Method for measuring similarity of cultural relic fragments based on multi-feature information

Abstract

The application discloses a kind of relics fragment similarity measurement methods based on multi-feature information, including collection relics fragment data, clustering relics fragment image, select a kind of clustering after relics fragment image conversion into binary image, extract the contour line of binary image one by one, identify the corner point of contour line and identify the similarity of two relics fragment image step.The application is according to the color information and texture features of ceramic fragments, before fragment splicing, different characteristic ceramic fragments are clustered, then the edge superpixel features and contour curvature features of ceramic fragments are fused, the similarity between fragments is calculated using difference and distance measurement formula, effectively reduce the time and complexity of fragment splicing, improve the accuracy of fragment similarity comparison.

Description

Technical Field

[0001] This invention relates to a method for measuring the similarity of cultural relic fragments, and more particularly to a method for measuring the similarity of cultural relic fragments based on multi-feature information, belonging to the field of computer vision technology. Background Technology

[0002] Cultural relics, as material remains of history, are receiving increasing attention. Archaeologists unearth numerous pottery fragments at archaeological sites. These fragments need to be pieced together. After collection, archaeological technicians clean, classify, and recombine the fragments based on their experience. However, manually piecing together fragments carries the risk of further damage to the relics, and manually piecing together large quantities of ceramic fragments is time-consuming and costly. With the development of computer technology and image processing, archaeologists can use computers to cluster and identify similarities in the fragments for subsequent piecing together work. Currently, many scholars have researched fragment piecing together. Ji Zhouqiang, in his 2009 paper "A New Contour Matching Method for Two-Dimensional Fragments" published in *Computer Applications Research*, Volume 26, Issue 8, proposed a two-dimensional fragment contour matching algorithm that uses the positional relationship between each point and its six adjacent points to represent the contour. This method is very simple in computation and can efficiently improve the speed of fragment contour matching; however, this algorithm cannot handle situations where the fragments are rotated. In his 2014 paper "A Two-Dimensional Fragment Stitching Based on Contour Corner Points and Gray Scale" published in *Laser Journal*, Volume 35, Issue 12, Liao Haibo proposed a method using the angle and distance between corner points as matching features. However, this method only extracts corner points as features for matching and does not consider cases where contour features are not obvious. In her 2010 paper "Automatic color based reassembly of fragmented images and paintings" published in *IEEE Transactions on Image Processing*, Volume 19, Issue 3, Efthymia Tsamoura proposed automatic stitching of colored oil painting fragments based on color information; however, this method only utilizes the color information of the contour and not the shape information. In her 2018 paper "A Two-Way Distance Field Algorithm for Stitching Cultural Relic Fragments Based on Contour Lines" published in *Computer Engineering*, Volume 44, Issue 6, Yuan Jie proposed extracting the display ridge lines of cultural relic fragments to obtain the decorative feature points of the fragments. This method can achieve good stitching results for cultural relic fragments with relatively clear patterns. However, this method does not take into account the possibility that the surface texture of the fragments may be worn away after being buried deep underground for a long time. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method for measuring the similarity of cultural relic fragments based on multi-feature information.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0005] A method for measuring the similarity of cultural relic fragments based on multi-feature information includes the following steps:

[0006] Step 1: Collect artifact fragment data: Obtain images of one or more artifact fragments, denoted as Pm = {p1, p2, ..., p...} mm}, where mm represents the number of artifact fragments;

[0007] Step 2: Clustering of artifact fragment images: The clustered categories C = {C1, C2, ..., C6} nn}, where n is the number of clusters. The i-th fragment belongs to the j-th category;

[0008] Step 3: Select one cluster of cultural relic fragment images and convert it into a binarized image;

[0009] Step 4: Extract the contour lines of the binarized image one by one;

[0010] Step 5: Identify the corner points of the contour line and divide the contour line into contour line segments based on the corner points;

[0011] Step 6: Identify the similarity between the two artifact fragment images, which consists of the following specific steps:

[0012] Step 6-1: Predict the wear on the edges of the artifact fragments, delineate the fragment edge pixels along the contour line according to the pre-set prediction bandwidth; predict fragment edge pixels using the neighboring pixels of the fragment edge pixels;

[0013] Step 6-2: Extract the first... The superpixels of the edge pixels of the fragment image constitute the first fragment. A collection of superpixel edges of artifact fragments ,in Representing the The first fragment of the cultural relic The edge superpixel, in is the th edge superpixel The number of superpixels at the edge of each artifact fragment;

[0014] Step 6-3: Use the vertical distance method to extract the feature point set on each contour line. Each pair of feature points, separated by one feature point, is combined into a feature point pair. A feature line containing the feature point pair is fitted, and the perpendicular distance between the middle feature point and the corresponding feature line is calculated. If the perpendicular distance is less than a preset distance threshold, the middle feature point is deleted. The resulting set of feature points is... Every three consecutive filtered feature points form a feature segment, and the curvature set of each feature segment is calculated. ,in Representing the The first outline of the contour line One curvature;

[0015] Step 6-4: Calculate the chord length of the feature segment formed by every 3 consecutive filtered feature points:

[0016] (1)

[0017] Construct a set of chord lengths ;

[0018] Step 6-5: Calculate the difference and SoD distance between the superpixel sets of the two contours:

[0019] (2)

[0020] (3)

[0021] (4)

[0022] (5)

[0023] Step 7: Similarity judgment: When SOD is less than the similarity threshold, the two contour feature segments are considered similar.

[0024] Furthermore, step 5 comprises the following specific steps:

[0025] Step 5-1: Use a polygon fitting function to approximate the outline of the artifact fragment, fitting the shape of the fragment into a polygon. Then, extract the key points of the fitted polygon as the corner points of the artifact fragment. Contours = {(x1, y1), (x2, y2), ..., (xnc, ync)} is used to segment the fragment outline to obtain the outline segments. c represents the total number of key points fitted;

[0026] Step 5-2: Based on corner points The contour line is segmented, and the set of segmented contour line segments is represented as follows:

[0027] L={l1, l2,..., lnl} (6)

[0028] This represents the i'-th contour line segment after segmentation. .

[0029] Furthermore, step 5-1 uses the B-spline interpolation function to fit the contour line.

[0030] The beneficial effects of the above technical solution are as follows:

[0031] (1) This invention uses the Meanshift method to cluster ceramic fragments with different characteristics based on their color information and texture features. By clustering the ceramic fragments before splicing them together, the time and complexity of splicing can be effectively reduced.

[0032] (2) The present invention predicts the edge wear of ceramic fragments, which is a method that can effectively improve the accuracy of fragment similarity identification and greatly reduces the impact of fragment fading and corrosion caused by long-term deep burial of ceramic fragments on similarity identification;

[0033] (3) The present invention is based on the method of fitting contour lines using B-spline interpolation function and extracting feature points using the vertical distance method. It accurately extracts the feature points of the contour lines and simplifies the data, reducing the time complexity of the algorithm and improving the computational efficiency of the algorithm. It replaces the complex and redundant process with a simple task.

[0034] (4) The present invention integrates the edge superpixel features and contour curvature features of ceramic fragments, and uses difference and distance metric formulas to calculate the similarity between fragments, thereby improving the accuracy of fragment similarity comparison. Attached Figure Description

[0035] Figure 1 This is a flowchart of the present invention.

[0036] Figure 2 This is a schematic diagram illustrating the extraction of fragment color information in an embodiment of the present invention.

[0037] Figure 3 This is a contour coordinate diagram of the contour lines extracted in an embodiment of the present invention.

[0038] Figure 4 This is a diagram showing the corner points after polygon fitting in an embodiment of the present invention. Detailed Implementation

[0039] Example 1

[0040] A method for measuring the similarity of cultural relic fragments based on multi-feature information includes the following steps:

[0041] Step 1: Collect artifact fragment data: Obtain images of one or more artifact fragments, denoted as Pm = {p1, p2, ..., p...} mm}, where mm represents the number of artifact fragments;

[0042] Step 2: Clustering artifact fragment images: This consists of the following specific steps:

[0043] Step 2-1: Extract the texture features Am = {a1, a2, ..., a...} of the artifact fragment images one by one. am This embodiment uses the Local Binary Pattern (LBP) operator to extract texture features from images of artifact fragments.

[0044] Step 2-2: Calculate the color histogram {h1, h2, ..., h} of each artifact fragment image. hm} and color information entropy B m = Where bm is the number of color grades;

[0045] Steps 2-3: This embodiment uses the Meanshift method to cluster images of cultural relic fragments, grouping fragments with different colors and textures into different categories. Let the categories be C = {C1, C2, ..., C...} nn}, where n is the number of clusters. The i-th fragment belongs to the j-th category.

[0046] Step 3: Select one cluster of cultural relic fragment images and convert it into a binarized image; this embodiment uses... The algorithm binarizes the clustered fragments;

[0047] Step 4: Extract the contour lines of the binarized image one by one;

[0048] Step 5: Identify the corner points of the contour line and divide the contour line into contour line segments based on the corner points;

[0049] Step 6: Identify the similarity between the two artifact fragment images, which consists of the following specific steps:

[0050] Step 6-1: Predict the wear on the edge of the artifact fragment, and delineate the fragment edge pixels along the contour line according to the pre-set prediction bandwidth; predict the fragment edge pixels using the neighboring pixels of the fragment edge pixels; in this embodiment, the prediction bandwidth is 2% of the fragment perimeter.

[0051] Step 6-2: Extract the first... The superpixels of the edge pixels of the fragment image constitute the first fragment. A collection of superpixel edges of artifact fragments ,in Representing the The first fragment of the cultural relic The edge superpixel, in is the th edge superpixel The number of superpixels at the edge of each artifact fragment;

[0052] Step 6-3: Use the vertical distance method to extract the feature point set on each contour line. Each pair of feature points, separated by one feature point, is combined into a feature point pair. A feature line containing the feature point pair is fitted, and the perpendicular distance between the middle feature point and the corresponding feature line is calculated. If the perpendicular distance is less than a preset distance threshold, the middle feature point is deleted. The resulting set of feature points is... Every three consecutive filtered feature points form a feature segment, and the curvature set of each feature segment is calculated. ,in Representing the The first outline of the contour line One curvature;

[0053] Step 6-4: Calculate the chord length of the feature segment formed by every 3 consecutive filtered feature points:

[0054] (1)

[0055] Construct a set of chord lengths ;

[0056] Step 6-5: Calculate the difference and SoD distance between the superpixel sets of the two contours:

[0057] (2)

[0058] (3)

[0059] (4)

[0060] (5)

[0061] Step 7: Similarity judgment: When SOD is less than the similarity threshold, the two contour feature segments are considered similar.

[0062] Step 5 consists of the following specific steps:

[0063] Step 5-1: Use a polygon fitting function to approximate the outline of the artifact fragment, fitting the shape of the fragment into a polygon. Then, extract the key points of the fitted polygon as the corner points of the artifact fragment. Contours = {(x1, y1), (x2, y2), ..., (xnc, ync)} is used to segment the fragment outline to obtain the outline segments. This represents the total number of key points fitted; in this embodiment, a cubic spline interpolation function is used to fit the contour line.

[0064] Step 5-2: Based on corner points The contour line is segmented, and the set of segmented contour line segments is represented as follows:

[0065] L={l1, l2,..., lnl} (6)

[0066] This represents the i'-th contour line segment after segmentation. .

Claims

1. A method for measuring the similarity of cultural relic fragments based on multi-feature information, characterized in that: Includes the following steps: Step 1: Collect artifact fragment data: Obtain images of one or more artifact fragments, denoted as Pm = {p1, p2, ..., p...} mm }, where mm represents the number of artifact fragments; Step 2: Clustering artifact fragment images: Clustered categories Where n is the number of clusters after clustering. The i-th fragment belongs to the j-th category; Step 3: Select one cluster of cultural relic fragment images and convert it into a binarized image; Step 4: Extract the contour lines of the binarized image one by one; Step 5: Identify the corner points of the contour line and divide the contour line into contour line segments based on the corner points; Step 6: Identify the similarity between the two artifact fragment images, which consists of the following specific steps: Step 6-1: Predict the wear on the edges of the artifact fragments and delineate the fragment edge pixels along the contour line according to the pre-set prediction bandwidth; Predict fragment edge pixels using neighboring pixels of fragment edge pixels; Step 6-2: Extract the first... The superpixels of the edge pixels of the fragment image constitute the first fragment. A collection of superpixel edges of artifact fragments ,in Representing the The first fragment of the cultural relic The edge superpixel, in is the th edge superpixel The number of superpixels at the edge of each artifact fragment; Step 6-3: Use the vertical distance method to extract the feature point set on each contour line. Each pair of feature points, separated by one feature point, is combined into a feature point pair. A feature line containing the feature point pair is fitted, and the perpendicular distance between the middle feature point and the corresponding feature line is calculated. If the perpendicular distance is less than a preset distance threshold, the middle feature point is deleted. The resulting set of feature points is... Every three consecutive filtered feature points form a feature segment, and the curvature set of each feature segment is calculated. ,in Representing the The first outline of the contour line One curvature; Step 6-4: Calculate the chord length of the feature segment formed by every 3 consecutive filtered feature points: (1) Construct a set of chord lengths ; Step 6-5: Calculate the difference and SoD distance between the superpixel sets of the two contours: (2) (3) (4) (5) Step 7: Similarity judgment: When SOD is less than the similarity threshold, the two contour feature segments are considered similar.

2. The method for measuring the similarity of cultural relic fragments based on multi-feature information according to claim 1, characterized in that: Step 5 is described in detail below: Step 5-1: Use a polygon fitting function to approximate the outline of the artifact fragment, fitting the shape of the artifact fragment into a polygon, and then extract the key points of the fitted polygon as the corner points of the artifact fragment. , is used to segment the outline of the fragments to obtain the outline segments, where nc represents the total number of fitted lines; Step 5-2: Based on corner points The contour line is segmented, and the set of segmented contour line segments is represented as follows: L={l1, l2,..., lnl} (6) This represents the i-th contour line segment after segmentation. .

3. The method for measuring the similarity of cultural relic fragments based on multi-feature information according to claim 2, characterized in that: Step 5-1: Fit the contour line using the B-spline interpolation function.

4. The method for measuring the similarity of cultural relic fragments based on multi-feature information according to claim 2, characterized in that: Step 5-1 uses a third-order spline interpolation function to fit the contour line.

5. The method for measuring the similarity of cultural relic fragments based on multi-feature information according to claim 2, characterized in that: Step 5-1 uses a fourth-order spline interpolation function to fit the contour line.

6. The method for measuring the similarity of cultural relic fragments based on multi-feature information according to claim 2, characterized in that: Step 5-1 uses a fifth-order spline interpolation function to fit the contour line.

7. The method for measuring the similarity of cultural relic fragments based on multi-feature information according to claim 2, characterized in that: Step 5-1 uses a sixth-order spline interpolation function to fit the contour line.

Images