KR101334794B1 - Flower recognition method and apparatus using feature extraction - Google Patents

Abstract

A disclosed technique relates to a method and an apparatus for flower recognition using feature information. The flower recognition method which receives an image including a flower using a computer or a mobile terminal and recognizes the flower by extracting feature information from the image comprises the following steps of: inputting an area including a flower as an image by an image device; detecting a flower area from the image; extracting a first piece of feature information about color from the flower area; obtaining the center of mass of the flower, the outline of the flower, and the distance between the center of mass and the outline of the flower from the flower area, and extracting a second piece of feature information about the flower shape; extracting a third piece of feature information by applying an Euclidean Distance value between eight color values stored previously and color information about each pixel in the flower area; dividing the distance between the center of mass and the outline of the flower into three parts which are a first area, a second area and a third area from the center of mass of the flower, and extracting a fourth piece of feature information about each color of the first, second and third areas; and generating a recognition candidate model about the flower using the first, second, third and fourth pieces of feature information. As a result, the performance of recognizing similar flowers or various flowers can be enhanced by extracting morphological features of a flower. [Reference numerals] (110) Input an image;(120) Detect a flower area;(130) Extract first feature information;(140) Extract second feature information;(150) Extract third feature information;(160) Extract fourth feature information;(170) Generate a recognition candidate model;(AA) Start;(BB) End

Description

Flower Recognition Method and Apparatus Using Feature Extraction

The disclosed technology relates to an apparatus and method for recognizing flowers using feature information. More particularly, the present disclosure relates to a method and apparatus for accurately recognizing flowers by receiving an image captured by an imaging apparatus.

Analyzing the flower or leaf of a plant to recognize what kind of plant it is is a field studied in classical plant taxonomy. In other words, the type of plant is determined by analyzing the characteristic information according to the flower or leaf of the plant to find the most similar plant.

In determining the type of a plant, conventional techniques use a method of converting an input image into a gray image and extracting a feature of a flower outline. Since this method uses the shape information according to the contour of the flower without using the characteristics of various colors of the image, there is a problem that the recognition rate decreases as a result of the increase in the recognition rate for other plants similar to the plants to be recognized.

As a conventional technique for a method and apparatus for discriminating a plant using feature information, Korean Patent Application Publication No. 10-2009-0109727 (Name of the Invention: Object Recognition Method and Apparatus Using Point of Interest Designation and Contour Image in Image) and Korean Patent Publication No. 10-2010-0076795 (name of the invention: an apparatus and method for recognizing flowers in digital image processing).

The disclosed technology provides an apparatus and method for accurately recognizing flowers using feature information on color and shape.

In order to achieve the above technical problem, a first aspect of the disclosed technology is to receive an image including a flower using a computer or a portable terminal, extract feature information from the image, and recognize the flower. Inputting a region including the image as an image, detecting a flower region, extracting first feature information about color in the flower region, the center of gravity of the flower in the flower region, the outline of the flower, and the center of gravity of the flower Extracting the second feature information on the shape of the flower, and extracting the third feature information by applying Euclidean distance value between the eight pre-stored color values and the pixel color information of each pixel of the flower area. Step, dividing the distance from the center of gravity of the flower to the contour of the flower by dividing the distance from the center of gravity of the flower into the first area, the second area and the third area Extracting the fourth feature information for each color of the first region, the second region, and the third region, and flower recognition using the first feature information, the second feature information, the third feature information, and the fourth feature information. A method of generating a flower recognition candidate model using feature information including generating a candidate model is provided.

A second aspect of the disclosed technology to achieve the above technical problem is a method of receiving a flower including a flower using a computer or a portable terminal, and extracting feature information from the image to recognize the flower, An input unit for inputting a region to be included as an image, a first feature information extractor for extracting first feature information about color from a flower area, a flower center from a flower center, a flower outline, and a flower center from a flower area The second feature information extracting unit for extracting the distance to the second feature information about the shape of the flower, and applying the Euclidean distance value between the eight stored color values and the color information of each pixel of the flower area A third feature information extracting unit for extracting feature information, by dividing the distance from the center of the flower to the contour of the flower in three parts, the first region, A fourth feature information extracting unit, a first feature information, and a second feature information, each of which is divided into a second area and a third area, and extracts fourth feature information about colors of each of the first area, the second area, and the third area; The present invention provides a flower recognition candidate model generation apparatus using feature information including a candidate model generator for generating a recognition candidate model for a flower using third and fourth feature information.

Embodiments of the disclosed technique may have effects that include the following advantages. It should be understood, however, that the scope of the disclosed technology is not to be construed as limited thereby, since the embodiments of the disclosed technology are not meant to include all such embodiments.

According to an embodiment of the disclosed technology, by extracting feature information using morphological features of a flower image, it provides an effect of improving the recognition performance for similar flowers and images of various flowers. In addition, it provides an effect of minimizing the influence of light by using the feature and shape information on the color of the entire and segmented flower area.

1 is a flowchart of a method of generating a flower recognition candidate model using feature information according to an embodiment of the disclosed technology.
2 is a flowchart illustrating a flower recognition method using feature information according to another exemplary embodiment of the disclosed technology.
3 is a diagram illustrating eight basic colors used in an embodiment of the disclosed technology.
4 is a diagram for extracting feature information about a flower shape in an embodiment of the disclosed technology.
FIG. 5 is a diagram illustrating a process of dividing a flower into three pieces in one embodiment of the disclosed technology.
6 is a diagram illustrating flower image recognition performance of the disclosed technology.

The description of the disclosed technique is merely an example for structural or functional explanation and the scope of the disclosed technology should not be construed as being limited by the embodiments described in the text. That is, the embodiments are to be construed as being variously embodied and having various forms, so that the scope of the disclosed technology should be understood to include equivalents capable of realizing technical ideas.

Meanwhile, the meaning of the terms described in the present application should be understood as follows.

The terms "first" and "second" are intended to distinguish one component from other components, and the scope of rights should not be limited by these terms. For example, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component.

It is to be understood that when an element is referred to as being "connected" to another element, it may be directly connected to the other element, but there may be other elements in between. On the other hand, when an element is referred to as being "directly connected" to another element, it should be understood that there are no other elements in between. On the other hand, other expressions describing the relationship between the components, ie, between, and between, "between" or "neighboring to," and "directly neighboring to" should be interpreted as well.

Singular expressions should be understood to include plural expressions unless the context clearly indicates otherwise, and terms such as "include" or "have" refer to features, numbers, steps, operations, components, parts, or parts thereof described. It is to be understood that the combination is intended to be present, but not to exclude in advance the possibility of the presence or addition of one or more other features or numbers, steps, operations, components, parts or combinations thereof.

Each step may occur differently from the stated order unless the context clearly dictates the specific order. That is, each step may occur in the same order as described, may be performed substantially concurrently, or may be performed in reverse order.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the disclosed technology belongs, unless otherwise defined. Terms such as those defined in the commonly used dictionaries should be construed to be consistent with the meanings in the context of the related art and should not be construed as having ideal or overly formal meanings unless expressly defined in this application. .

1 is a flowchart of a method of generating a flower recognition candidate model using feature information according to an embodiment of the disclosed technology. Referring to FIG. 1, in the method of generating a flower recognition candidate model using feature information, inputting an image in an imaging apparatus 110, detecting a flower region in an image 120, and first feature information in a flower region Extracting step 130, extracting the second feature information from the flower area 140, extracting the third feature information from the flower area 150, and extracting the fourth feature information from the flower area. And step 160 of generating a recognition candidate model for the flower. Hereinafter, the technical features performed in each step will be described in detail.

In step 110, an image is input from the imaging apparatus. In one embodiment of the disclosed technology, an imaging device means a computer or a portable terminal. Of course, it will be understood by those skilled in the art that other electronic devices or photographing apparatuses capable of photographing flower images can be used.

That is, in step 110, an image is input using a computer, a portable terminal, an electronic device capable of photographing, or a photographing device. And it is preferable to include a flower in the input image so as not to violate the purpose of the disclosed technology. For example, in one embodiment of the disclosed technology, the image may be a picture or a video including a flower.

In operation 120, a flower region is detected from the input image. As a method of detecting a region of a flower, for example, an Intellgent Scissors algorithm or a watershed algorithm may be used in an image including a flower. Intellgent Scissors and Watershed are conventional image detection algorithms. Using such image detection algorithms, it is possible to detect areas of flowers in an image.

In operation 130, first feature information on color is extracted from the entire flower region included in the image. Before describing step 130, the entire area of the flower refers to an area including the entire flower, not an area including a part of the flower, in the image. The area in the image is a specific range that contains the flowers you want to recognize in the image. Therefore, in operation 130, first feature information on color is extracted from the entire flower area.

As described above, the first feature information is feature information about a color. That is, the first feature information is obtained by extracting an average value of RGB or YCbCr. Here, the average value of RGB includes the average of red, average of green, and average of blue in the flower area. And the average value of YCbCr includes the average of Y, the average of Cb and the average of Cr in the flower area. For a detailed description of the process of calculating the average value of RGB and the average value of YCbCr, the following description will be given with reference to Equations (1) and (2) below.

Average of Red  = Sum of Red  Of Flower pixel

Average of Green = Sum of Green / Flower pixel (One)

Average of Blue  = Sum of Blue  Of Flower pixel

Referring to equation (1) above Red Average (Average of of Red is the sum of the total reds in the flower area ( Sum of Red is the total number of pixels in the flower area ( Flower Divide by pixel and average of Red ( Average of Red ) is obtained. And the average (Average of Green) is the Green's (Sum of Green) the sum of the total number of pixels in the area of the flower Green flower area (Flower The average of Green is obtained by dividing by pixel ) and the average of Blue ( Average of Blue) is the sum of the Blue (Sum of Blue is the total number of pixels in the flower area ( Flower Divide with pixel and average of Blue ( Average of Blue ).

Average of  Y = Sum of  Y / Flower pixel

Average of Cb = Sum of Cb  Of Flower pixel                      (2)

Average of Cr  = Sum of Cr  Of Flower pixel

Meanwhile, referring to the above formula (2), the average of Y ( Average of Y) Y is the sum (Sum of of Y is the total number of pixels in the flower area ( Flower Divide by pixel and average of Y ( Average of Y ). And the average (Average of Cb of Cb ) is the sum of Cb ( Sum of Cb ) is the total number of pixels in the flower area ( Flower Divide by pixel and average of Cb ( Average of Cb ), and the average of Cr ( Average of Cr) is the sum of Cr (Sum of Cr ) is the total number of pixels in the flower area ( Flower pixel) and by dividing the average of Cr (Average of Cr ).

Each of the six average values extracted as above is used as feature information about the flower area. In the case of the conventional technologies disclosed above, since the feature information for the various colors is not used as described above, only the feature information about the flower shape included in the image and the judgment of the flower using the color are used, so that the recognition rate of the flower is lowered. There was this. However, in the disclosed technology, all feature information extracted by extracting feature information about a shape and extracting feature information on a shape is used, thereby providing a high recognition rate.

In operation 140, the second feature information on the shape of the flower is extracted. Before extracting the feature information on the shape, in step 140, the center of gravity and the contour of the flower and the distance from the center of gravity to the flower's contour are obtained. In obtaining the center of gravity, one embodiment of the disclosed technique uses the following equation (1).

및

은 해당 픽셀의 좌표값이다. 이 수학식 1을 이용하면 영상이 포함하는 꽃 영역에서 꽃의 무게중심을 계산하는 것이 가능하다.In Equation 1, C is the center of gravity, N represents the number of pixels of the flower area,

And

Is the coordinate value of the pixel. Using Equation 1, it is possible to calculate the center of gravity of the flower in the flower area included in the image.

In step 140, the distance from the center of gravity to the contour of the flower is calculated along with the center of gravity of the flower. The distance information from the center of gravity of the flower to the contour of the flower is needed to obtain the second characteristic information about the shape of the flower. That is, the second characteristic information, which is characteristic information on the shape of the flower, is obtained by using the distance information from the center of gravity and the center of gravity of the flower to the outline of the flower. The second characteristic information includes information on the number of petals of the flower. In an embodiment of the disclosed technique, Equation 2 below is used to calculate the number of petals.

내에서 객체 영상의 무게중심으로부터의 윤곽선의 거리가 기준점을 지나는 양의 부호를 판단하며 이는 곧 꽃 객체의 꽃잎 수를 판별할 수 있는 중요한 특징으로 사용된다. 여기서

는 꽃 영상의 무게중심으로부터 윤곽선까지의 거리이며,

은 평균거리를 의미한다. 개시된 기술의 일 실시예에서는 영교차율에 따라 꽃잎 수를 계산하여 제 2 특징정보로 이용한다.In Equation 2, the reference value of Zero Crossing Rate is an average distance value of an individual distance from the center of gravity of the object image as a zero point, and the entire circumference length of the extracted contour line is extracted.

The distance of the contour from the center of gravity of the object image determines the positive sign that passes the reference point, which is used as an important feature to determine the number of petals of the flower object. here

Is the distance from the center of gravity of the flower image to the outline,

Means the average distance. In one embodiment of the disclosed technology, the number of petals is calculated and used as the second feature information according to the zero crossing rate.

In step 150, Euclidean distance values are applied between the eight stored color values and the pixel color information of each flower area. The eight color values here are white, yellow, cyan, green, magenta, red, blue, and black. The Euclidean Distance value is the distance between two vectors or two points in n-dimensional space.

In operation 160, the area of the flower may be divided into three parts to extract first and third feature information about each color of the divided areas. In step 160, the flower area is divided into three areas by using the center of gravity of the flower obtained in step 140 and the average distance information to the contour of the flower, and divided into a first area, a second area, and a third area, respectively.

In one embodiment of the disclosed technology, the first, second, and third regions are divided into three regions in the order close to the center of gravity, but the first, second, and third numbers like these simply distinguish each region. It is only a number, and three or less and more are possible, and the order of the numbers is not interpreted in a sequential manner. That is, there is a possibility that the number of regions and the order of the first region, the second region, and the third region are sufficiently changed according to the intention of the user. Those skilled in the art will appreciate this point.

On the other hand, since the area of the flower is divided into three divisions in step 160, it is preferable to extract feature information about each color. Therefore, the fourth feature information includes feature information about the color of the first region, feature information about the color of the second region, and feature information about the color of the third region.

Therefore, in step 160, the average value of red, green, and blue, and the average value of Y, Cb, and Cr of the first region are obtained as the fourth feature information, and eight pre-stored color values of white, yellow, cyan, green, magenta, and red are obtained. Euclidean distance values are used between the colors of blue, black and black and the pixel color information of the flower area. The average values of red, green, and blue, and Y, Cb, and Cr of the second area are obtained, and the color and flower areas of the eight color values, white, yellow, cyan, green, magenta, red, blue, and black, respectively. Euclidean distance is used between pixel color information. The average values of the red, green, and blue colors, the average values of the Y, Cb, and Cr of the third area are obtained, and the color and flower areas of the eight color values, white, yellow, cyan, green, magenta, red, blue, and black, respectively. Euclidean distance is used between pixel color information. That is, 14 pieces of feature information are extracted for each region, and a total of 42 pieces of feature information is extracted as the fourth feature information.

In step 170, a candidate model is generated to recognize flowers. In the process of recognizing a flower, one embodiment of the disclosed technique uses a recognition candidate model. In step 170, the first feature information, the second feature information, the third feature information, and the fourth feature information extracted in steps 130, 140, 150, and 160 are used as one of the existing pattern recognition algorithms. To generate a recognition candidate model.

2 is a flowchart of a flower recognition apparatus using feature information according to an embodiment of the disclosed technology. Referring to FIG. 2, the flower recognition apparatus using the feature information includes an input unit 210 for inputting an image in an image device, a detector 220 for detecting a flower region in an image, and a first feature for extracting first feature information from an image. The information extractor 230, the second feature information extractor 240 to extract the second feature information from the image, the third feature information extractor 250 to extract the third feature information from the image, and the fourth feature from the image. And a candidate model generator 280 using a fourth feature information extractor 260 for extracting information and a candidate model DB 270 for storing at least one recognition candidate model. Hereinafter, a detailed description of the technical features of each component.

The input unit 210 receives an image including a flower from the image device. As described above with reference to FIG. 1, an embodiment of the disclosed technology, an imaging device in the embodiment of the disclosed technology means a computer or a portable terminal. In addition, as described above, the fact that other electronic devices or photographing apparatuses capable of photographing can be used will be fully understood by those skilled in the art.

On the other hand, the input unit 210 preferably inputs an image including the entire area of the flower. The input image is extracted by the first feature information extractor 230, the second feature information extractor 240, the third feature information extractor 250, and the fourth feature information via the detector 220 that detects the flower region. It is transmitted to the unit 260 to extract different feature information from each component. Hereinafter, the detector 220, the first feature information extractor 230, the second feature information extractor 240, the third feature information extractor 250, the fourth feature information extractor 260, and the candidate model are generated. The technical features performed in the unit 280 will be described in detail.

As described above with reference to FIG. 1, the detector 220 detects an image of a flower area using an image detection algorithm such as an Intellgent Scissors algorithm or a watershed algorithm from an image including the flower. Intellgent Scissors and Watershed are conventional algorithms used to detect images. Using this algorithm, it is possible to detect flowers that the image contains.

The first feature information extractor 230 extracts feature information about a color from the entire flower area. Hereinafter, the characteristic information about the color is referred to as first characteristic information. The first feature information extractor 230 extracts average values of RGB and YCbCr as first feature information. Here, the average value of RGB means that the average of red, average of green and average of blue are included in the entire flower area. And the mean value of YCbCr means to include the average of Y, the average of Cb and the average of Cr in the flower area. That is, the average value of RGB includes an average of red, an average of green, and an average of blue. And the average value of YCbCr includes the average of Y, the average of Cb and the average of Cr. The first feature information extractor 230 extracts feature information about the color of the flower area.

The second feature information extractor 240 extracts feature information about the shape of the flower. Hereinafter, the characteristic information on the shape of the flower is referred to as second characteristic information. In extracting the second feature information, the second feature information extracting unit 240 obtains the distance information from the center of gravity of the flower and the center of gravity of the flower to the contour, and extracts the feature information on the shape of the flower using them. There are various feature information about the shape of the flower, but since the disclosed technology aims to accurately recognize the flower, the distance from the center of gravity of the flower area to the flower area outline, which is one of the main feature information of the flower Extracting the information is used as the second feature information.

On the other hand, in calculating the distance information from the center of gravity of the petal number and the flower area to the contour of the flower area, one embodiment of the disclosed technique uses a zero crossing rate. The zero crossing rate is the number of times that the flower outline crosses the reference axis. That is, the distance information from the center of gravity of the flower area to the contour of the flower area is used for the feature, and the information determines the number of petals of the flower included in the flower area. Accordingly, the second feature information extracting unit 240 calculates the number of petals according to the distance information and the zero crossing rate from the flower area center of gravity to the flower area contour and uses it as the second feature information.

The third feature information extractor 250 applies third specific information to which Euclidean distance values are applied between eight pre-stored color values of white, yellow, cyan, green, magenta, red, blue, and black, and pixel color information of each flower area. Extract The Euclidean Distance value is the distance between two vectors or two points in n-dimensional space.

The fourth feature information extractor 260 divides the flower region by using the distance information from the flower center of gravity and the flower region weight center obtained by the second feature information extractor 240 to the flower region contour. First characteristic information and third characteristic information on color are extracted from each of the three divided regions. For detailed description of one embodiment of the disclosed technique, each divided area is referred to as a first region, a second region, and a third region. The fourth feature information extractor 260 extracts the first feature information and the third feature information from the first area. The first feature information and the third feature information are extracted from the second area, and the first feature information and the third feature information are extracted from the third area.

The first feature information and the third feature information extracted from the respective regions have the same characteristics in that they are feature information on colors, but the extracted values may be different because the ranges for extracting feature information are different. The fourth characteristic information is a value of the first characteristic information and the third characteristic information extracted from each of these areas.

The recognition candidate model DB 270 is used by the candidate model generator in an embodiment of the disclosed technology. The recognition candidate model DB 270 extracts and stores feature information about at least one flower that is a candidate model in advance. Therefore, the flowers photographed using the candidate model are recognized.

Meanwhile, as shown in FIG. 2, the candidate candidate model DB 270 may be configured separately by the candidate model generator 280. Of course, those skilled in the art that it can be used as a component included in the candidate model generator 280 will be fully understood.

The candidate model generator 280 recognizes flowers by measuring similarity of the input flower image feature information using candidate models stored in the recognition candidate model DB 270 in advance. Here, the candidate model generation unit 280 preferably compares the candidate information with feature information extracted from the feature of the input flower image and classifies the class into the highest similarity class to recognize the flower. In other words, it is desirable to compare the flowers to be recognized with a large number of candidate models having similar feature information in order to reduce the false recognition rate.

Meanwhile, the classification of the recognition candidate model DB and the photographed or photographed flower image feature information into a class having the highest similarity using Multi Class SVM, which is one of the existing pattern recognition algorithms, is a common knowledge in the art. If you have it, you will understand it.

In addition, in one embodiment of the disclosed technology it is possible to further provide the following components as shown in FIG. 290 is a result output unit for outputting a result according to the recognition. The result output unit 290 outputs the results of scientific name, plant name, related information, and the like of the recognized flower image. Therefore, in the related arts, it is possible to solve the problem of lowering the recognition rate due to recognizing a flower depending on whether the flower is judged according to the shape and color of the flower. For example, the result output unit 290 may be an image device such as a monitor or a screen, and may output the recognition result through the image device.

3 illustrates eight color values used in an embodiment of the disclosed technology. Referring to FIG. 3, eight color values previously stored in an embodiment of the disclosed technology are white, yellow, cyan, green, magenta, red, blue, and black, respectively. As shown in the figure, the RGB ranges from 0 to 255, respectively. In addition, referring to FIG. 3, Red, Green, and Blue of RGB each include different ranges from 0 to 255 for eight reference color values.

In one embodiment of the disclosed technique, Euclidean distance values are used between the above eight color values for the pixel colors of the entire flower area. In addition, each of the regions obtained by dividing the area of the flower into three uses first characteristic information and third characteristic information with respect to color. Therefore, in the related art, it is possible to provide better recognition performance compared to using only information on the shape of a flower.

4 is a diagram for extracting feature information about a flower shape in an embodiment of the disclosed technology. As described above with reference to FIGS. 1 and 2, in an embodiment of the disclosed technology, information is obtained from the center of gravity of a flower and the center of gravity of a flower to the contour of the flower, using the second information, which is characteristic information on the shape of the flower. Extract feature information. Referring to FIG. 4, the distance, which is the distance between the edge lengths of the flower outline and the center of gravity, may be obtained, and the average distance, which is the average of the distances, may be calculated.

The flower to be recognized in the disclosed technology may not have a true appearance in its original form. In addition, the deviation may be greatly increased depending on where the edge length value of FIG. 4 is obtained according to the photographing angle and position. Therefore, in one embodiment of the disclosed technology, each of these distance values is obtained and the average distance thereof is used. The average distance is used as a criterion for dividing the flowers included in the image. The process of dividing the flower will be described later with reference to FIG. 5.

5 illustrates a process of dividing a flower in an embodiment of the disclosed technology. Referring to FIG. 5, in dividing a flower included in an image, it can be seen that the flower is divided into three regions based on the center of gravity. As described above with reference to FIG. 4, the distance between the center of gravity of the flower and the contour of the flower may vary depending on where the contour of the flower is held. Therefore, all distances are obtained and the average is divided into three.

On the other hand, in one embodiment of the disclosed technology, the flowers are divided into three by basically dividing the area into three, but it is possible to divide the area into three according to the setting value, not the third, according to the user's setting. Will be understood by those of ordinary skill in the art.

In an embodiment of the disclosed technology, feature information on color is extracted again in each of the divided regions. Therefore, it provides an effect of reducing the error due to the recognition of flowers.

6 is a diagram illustrating flower recognition performance of the disclosed technology. Referring to FIG. 6, a result of recognizing a flower included in an image in the disclosed technology may be seen. Experiments were performed to recognize flowers according to the technical characteristics of each step and components described above with reference to FIGS. 1 to 5. We performed 1092 recognition experiments on 30 different flower images.

The horizontal axis of FIG. 6 is different flowers input from the image. The vertical axis is previously stored candidate models by extracting feature information. As a result of the experiment, 1068 out of a total of 1092 were recognized and 24 failed. As a result, a high recognition success rate of about 97% was derived.

Flower recognition apparatus and method using the feature information through the embodiments of the disclosed technology has been described with reference to the embodiment shown in the drawings for clarity, but this is only an example, those skilled in the art It will be understood that various modifications and equivalent other embodiments are possible. Therefore, the true technical protection scope of the disclosed technology should be defined by the appended claims.

110: image input step 120: flower area detection step
130: extracting the first feature information 140: extracting the second feature information
150: third feature information extraction step 160: fourth feature information extraction step
170: generating a recognition candidate model
210: Image input 220: Flower area detection
230: Extract first feature information 240: Extract second feature information
250: extraction of the third feature information 260: extraction of the fourth feature information
270: Recognition candidate model DB 280: Similarity measurement
290: recognition result

Claims (14)

In the method for receiving an image including a flower using a computer or a portable terminal, extracting feature information from the image to recognize the flower,
Inputting an image including a flower in the imaging apparatus;
Detecting flower regions in the image;
Extracting first characteristic information about a color from the flower area;
Obtaining a distance from the center of gravity of the flower, the outline of the flower and the center of gravity of the flower to the outline of the flower in the flower area, and extracting second feature information on the shape of the flower;
Extracting third feature information by applying an Euclidean distance value between eight stored color values and color information of each pixel of the flower area;
Divide the distance from the center of gravity of the flower to the contour of the flower by dividing the distance into the first area, the second area and the third area from the center of gravity of the flower, and the first area, the second area and the third area. Extracting fourth feature information for each color of the area; And
Generating a recognition candidate model for the flower using the first feature information, the second feature information, the third feature information, and the fourth feature information. . The method of claim 1, wherein the first feature information,
An average value of RGB or YCbCr is extracted, and the RGB includes an average of Red, an average of Green and an average of Blue in the flower area, and the YCbCr includes an average of Y, an average of Cb, and an average of Cr in the area. Flower recognition method using the feature information. The method of claim 1, wherein the second characteristic information,
And calculating the number of petals of the flower using the center of gravity of the flower and the distance to the contour of the flower. The method of claim 1, wherein the center of gravity of the flower,
Flower recognition method using the feature information obtained according to the following equation.

(In the above formula

Represents the number of pixels, and

Indicates the coordinate value of the corresponding pixel.) delete The method of claim 1, wherein the pre-stored eight color values,
Flower recognition method using feature information of white, yellow, cyan, green, magenta, red, blue and black. The method of claim 1, wherein the generating of the recognition candidate model for the flower comprises:
Flower recognition method using feature information for generating at least one candidate model of the flower using a multi-class SVM, and measuring the similarity of the candidate models and the flower. An apparatus for receiving an image including a flower using a computer or a portable terminal, and extracting feature information from the image to recognize the flower,
An input unit for receiving an image including a flower from an image device;
A first feature information extracting unit which extracts first feature information about a color from a flower area of the image;
A second feature information extracting unit for obtaining a distance from the center of gravity of the flower, the outline of the flower and the center of gravity of the flower to the outline of the flower in the flower area, and extracting second feature information on the shape of the flower ;
A third feature information extracting unit extracting third feature information by applying an Euclidean distance value between eight stored color values and color information of each pixel of the flower area;
Divide the distance from the center of gravity of the flower to the contour of the flower by dividing the distance into the first area, the second area and the third area from the center of gravity of the flower, and the first area, the second area and the third area. A fourth feature information extraction unit for extracting fourth feature information for each color of the area; And
A flower recognition candidate model using feature information comprising a candidate model generator for generating a recognition candidate model for the flower using the first feature information, the second feature information, the third feature information, and the fourth feature information. Generating device. The method of claim 8, wherein the first feature information extracting unit,
Extracting an average value of RGB or YCbCr, the RGB includes an average of Red, an average of Green and an average of Blue in the flower area, wherein the YCbCr includes an average of Y, an average of Cb and an average of Cr in the area Flower recognition apparatus using the feature information. The method of claim 8, wherein the second feature information extracting unit,
Flower recognition apparatus using feature information for calculating the number of petals of the flower using the center of gravity of the flower and the distance to the contour of the flower. The method of claim 8, wherein the center of gravity of the flower,
A flower recognition device using feature information obtained according to the following equation.

(In the above formula

Represents the number of pixels in the region,

Indicates the coordinate value of the pixel.) delete The method of claim 9, wherein the pre-stored eight color values,
Flower recognition device using the characteristic information of white, yellow, cyan, green, magenta, red, blue and black. The method of claim 8, wherein the candidate model generator,
A flower recognition apparatus using feature information for generating at least one candidate model of the flower using a multi class SVM, and measuring similarity between the candidate models and the flower.

Images