201911360618.8A
Other languages
Chinese (zh)
Inventor
王振洲
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong University of Technology
Original Assignee
Shandong University of Technology
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 Shandong University of Technology filed Critical Shandong University of Technology
Priority to CN201911360618.8A priority Critical patent/CN113033592A/en
Publication of CN113033592A publication Critical patent/CN113033592A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • 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
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a shape matching and object identification method based on gradient difference distribution. The method comprises the steps of obtaining one-dimensional distance distribution by calculating the distance from the center of an object to all points on the outline of the object, filtering the one-dimensional distance distribution in a frequency domain through discrete Fourier transform, then calculating slope difference distribution of the filtered one-dimensional distance distribution, solving valley positions and peak positions of the slope difference distribution by enabling a derivative of the slope difference distribution to be equal to zero, mapping the valley positions and the peak positions to the outline of the object to obtain two-dimensional slope difference characteristic points, modeling the shape of each type of object by utilizing the normalized two-dimensional slope difference characteristic points, and identifying the type of the detected binary object by calculating the minimum distance sum of the normalized two-dimensional slope difference characteristic points of the online detected binary object and the normalized two-dimensional slope difference characteristic points of the shape model of each type of object.

Description

Shape matching and object identification method based on slope difference distribution
Technical Field
The invention relates to a shape matching and recognition technology of a binary object, in particular to a method for obtaining one-dimensional distance distribution by calculating the distance from the center of the object to all points on the outline of the object, filtering the one-dimensional distance distribution in the frequency domain by discrete Fourier transform, calculating the slope difference distribution of the filtered one-dimensional distance distribution, solving the valley position and peak position of the slope difference distribution by making the derivative of the slope difference distribution equal to zero, mapping the valley position and peak position to the object outline to obtain two-dimensional slope difference characteristic points, modeling the shape of each type of object by utilizing the normalized two-dimensional slope difference characteristic points, and identifying the category of the detected binary object by calculating the minimum distance sum of the normalized two-dimensional slope difference characteristic point of the online detected binary object and the normalized two-dimensional slope difference characteristic point of the shape model of each type of object.
Background
The invention relates to a shape matching and recognition technology of a binary object. Shape recognition of binary objects is an important research content of pattern recognition, and many conventional binary object shape recognition technologies, such as a shape recognition method based on a fourier descriptor, a shape recognition method based on principal component analysis, a shape recognition method based on an invariance distance, and the like, are used. However, the conventional methods have low recognition accuracy, and the application and development of object target recognition are severely limited. The invention carries out shape matching and object identification based on slope difference distribution characteristic points, slope difference distribution is a new one-dimensional distribution characteristic point calculation method proposed by the applicant in recent years, the method is successfully applied to the field of threshold selection and clustering, subversive precision surpassing is achieved compared with the traditional method, please refer to the document Z.Z. Wang, "A new approach for segmentation and quantification of cells or nanoparticles," IEEE T Info rm, 12(3): 962 and 971 (2016). Z.Z. Wang, "Determining the calibration centers by slope difference distribution," IEEE Access, 5, 10995-. The invention expands the application of the slope difference distribution to the solution of the characteristic points of the two-dimensional object outline, successfully obtains a series of two-dimensional slope difference characteristic points by a method of solving the one-dimensional slope difference characteristic points on the one-dimensional distance distribution and then mapping the one-dimensional slope difference characteristic points to the two-dimensional object outline, and a large number of experiments show that the shape matching and object identification method based on the slope difference distribution also has the precision and the potential of subverting the traditional object identification method.
Disclosure of Invention
The invention aims to provide a shape matching and object recognition method based on slope difference distribution, aiming at the problems that the existing object shape recognition method has low precision and can not meet the requirements of some high-precision recognition applications.
In order to achieve the purpose of the invention, the invention is realized by adopting the following technical scheme:
the method comprises the steps of obtaining one-dimensional distance distribution by calculating the distance from the center of an object to all points on the outline of the object, filtering the one-dimensional distance distribution in a frequency domain through discrete Fourier transform, then calculating slope difference distribution of the filtered one-dimensional distance distribution, mapping one-dimensional slope difference characteristic points to the outline of the object by calculating slope difference distribution and slope difference characteristic points of the one-dimensional distance distribution and utilizing the one-to-one correspondence relationship between the one-dimensional distance distribution points and the outline points of the object to obtain two-dimensional slope difference characteristic points, modeling the shape of each type of object by utilizing the normalized two-dimensional slope difference characteristic points, and identifying the type of the detected binary object by calculating the minimum distance sum of the normalized two-dimensional slope difference characteristic points of the online detected binary object and the normalized two-dimensional slope difference characteristic points of the shape model of each type of the object.
Compared with the prior art, the invention has the following advantages:
the shape matching and object identification method based on the slope difference distribution can robustly calculate the slope difference characteristic points of the object shape, and carries out object identification by matching the normalized slope difference characteristic points, so that the identification accuracy is obviously higher than that of the existing object identification technology, and the efficiency is also improved.
Drawings
FIG. 1 is a flow chart of the operation of the present invention.
Detailed Description
The present invention will be described in detail below based on a work flow chart.
FIG. 1 is a flow chart of the present invention, which comprises inputting a binary image of an object, obtaining a center portion of the binary image by morphological erosion, calculating an object center according to a pixel mean of the center portion, obtaining a two-dimensional contour of the object by boundary extraction, obtaining a one-dimensional distance distribution by calculating a distance from the object center to each pixel on the contour of the object, calculating a slope difference distribution of the one-dimensional distance distribution, selecting a valley position and a peak position of the slope difference distribution as candidate features, forming a one-dimensional slope difference feature by selecting features that best meet conditions, mapping the one-dimensional slope difference feature to the contour of the two-dimensional object by dimension conversion to obtain a two-dimensional slope difference feature, normalizing the two-dimensional slope difference feature to obtain a scale-invariant feature, and identifying the object by shape matching.
The object center point calculation method is as follows:
step 1: calculate the originalStarting binary objectO b Area of (1) byS o Represents;
step 2: for binary objects by the following equationO b Performing iterative morphological etching:
Figure 861657DEST_PATH_IMAGE001
(1)
Figure 537751DEST_PATH_IMAGE002
(2)
wherein
Figure 138497DEST_PATH_IMAGE003
Is a spherical structural element with the radius of 3, and calculates the corroded binary objectO e Area of (1) byS e Represents;
and step 3: repeating step 2 until the area of the binary object is erodedS e Smaller than the original binary object areaS o Is/are as followsQOne-third:
Figure 132998DEST_PATH_IMAGE004
(3)
original binary objectO b By calculating the corroded binary objectO e The center of (c) is found:
Figure 742971DEST_PATH_IMAGE005
(4)
wherein (A) and (B)x i , y i ), i=1,2,…,MRepresenting a binary object to be erodediThe image coordinates of the individual pixels are,Mindicating that the eroded binary object contains the total number of pixels.
Original binary objectO b Outermost peripheral pixel ofExtracted as the object profileC j 2D , i=1,2,…,LThe points on the contour are represented asP(x j , y j ), j=1,2,…,L, LRepresenting the total number of points on the object contour. Then one-dimensional distance distributionD j 1D , i=1,2,…,LCalculated from the following formula:
Figure 506527DEST_PATH_IMAGE006
(5)。
one-dimensional distance distribution by the following methodD j 1D , i=1,2,…,LAnd (3) filtering:
step 1: one-dimensional distance distribution byD j 1D , i=1,2,…,LConversion to the frequency domain:
Figure 543754DEST_PATH_IMAGE007
(6)
step 2: by removingF(k) High-frequency components:
Figure 759971DEST_PATH_IMAGE008
(7)
whereinWIs based on the cut-off frequency of a discrete fourier transform low-pass filter;
and step 3: obtaining a filtered one-dimensional distance distribution byD j 1D’ , i=1,2,…,L：
Figure 908056DEST_PATH_IMAGE009
(8)。
One-dimensional distance distribution after filteringD j 1D’ , i=1,2,…,LThe slope difference distribution of (a) is obtained by the following method:
step 1: after filteringA certain point on the one-dimensional distance distribution (j, D j 1D’ ) Left selection ofNPoints (A)j, D j 1D’ ), x= j,j-1,…,j-N+1Fitting a straight line:
Figure 526119DEST_PATH_IMAGE010
(9)
whereina l Is the slope of the fitted line, and is calculated by the following formula:
Figure 999826DEST_PATH_IMAGE011
(10)
Figure 703339DEST_PATH_IMAGE012
(11)
Figure 389536DEST_PATH_IMAGE013
(12)
step 2: at a point on the filtered one-dimensional distance distribution (j, D j 1D’ ) Right selection ofNPoints (A)j, D j 1D’ ), x= j,j+1,…,j+N-1Fitting a straight line:
Figure 127685DEST_PATH_IMAGE014
(13)
whereina r Is the slope of the fitted line, and is calculated by the following formula:
Figure 8178DEST_PATH_IMAGE015
(14)
Figure 198988DEST_PATH_IMAGE016
(15)
Figure 688875DEST_PATH_IMAGE017
(16)
and step 3: (a certain point on the filtered one-dimensional distance distributionj, D j 1D’ ) The slope difference of (d) is calculated by:
Figure 15951DEST_PATH_IMAGE018
(17)
the slope difference of a series of points on the filtered one-dimensional distance distribution forms a slope difference distribution consisting ofS(x),x=N+1,2,…, L-NIt is shown that let the derivative of the slope difference distribution be zero:
Figure 565881DEST_PATH_IMAGE019
(18)
the solution of the above equation is the valley position in the slope difference distributionV i , i=1,2,…,N V And peak positionP i , i=1,2,…, N P And valley levelM i V , i=1,2,…,N V And peak magnitudeM i P , i=1,2,…,N P 。
These valley positionsV i , i=1,2,…,N V And peak positionP i , i=1,2,…,N P One-dimensional feature points of composition slope difference distributionF i 1D , i=1,2,…,N F WhereinN F =N V +N P Two-dimensional feature points of slope difference distributionF i 2D =(x i 2D , y i 2D ),i=1,2,…,N F Is calculated by the following formulaCalculating:
Figure 509567DEST_PATH_IMAGE020
(19)
two-dimensional feature points with distributed slope differencesF i 2D =(x i 2D , y i 2D ), i=1,2,…,N F Normalized by the following formula:
Figure 803145DEST_PATH_IMAGE021
(20)
Figure 250307DEST_PATH_IMAGE022
(21)
each type of object is described by a shape model consisting of normalized two-dimensional slope difference characteristic points, the generated shape model is representative and general, namely the two-dimensional slope difference characteristic points in the shape model are characteristic points of the shape of the type of object, and the generated shape model is formed byF i M =(x i M , y i M ), i=1,2,…,N M Is shown in whichN M Is less than or equal toN F 。
Is calculated by the following formulaSMinimum distance sum of slope difference characteristic point in class shape model and slope difference characteristic point of online detected binary objectd S min ：
Figure 705559DEST_PATH_IMAGE023
(22)
WhereinF i M (
Figure 870961DEST_PATH_IMAGE024
) To representClockwise rotating slope difference characteristic points in the shape model
Figure 702651DEST_PATH_IMAGE024
And (4) degree.
And for each type of object, manually selecting representative and common slope difference characteristic points under lines to generate a shape model.
If the individual shapes of a certain class of objects are very different, a plurality of shape models composed of slope difference characteristic points need to be generated for the class of objects.
When the online object is identified, the minimum distance sum of the slope difference characteristic point of the online detected binary object and the characteristic points of all object shape models is calculated by a formula (22)d S min , S=1,2,…,N S WhereinN S Representing the sum of all generated shape models, the class of the detected binary object
Figure 535477DEST_PATH_IMAGE025
Calculated from the following formula:
Figure 161631DEST_PATH_IMAGE026
(23)。

Claims (8)

1. a method for matching the shape of binary object and recognizing the object features that the one-dimensional distance distribution is obtained by calculating the distances from the center of object to all points on the outline of object, filtering the one-dimensional distance distribution in the frequency domain by discrete Fourier transform, calculating the slope difference distribution of the filtered one-dimensional distance distribution, solving the valley position and peak position of the slope difference distribution by making the derivative of the slope difference distribution equal to zero, mapping the valley position and peak position to the object outline to obtain two-dimensional slope difference characteristic points, modeling the shape of each type of object by utilizing the normalized two-dimensional slope difference characteristic points, and identifying the category of the detected binary object by calculating the minimum distance sum of the normalized two-dimensional slope difference characteristic point of the online detected binary object and the normalized two-dimensional slope difference characteristic point of the shape model of each type of object.
2. The method of claim 1, wherein the center of the binary object is calculated by iteratively corroding the binary object by a morphological corrosion method until the area of the corroded object is smaller than a predetermined or pre-calculated area threshold, and then calculating the center of the binary object by the pixel mean of the final corroded object.
3. The method of claim 1, wherein the one-dimensional distance distribution is formed by calculating distances from a center of the object to each point on the contour of the object, and finally arranging all the distances in sequence to form the one-dimensional distance distribution, wherein the points on the one-dimensional distance distribution correspond to the points on the contour of the object one-to-one.
4. The method of claim 1, wherein the one-dimensional distance distribution is filtered before the slope difference distribution is calculated from the one-dimensional distance distribution, and wherein the filtering is performed in either a time domain or a frequency domain.
5. The method of claim 1, wherein the slope difference is calculated by applying a left side to an arbitrary point on the one-dimensional distance distributionNFitting a straight line to the adjacent points to obtain a left slope, and then distributing any point on the right side of the one-dimensional distance distributionNAnd fitting a straight line to the adjacent points to obtain a right slope, subtracting the left slope from the right slope to obtain the slope difference of the points, and finally arranging the slope differences of all the points together to form slope difference distribution.
6. The method of claim 1, wherein the slope difference feature points are solved by solving for valley positions and peak positions of the slope difference distribution with the derivative of the slope difference distribution equal to zero, and mapping the valley positions and peak positions onto the object contour to obtain two-dimensional slope difference feature points.
7. The method of claim 1, wherein the shape model of the object is comprised of two-dimensional slope difference feature points that are manually selected, wherein the manual selection criteria is selection of feature points having general and representative two-dimensional slope differences, and wherein all of the two-dimensional slope difference feature points are normalized.
8. The method of claim 1, wherein the object is identified by calculating a minimum distance sum of a normalized two-dimensional slope difference feature point of the binary object detected on-line and a normalized two-dimensional slope difference feature point of the shape model of each type of object to identify the type of the binary object detected.
CN201911360618.8A 2019-12-25 2019-12-25 Shape matching and object identification method based on slope difference distribution Withdrawn CN113033592A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911360618.8A CN113033592A (en) 2019-12-25 2019-12-25 Shape matching and object identification method based on slope difference distribution

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911360618.8A CN113033592A (en) 2019-12-25 2019-12-25 Shape matching and object identification method based on slope difference distribution

Publications (1)

Publication Number Publication Date
CN113033592A true CN113033592A (en) 2021-06-25

Family

ID=76458464

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911360618.8A Withdrawn CN113033592A (en) 2019-12-25 2019-12-25 Shape matching and object identification method based on slope difference distribution

Country Status (1)

Country Link
CN (1) CN113033592A (en)

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1786980A (en) * 2005-12-08 2006-06-14 上海交通大学 Melthod for realizing searching new position of person's face feature point by tow-dimensional profile
CN101359403A (en) * 2008-07-28 2009-02-04 上海同盛工程建设配套管理有限公司 Method for extracting contour outline of buildings from satellite imagery
CN102880877A (en) * 2012-09-28 2013-01-16 中科院成都信息技术有限公司 Target identification method based on contour features
CN107103323A (en) * 2017-03-09 2017-08-29 广东顺德中山大学卡内基梅隆大学国际联合研究院 A kind of target identification method based on image outline feature
CN107909047A (en) * 2017-11-28 2018-04-13 上海信耀电子有限公司 A kind of automobile and its lane detection method and system of application
CN109299720A (en) * 2018-07-13 2019-02-01 沈阳理工大学 A kind of target identification method based on profile segment spatial relationship
CN109949326A (en) * 2019-03-21 2019-06-28 苏州工业园区测绘地理信息有限公司 Contour of building line drawing method based on Backpack type three-dimensional laser point cloud data
CN110487579A (en) * 2019-08-28 2019-11-22 湘潭大学 A kind of girder construction damnification recognition method based on inclination angle slope

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1786980A (en) * 2005-12-08 2006-06-14 上海交通大学 Melthod for realizing searching new position of person's face feature point by tow-dimensional profile
CN101359403A (en) * 2008-07-28 2009-02-04 上海同盛工程建设配套管理有限公司 Method for extracting contour outline of buildings from satellite imagery
CN102880877A (en) * 2012-09-28 2013-01-16 中科院成都信息技术有限公司 Target identification method based on contour features
CN107103323A (en) * 2017-03-09 2017-08-29 广东顺德中山大学卡内基梅隆大学国际联合研究院 A kind of target identification method based on image outline feature
CN107909047A (en) * 2017-11-28 2018-04-13 上海信耀电子有限公司 A kind of automobile and its lane detection method and system of application
CN109299720A (en) * 2018-07-13 2019-02-01 沈阳理工大学 A kind of target identification method based on profile segment spatial relationship
CN109949326A (en) * 2019-03-21 2019-06-28 苏州工业园区测绘地理信息有限公司 Contour of building line drawing method based on Backpack type three-dimensional laser point cloud data
CN110487579A (en) * 2019-08-28 2019-11-22 湘潭大学 A kind of girder construction damnification recognition method based on inclination angle slope

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ZHENZHOU WANG: "Slope Difference Distribution and Its Computer Vision Applications", 《HTTPS://ARXIV.ORG/ABS/1910.05704V1》 *
尘风断弦: "计算图像区域的重心（区域像素值的均值）", 《HTTPS://BBS.CSDN.NET/TOPICS/390445741?LIST=LZ》 *

Similar Documents

Publication Publication Date Title
CN112819845B (en) Flexible package substrate contour, line width and line distance defect detection method, medium and equipment
CN109299720B (en) Target identification method based on contour segment spatial relationship
CN103077529B (en) Based on the plant leaf blade characteristic analysis system of image scanning
CN113610917B (en) Circular array target center image point positioning method based on blanking points
CN103839265B (en) SAR image registration method based on SIFT and normalized mutual information
CN104102920A (en) Pest image classification method and pest image classification system based on morphological multi-feature fusion
CN108830888B (en) Coarse matching method based on improved multi-scale covariance matrix characteristic descriptor
CN105740753A (en) Fingerprint identification method and fingerprint identification system
CN106446894A (en) Method for recognizing position of spherical object based on contour
CN109508709B (en) Single pointer instrument reading method based on machine vision
CN112734816B (en) Heterologous image registration method based on CSS-Delaunay
CN111027530A (en) Preprocessing method based on tire embossed character recognition
CN111950559A (en) Pointer instrument automatic reading method based on radial gray scale
CN109359653B (en) Cotton leaf adhesion lesion image segmentation method and system
CN109949324B (en) Contour detection method based on non-classical receptive field nonlinear subunit response
CN117115390A (en) Three-dimensional model layout method of power transformation equipment in transformer substation
CN105550646B (en) Generalized illumination invariant face feature description method based on logarithmic gradient histogram
CN113033592A (en) Shape matching and object identification method based on slope difference distribution
CN111079208A (en) Particle swarm optimization algorithm-based method for identifying surface correspondence between CAD models
CN113781413A (en) Electrolytic capacitor positioning method based on Hough gradient method
JPWO2019008402A5 (en)
CN104036232A (en) Image edge feature analysis-based necktie pattern retrieval method
CN106127851A (en) A kind of method of three-dimensional point cloud object detection based on curved surface segmentation
CN116704446A (en) Real-time detection method and system for foreign matters on airport runway pavement
CN103871048A (en) Straight line primitive-based geometric hash method real-time positioning and matching method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WW01 Invention patent application withdrawn after publication

Application publication date: 20210625

WW01 Invention patent application withdrawn after publication