KR20170062661A - Automatic source classification method and apparatus using mean-shift clustering and stepwise merging in color image - Google Patents
Automatic source classification method and apparatus using mean-shift clustering and stepwise merging in color image Download PDFInfo
- Publication number
- KR20170062661A KR20170062661A KR1020150168074A KR20150168074A KR20170062661A KR 20170062661 A KR20170062661 A KR 20170062661A KR 1020150168074 A KR1020150168074 A KR 1020150168074A KR 20150168074 A KR20150168074 A KR 20150168074A KR 20170062661 A KR20170062661 A KR 20170062661A
- Authority
- KR
- South Korea
- Prior art keywords
- color
- cluster
- mean
- image
- data
- Prior art date
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
- B07C5/34—Sorting according to other particular properties
-
- G06K9/4652—
-
- G06K9/6215—
Abstract
More particularly, the present invention relates to a method and apparatus for automatically selecting raw materials by color-image clustering using mean-shift clustering and stepwise merging. More specifically, (3) dividing the foreground map image into N clusters by applying the extracted foreground map image to a mean-shift clustering algorithm, (4) dividing the foreground map image into N clusters, Selecting the largest cluster in the true N clusters as the seed cluster, (5) selecting the seed cluster and the remaining clusters as the seed cluster, (5) locating proximity between the seed clusters and the surrounding clusters, (6) merging the two closures when the proximity and color similarity are less than or equal to a threshold value, (6) Transforming the RG / GB / BR into a two-dimensional color distribution diagram of RG / GB / BR, and generating ellipses based on the cluster data of the respective color distribution diagrams; and (7) And a step of classifying the good and the defective of the defective product.
According to the automatic raw material sorting and apparatus using the mean-shift clustering and the stepwise merging method in the color image proposed in the present invention, the raw materials are automatically selected using the mean-shift clustering based on the color image and the stepwise merging method, It is possible to select raw materials more accurately and automatically with less artificial manipulation of users for various colors of raw materials.
Description
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automatic raw material sorting method and apparatus, and more particularly, to a method and apparatus for automatic raw material sorting using mean-shift clustering and stepwise merging in a color image.
Recently, as a result of frequent occurrence of safety accidents due to agricultural and marine products produced as a bad raw material, there is a growing need for technology research for selecting various raw materials as good and defective products. Such raw material sorting technology can be applied not only to agricultural and marine products but also to various kinds of raw materials such as ores and waste materials, and is a necessary technique for processing high quality goods.
The method of screening raw materials which is mainly used at present is a visual screening method by skillful workers. However, this method requires skilled workers for decades, and even if skilled workers increase their fatigue due to prolonged repetitive labor, May decrease. In order to solve these problems, a method of selecting the raw materials using the brightness information of the raw material from the images input from the Mono CCD camera has been developed. However, this method is effective for monochromatic raw materials such as grain, but there is a problem that the screening performance is reduced for raw materials containing various colors such as ores and waste materials. In order to solve these problems, classification filters using raw material colors have been developed by using a Trichromatic CCD camera or a color CCD camera for increasing the sorting power of raw materials of various colors. However, this method also has a problem that the raw material itself contains noise There was a problem that there were many misclassifications.
With respect to such a raw material selection method and apparatus, Japanese Patent Application Laid-Open No. 10-2014-0055042 (titled invention: raw material sorting apparatus and sorting method, public date: May 9, 2014) has been disclosed.
The present invention has been proposed in order to solve the above-mentioned problems of the previously proposed methods. By automatically selecting raw materials using mean-shift clustering and stepwise merging based on color images, It is an object of the present invention to provide a method and apparatus for automatic raw material sorting using mean-shift clustering and stepwise merging methods in color images, which can automatically select raw materials with less human manipulation and more accurately.
According to an aspect of the present invention, there is provided an automatic material selection method using mean-shift clustering and stepwise merging in a color image,
As an automatic raw material selection method,
(1) receiving a raw material image photographed in color;
(2) extracting a foreground map image from the input color material image by removing the background;
(3) applying the extracted foreground map image to a mean-shift clustering algorithm, and dividing the foreground map image into N clusters;
(4) selecting the largest cluster in the divided N clusters as a seed cluster;
(5) For each selected seed cluster and each of the remaining clusters, color similarity between the representative proximity of the seed cluster and the neighboring cluster and the representative color of the seed cluster and the representative cluster of the surrounding cluster is measured, Merging the two clusters if the threshold value is less than the threshold value;
(6) transforming the merged cluster data into a two-dimensional color distribution diagram of RG / GB / BR, and generating ellipses based on the cluster data of the respective color distribution diagrams; And
(7) classifying the good and defective products of the raw material on the basis of whether or not they are included in the generated ellipse.
Preferably, the step (3)
(3-1) designating a window area based on an arbitrary starting pixel in the extracted foreground map image.
More preferably, the step (3)
(3-2) determining a color of the starting pixel as a reference color value, and searching for data similar to the starting pixel in the designated window area.
More preferably, the step (3)
(3-3) determining a coordinate average of the similar data as a new center of gravity, and determining an average color of the similar data as a new reference color value.
More preferably, the step (3)
(3-4) re-determining the window area based on the newly defined center of gravity, and searching for data similar to the newly-determined reference color value in the designated window area.
Preferably, the step (3)
(3-5) repeating the operations of steps (3-1) to (3-4) until the center of gravity converges to determine the average color of the data obtained at the last time as the reference color value have.
Preferably, the step (3)
(3-6) The other pixels of the extracted foreground map image are also subjected to the operations of steps (3-1) to (3-5) to obtain respective reference color values, and based on the similarity between the reference color values And dividing the foreground map image into N clusters.
Preferably, in said step (5)
The threshold values of the position proximity and color similarity can be predetermined.
Advantageously, said step (6)
(6-1) calculating a slope of the cluster data by calculating a linear regression function from the cluster data of the respective color distribution diagrams, and calculating a height and a width of the cluster data of the respective color distribution maps, And obtaining a shortening value.
More preferably, the step (6)
(6-2) generating an ellipse including the cluster data using the slope, the long axis value, the short axis value, and the center point.
According to another aspect of the present invention, there is provided an automatic material selection apparatus using mean-shift clustering and stepwise merging in a color image,
An automatic material sorting apparatus using mean-shift clustering and stepwise merging in a color image by the processor,
The processor comprising:
The raw material image photographed in color is input,
Extracting a foreground map image from the input color material image by removing the background,
Applying the extracted foreground map image to a mean-shift clustering algorithm, dividing the foreground map image into N clusters,
The largest cluster in the divided N clusters is selected as the seed cluster,
Wherein the color similarity between the representative proximity of the seed cluster and the neighboring cluster and the representative color of the seed cluster and the representative color of the neighboring cluster are measured for the selected seed cluster and the remaining populations, , The two clusters are merged,
Transforming the merged cluster data into a two-dimensional color distribution diagram of RG / GB / BR, generating ellipses based on the cluster data of the respective color distribution diagrams,
And classifies the good and defective parts of the raw material on the basis of whether or not they are included in the generated ellipse.
Advantageously,
And may designate a window area based on an arbitrary start pixel in the extracted foreground map image.
More preferably, the processor is configured to:
Determining a color of the starting pixel as a reference color value and looking for data similar to the starting pixel within the designated window area.
More preferably, the processor is configured to:
Determine a coordinate mean of the similar data as a new center of gravity and define an average color of the similar data as a new reference color value.
More preferably, the processor is configured to:
The window region may be re-determined on the basis of the newly defined center of gravity, and data similar to the newly-determined reference color value may be searched in the designated window region.
Advantageously,
And repeating the operation until the center of gravity converges to define the average color of the last data as a reference color value.
Advantageously,
Other pixels of the extracted foreground map image may be implemented to obtain respective reference color values through repetitive operations and to divide the foreground map image into N clusters based on the similarity between the respective reference color values.
Advantageously,
The threshold value of the position proximity and color similarity can be predetermined.
Advantageously,
Calculating a slope of the cluster data by calculating a linear regression function from the cluster data of the respective color distribution diagrams and calculating a height and a width of the cluster data of the respective color distribution diagrams to calculate a center point of data and a long axis value and a short axis value of the ellipse, . ≪ / RTI >
Advantageously,
The ellipse including the cluster data may be generated using the slope, the long axis value, the short axis value, and the center point.
According to the automatic raw material sorting and apparatus using the mean-shift clustering and the stepwise merging method in the color image proposed in the present invention, the raw materials are automatically selected using the mean-shift clustering based on the color image and the stepwise merging method, It is possible to select raw materials more accurately and automatically with less artificial manipulation of users for various colors of raw materials.
FIG. 1 is a flowchart illustrating an entire process of an automatic raw material sorting method using mean-shift clustering and stepwise merging in a color image, according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating a raw material image picked up in color in an automatic raw material sorting method using mean-shift clustering and stepwise merging method in a color image according to an embodiment of the present invention. FIG.
FIG. 3 is a diagram illustrating an automatic raw material selection method using mean-shift clustering and a stepwise merging method in a color image according to an embodiment of the present invention. A drawing.
FIG. 4 is a flowchart illustrating a method of selecting a foreground map by applying a mean-shift clustering algorithm and a stepwise merging method in a color image according to an embodiment of the present invention. Fig. 2 is a diagram showing a flow of a process of dividing into N clusters;
FIG. 5 is a diagram illustrating an automatic raw material selection method using a mean-shift clustering and a stepwise merging method in a color image according to an embodiment of the present invention. In the automatic raw material selection method, a window region is designated based on an arbitrary start pixel in the extracted foreground map FIG.
FIG. 6 is a flowchart illustrating a method of selecting a raw material using mean-shift clustering and stepwise merging in a color image according to an embodiment of the present invention. In the automatic raw material sorting method, the largest cluster among N clusters is called a seed Fig.
FIG. 7 is a graph showing the relationship between the proximity of selected seed clusters and neighboring clusters, and the representative colors of selected seed clusters and neighboring clusters in the automatic raw material sorting method using Mean-Shift clustering and stepwise merging method in a color image according to an embodiment of the present invention. The color similarity between the representative colors of the community is measured to show the merging of the two populations when the location proximity and color similarity are below the threshold value
FIG. 8 is a flowchart illustrating a method of automatically sorting raw materials using mean-shift clustering and stepwise merging in a color image according to an exemplary embodiment of the present invention. Referring to FIG. 8, a linear regression function is calculated from the cluster data of each color distribution, Drawing of a slope rule.
9 is a flowchart illustrating a method of automatically selecting a raw material using mean-shift clustering and stepwise merging in a color image according to an embodiment of the present invention. The automatic data sorting method rotates the slope of the cluster data distribution to be 0, Drawing the long axis, the short axis and the center point of the ellipse to be generated using the width and the height.
10 illustrates generation of an ellipse using a slope, a long axis, a short axis, and a center point in a method of automatic material selection using mean-shift clustering and a stepwise merging method in a color image, according to an embodiment of the present invention.
Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention. In the following detailed description of the preferred embodiments of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. The same or similar reference numerals are used throughout the drawings for portions having similar functions and functions.
In addition, in the entire specification, when a part is referred to as being 'connected' to another part, it may be referred to as 'indirectly connected' not only with 'directly connected' . Also, to "include" an element means that it may include other elements, rather than excluding other elements, unless specifically stated otherwise.
1 is a flowchart illustrating an entire process of an automatic material selection method using Mean-Shift clustering and a stepwise merging method in a color image, according to an embodiment of the present invention. As shown in FIG. 1, an automatic material selection method using mean-shift clustering and stepwise merging method in a color image according to an embodiment of the present invention includes the steps of (1) receiving raw material images photographed in color Step S200 of extracting the foreground map image from the input color material image by removing the background, (3) applying the extracted foreground map image to the mean-shift clustering algorithm, (S300), (4) selecting the largest cluster in the N clusters divided into the seed cluster (S400), (5) selecting the seed cluster and the remaining clusters, (S500) of merging two clusters when the location proximity and color similarity are below the threshold value (S500), and (6) when the location proximity and the color similarity are below the threshold, the color similarity between the representative color of the seed cluster and the representative color of the surrounding cluster is measured Clustering data to RG / GB / BR (S600) of generating an ellipse based on the cluster data of each color distribution chart, and (7) classifying the good and defective items of the raw material based on whether or not the raw material is included in the generated ellipse Step S700.
Hereinafter, each step of the automatic material sorting method proposed by the present invention using mean-shift clustering and stepwise merging method in a color image will be described in detail with reference to the drawings.
In step S100, raw material images shot in color can be input. FIG. 2 is a diagram illustrating a raw material image picked up in color in an automatic material selection method using mean-shift clustering and stepwise merging in a color image according to an exemplary embodiment of the present invention. As shown in FIG. 2, in step S100, raw material images photographed in color can be input. Here, the camera for photographing the raw material image in color may be a color CCD camera. However, the color CCD camera is according to an embodiment of the present invention, and various types of color cameras capable of shooting in color with respect to raw materials of various colors can be used.
In step S200, the foreground map image may be extracted by removing the background from the color material image input through step S100. 3 is a view illustrating an automatic raw material selection method using mean-shift clustering and a stepwise merging method in a color image, in which a foreground image is extracted by removing a background from a raw material image input in color, according to an embodiment of the present invention FIG. As shown in FIG. 3, in step S200, the foreground map image may be extracted by removing the background from the color material image input through step S100.
In step S300, the extracted foreground map image is applied to a mean-shift clustering algorithm, and the foreground map image can be divided into N clusters. The process of dividing the extracted foreground map image into the N clusters by applying the mean-shift clustering algorithm will be described in detail with reference to FIG. 4 and FIG.
FIG. 4 is a flowchart illustrating a method of selecting a foreground map by applying a mean-shift clustering algorithm and a stepwise merging method in a color image according to an embodiment of the present invention. FIG. 3 is a view showing a flow of a process of dividing into N clusters. As shown in FIG. 4, the process of dividing the foreground map image into N clusters by applying the foreground map image to the mean-shift clustering algorithm includes (3-1) extracting an arbitrary start pixel in the extracted foreground map image (S320), (3-3) determining a color of the starting pixel as a reference color value and finding data similar to the starting pixel in the designated window area, (S330), determining a coordinate average of similar data as a new center of gravity, and setting an average color of similar data as a new reference color value, (3-4) redefining the window area based on the newly determined center of gravity, (S340). (3-5) Repeat the operations of steps S310 to S340 until the center of gravity converges to obtain the data of the last obtained data (S350), and (3-6) the other pixels of the extracted foreground map image also obtain the respective reference color values through the operations of steps S310 to S350, and determine the relationship between the respective reference color values And dividing the foreground map image into N clusters based on the degree of similarity (S360). Hereinafter, steps S310 to S360 will be described in more detail.
In step S310, a window area can be designated based on an arbitrary starting pixel in the extracted foreground map image. FIG. 5 is a diagram illustrating an automatic raw material selection method using a mean-shift clustering and a stepwise merging method in a color image according to an embodiment of the present invention. In the automatic raw material selection method, a window region is designated based on an arbitrary start pixel in the extracted foreground map Fig. As shown in FIG. 5, an arbitrary pixel may be designated as a start pixel within the extracted foreground map image, and a window region may be designated around the start pixel.
In step S320, the color of the starting pixel determined in step S310 is determined as the reference color value, and data similar to the color value of the starting pixel in the window area designated in step S310 can be found.
In step S330, the coordinates of the similar data found in step S320 may be averaged to determine a new center of gravity, and the average color of similar data may be set to a new reference color value.
In step S340, a new window area is re-centered around the center of gravity defined in step S330, and data similar to the reference color value newly defined in S330 in the newly designated window area can be found.
In step S350, the operations of steps S310 to S340 may be repeated until the center of gravity converges to determine the average color of the last data obtained as the reference color value.
In step S360, the other pixels of the extracted foreground map image are also subjected to the operations of steps S310 to S350 to obtain respective reference color values, and the foreground map images are divided into N clusters based on the similarity between the respective reference color values .
In step S400, the extracted foreground map image is applied to the mean-shift clustering algorithm, and the largest cluster among the clusters obtained by dividing the foreground map image by N can be selected as the seed cluster. FIG. 6 is a flowchart illustrating a method of selecting a raw material using mean-shift clustering and stepwise merging in a color image according to an embodiment of the present invention. In the automatic raw material sorting method, the largest cluster among N clusters is called a seed Fig. As shown in FIG. 6, in step S400, the extracted foreground map image is applied to the mean-shift clustering algorithm, and the largest cluster among the clusters obtained by dividing the foreground map image by N can be selected as the seed cluster.
In step S500, the color similarity between the representative proximity of the selected seed community and the neighboring communities and the representative color of the surrounding communities and the representative images of the surrounding communities are measured for the selected seed community and the remaining communities in step S400, Is less than or equal to the threshold, the two clusters can be merged. FIG. 7 is a graph showing the relationship between the proximity of selected seed clusters and neighboring clusters, and the representative colors of selected seed clusters and neighboring clusters in the automatic raw material sorting method using Mean-Shift clustering and stepwise merging method in a color image according to an embodiment of the present invention. The color similarity between the representative colors of the cluster is measured, and when the proximity of the position and the color similarity are equal to or less than the threshold value, the two clusters are merged. As shown in FIG. 7, in step S500, the proximity and color similarity between the selected seed cluster and the other clusters are compared on the basis of the threshold value. When the seed proximity and color similarity are below the threshold value, two clusters can be integrated, New clusters may be created. Here, the threshold values of the position proximity and the color similarity can be predetermined.
In addition, if the number of data forming the cluster is less than the predetermined number of data among the integrated clusters and the newly generated clusters, the cluster may be deleted. For example, if the number of data in the candidate cluster is less than or equal to 6% of all the raw materials, the cluster is regarded as a border and is set to be deleted. In step S500, among the integrated clusters and the newly generated clusters, If the data count is less than 6% of the total raw material, the cluster may be deleted.
In step S600, the merged cluster data is converted into a two-dimensional color distribution diagram of RG / GB / BR through step S500, and an ellipse can be generated based on the cluster data of each color distribution diagram. The process of generating ellipses based on the cluster data of each color distribution diagram will be described in detail with reference to FIG. 8 through FIG.
FIG. 8 is a flowchart illustrating a method of automatically sorting raw materials using mean-shift clustering and stepwise merging in a color image according to an exemplary embodiment of the present invention. Referring to FIG. 8, a linear regression function is calculated from the cluster data of each color distribution, And a slope rule is obtained. As shown in FIG. 8, the slope of the corresponding cluster data can be obtained by calculating a linear regression function from the cluster data converted into the color distribution diagram.
9 is a flowchart illustrating a method of automatically selecting a raw material using mean-shift clustering and stepwise merging in a color image according to an embodiment of the present invention. The automatic data sorting method rotates the slope of the cluster data distribution to be 0, Axis value, the short axis value and the center point of the ellipse to be generated using the width and height. As shown in FIG. 9, the long axis value, the short axis value and the center point of the ellipse to be generated can be obtained using the width and height of the data distributed in the cluster, so that the slope of the distributed cluster data is zero have. For example, by measuring the height and width of data distributed in a cluster, a long value of both lengths can be a long axis value of an ellipse to be generated, and a short value can be a short axis value of an ellipse have. In addition, the intersection of both axes can be the center point of the ellipse to be generated.
In step S610, as shown in Figs. 8 and 9, the linear regression function is calculated from the cluster data of each color distribution chart to obtain the slope of the corresponding cluster data, and the cluster data of each color distribution chart By calculating the height and width, the long axis value, the short axis value and the center point of the ellipse to be generated can be obtained.
In step S620, an ellipse including cluster data may be generated using the slope, the long axis value, the short axis value, and the center point obtained in step S610. 10 is a diagram illustrating generation of an ellipse using a slope, a long axis, a short axis, and a center point in a method of automatic material selection using mean-shift clustering and a stepwise merging method in a color image according to an embodiment of the present invention. As shown in FIG. 10, an ellipse including cluster data can be generated using the slope, long axis value, short axis value, and center point obtained from the cluster data.
In step S700, through the ellipses generated in step S600, it is possible to classify the good and defective products of the raw materials on the basis of whether they are included in the ellipse.
The automatic raw material selection method using the Mean-Shift clustering and the stepwise merging method in the color image is performed by using at least one processor 100 and mean-shift clustering and stepwise merging methods in the color image by the processor 100 And may be implemented as an automatic raw material sorting apparatus 10.
The present invention may be embodied in many other specific forms without departing from the spirit or essential characteristics of the invention.
S100: Receiving a raw material image photographed in color
S200: extracting the foreground map image by removing the background from the inputted color material image
S300: dividing the foreground map image into N clusters by applying the extracted foreground map image to the mean-shift clustering algorithm
S310: designating a window area based on an arbitrary starting pixel in the extracted foreground map image
S320: determining the color of the starting pixel as a reference color value and searching for data similar to the starting pixel within the designated window area
S330: determining a coordinate average of similar data as a new center of gravity, and determining an average color of similar data as a new reference color value
S340: Redetermining the window area based on the new center of gravity and searching for data similar to the new reference color value in the designated window area
S350: repeating the operations of steps S310 to S340 until the center of gravity converges, and determining an average color of data obtained at the last time as a reference color value
S360: Another pixel of the extracted foreground map image is obtained by obtaining the respective reference color values through the operations of steps S310 to S350, and dividing the foreground map image into N clusters based on the similarity between the respective reference color values
S400: Selecting the largest cluster in the divided N clusters as the seed cluster
S500: For the selected seed cluster and each of the remaining clusters, the color similarity between the seed proximity between the seed cluster and the surrounding cluster, and the representative color of the selected seed cluster and the representative cluster of the surrounding cluster are measured, and the location proximity and color similarity are equal to or less than the threshold , Merging the two clusters
S600: converting the merged cluster data into a two-dimensional color distribution diagram of RG / GB / BR and expressing it, and generating an ellipse based on the cluster data of each color distribution diagram
S610: Calculate the slope of the corresponding cluster data by calculating the linear regression function from the cluster data of each color distribution chart, and calculate the height and width of the cluster data of each color distribution map to calculate the center point of the cluster data, the long axis value and the short axis value of the ellipse Step to Obtain
S620: generating an ellipse including cluster data using the obtained slope, long axis value, short axis value and center point
S700: Classifying the good and defective parts of the raw materials based on whether or not they are included in the generated ellipses
10: Automatic raw material sorting apparatus using mean-shift clustering and stepwise merging method in a color image according to an embodiment of the present invention
100: Processor
Claims (20)
(1) receiving a raw material image photographed in color (S100);
(2) extracting a foreground map image from the input color material image by removing the background (S200);
(3) applying the extracted foreground map image to a mean-shift clustering algorithm to divide the foreground map image into N clusters (S300);
(4) selecting the largest cluster among the N clusters divided as seed clusters (S400);
(5) For each selected seed cluster and each of the remaining clusters, color similarity between the representative proximity of the seed cluster and the neighboring cluster and the representative color of the seed cluster and the representative cluster of the surrounding cluster is measured, Merging the two clusters if the threshold value is less than the threshold value (S500);
(6) transforming the merged cluster data into a two-dimensional color distribution diagram of RG / GB / BR, and generating ellipses based on the cluster data of the respective color distribution diagrams (S600); And
(7) classifying the good and defective parts of the raw material on the basis of whether or not they are included in the generated ellipse (S700). The automatic raw material sorting using the mean-shift clustering and the stepwise merging method in the color image Way.
(3-1) designating a window region based on an arbitrary starting pixel in the extracted foreground map image (S310), and performing a mean-shift clustering and a stepwise merging method on the color image Automatic raw material selection method.
(3-2) determining a color of the starting pixel as a reference color value, and searching for data similar to the starting pixel in the designated window area (S320) Automatic Raw Material Selection Method Using Clustering and Stepwise Merge Method.
(3-3) determining a coordinate average of the similar data as a new center of gravity, and setting an average color of the similar data as a new reference color value (S330). Automatic Raw Material Selection Method Using Clustering and Stepwise Merge Method.
(3-4) re-determining the window area based on the newly defined center of gravity, and searching for data similar to the newly-determined reference color value in the designated window area (S340) Automatic raw material selection method using mean-shift clustering and stepwise merging method in image.
(3-5) Repeating the operations of steps (3-1) to (3-4) until the center of gravity converges (S350), the average color of the data obtained at the end is determined as the reference color value The method comprising the steps of: means-shift clustering and stepwise merging in a color image.
(3-6) The other pixels of the extracted foreground map image are also subjected to the operations of steps (3-1) to (3-5) to obtain respective reference color values, and based on the similarity between the reference color values Further comprising a step S360 of dividing the foreground map image into N clusters by using the mean-shift clustering method and the stepwise merging method in the color image.
Wherein the threshold value of the position proximity and the color similarity can be determined in advance. The automatic raw material sorting method using mean-shift clustering and stepwise merging method in a color image.
(6-1) calculating a slope of the cluster data by calculating a linear regression function from the cluster data of the respective color distribution diagrams, and calculating a height and a width of the cluster data of the respective color distribution maps, (S610) of obtaining a short axis value by using a mean-shift clustering method and a stepwise merge method in a color image.
(6-2) generating an ellipse including the cluster data using the obtained slope, the long axis value, the short axis value, and the center point (S620). Automatic Raw Material Selection Method Using Stepwise Merge Method.
The processor (100)
The raw material image photographed in color is input,
Extracting a foreground map image from the input color material image by removing the background,
Applying the extracted foreground map image to a mean-shift clustering algorithm to divide the foreground map image into N clusters,
The largest cluster in the divided N clusters is selected as the seed cluster,
Wherein the color similarity between the representative proximity of the seed cluster and the neighboring cluster and the representative color of the seed cluster and the representative color of the neighboring cluster are measured for the selected seed cluster and the remaining populations, , The two clusters are merged,
Transforming the merged cluster data into a two-dimensional color distribution diagram of RG / GB / BR, generating ellipses based on the cluster data of the respective color distribution diagrams,
(10), wherein the raw material is classified into good and defective products based on whether the raw material is contained in the generated ellipse.
And a window region is specified based on an arbitrary starting pixel in the extracted foreground map image. The automatic raw material sorting apparatus (10) according to claim 10, wherein the window region is specified based on an arbitrary starting pixel in the extracted foreground map image.
Wherein the color of the starting pixel is determined as a reference color value and data similar to the starting pixel is found in the designated window area. The automatic material sorting method using the mean-shift clustering and the stepwise merging method in a color image Device (10).
Wherein the coordinate center of the similar data is set as a new center of gravity and the average color of the similar data is defined as a new reference color value. Device (10).
Wherein the window region is redetermined based on the new center of gravity and data similar to the newly defined reference color value is searched in the designated window region. (10).
And the average color of the data obtained at the end is determined as a reference color value by repeating the operation until the center of gravity converges. An automatic material sorting apparatus using mean-shift clustering and stepwise merging method in a color image, (10).
The other pixels of the extracted foreground map image are also obtained by repeatedly performing the operations to obtain respective reference color values and dividing the foreground map image into N clusters based on the similarity between the reference color values. (10) which uses mean-shift clustering and stepwise merging method in color image.
Wherein the threshold value of the position proximity and the color similarity are predetermined so that the threshold value of the proximity and color similarity can be determined in advance.
Calculating a slope of the cluster data by calculating a linear regression function from the cluster data of the respective color distribution diagrams and calculating a height and a width of the cluster data of the respective color distribution diagrams to calculate a center point of data and a long axis value and a short axis value of the ellipse, (10) using the mean-shift clustering and the stepwise merging method in a color image.
The method according to claim 1, wherein the step of generating the ellipses comprises generating the ellipses including the cluster data by using the slope, the long axis value, the short axis value, and the center point. ).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR1020150168074A KR101753101B1 (en) | 2015-11-28 | 2015-11-28 | Automatic source classification method and apparatus using mean-shift clustering and stepwise merging in color image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR1020150168074A KR101753101B1 (en) | 2015-11-28 | 2015-11-28 | Automatic source classification method and apparatus using mean-shift clustering and stepwise merging in color image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| KR20170062661A true KR20170062661A (en) | 2017-06-08 |
| KR101753101B1 KR101753101B1 (en) | 2017-07-03 |
Family
ID=59220862
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| KR1020150168074A KR101753101B1 (en) | 2015-11-28 | 2015-11-28 | Automatic source classification method and apparatus using mean-shift clustering and stepwise merging in color image |
Country Status (1)
| Country | Link |
|---|---|
| KR (1) | KR101753101B1 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112287913A (en) * | 2020-12-25 | 2021-01-29 | 浙江渔生泰科技有限公司 | Intelligent supervisory system for fish video identification |
| KR20230072170A (en) * | 2021-11-17 | 2023-05-24 | (주)심스리얼리티 | Method for lighting 3D map medeling data |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4908926B2 (en) | 2006-05-29 | 2012-04-04 | 日清製粉株式会社 | Granule sorter |
| KR101451097B1 (en) | 2012-12-23 | 2014-10-15 | 전남대학교산학협력단 | System and method for image classification using color histogram |
| KR101324338B1 (en) | 2012-12-27 | 2013-10-31 | 충남대학교산학협력단 | A compact quality sorting method and apparatus for cherry tomatoes based on real-time color image processing |
-
2015
- 2015-11-28 KR KR1020150168074A patent/KR101753101B1/en active IP Right Grant
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112287913A (en) * | 2020-12-25 | 2021-01-29 | 浙江渔生泰科技有限公司 | Intelligent supervisory system for fish video identification |
| KR20230072170A (en) * | 2021-11-17 | 2023-05-24 | (주)심스리얼리티 | Method for lighting 3D map medeling data |
Also Published As
| Publication number | Publication date |
|---|---|
| KR101753101B1 (en) | 2017-07-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106651872B (en) | Pavement crack identification method and system based on Prewitt operator | |
| US11282185B2 (en) | Information processing device, information processing method, and storage medium | |
| US8294794B2 (en) | Shadow removal in an image captured by a vehicle-based camera for clear path detection | |
| JP5458905B2 (en) | Apparatus and method for detecting shadow in image | |
| JP5432714B2 (en) | Composition analysis method, image apparatus having composition analysis function, composition analysis program, and computer-readable recording medium | |
| US20130326417A1 (en) | Textual attribute-based image categorization and search | |
| JP6497579B2 (en) | Image composition system, image composition method, image composition program | |
| JP5549345B2 (en) | Sky detection apparatus and method used in image acquisition apparatus | |
| JP6529315B2 (en) | Main subject detection method, main subject detection device and program | |
| JP2010055194A (en) | Image processing device and method, learning device and method, and program | |
| US20100040276A1 (en) | Method and apparatus for determining a cell contour of a cell | |
| CN104915943B (en) | Method and apparatus for determining main parallax value in disparity map | |
| AU2015324988A1 (en) | Method and system for unsupervised image segmentation using a trained quality metric | |
| JP2007072628A (en) | Face direction discriminating device | |
| CN104597057A (en) | Columnar diode surface defect detection device based on machine vision | |
| KR101753101B1 (en) | Automatic source classification method and apparatus using mean-shift clustering and stepwise merging in color image | |
| CN112001853A (en) | Image processing apparatus, image processing method, image capturing apparatus, and storage medium | |
| US9235773B2 (en) | Image processing device capable of determining types of images accurately | |
| US8830529B2 (en) | Image processing device for accurately identifying region in image without increase in memory requirement | |
| KR102010086B1 (en) | Method and apparatus for phase segmentation of microstructure | |
| KR20240019361A (en) | Methods, measuring devices, welding systems, and programs for measuring the behavior of welding phenomena | |
| JP5172995B2 (en) | Face orientation discrimination device and computer program | |
| Kim et al. | Automatic classification algorithm for raw materials using mean shift clustering and stepwise region merging in color | |
| KR101517805B1 (en) | Object Detection Method with Image Value based Division Scheme and Image Processing Apparatus using the same | |
| JP5339147B2 (en) | Method for dividing texture regions with different grain sizes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| A201 | Request for examination | ||
| E902 | Notification of reason for refusal | ||
| E701 | Decision to grant or registration of patent right | ||
| GRNT | Written decision to grant |