KR20130133411A - Plant classification method and system using autorecognition of leaf image - Google Patents

Abstract

A plant classification method using automatic leaf image recognition according to the present invention comprises the following steps: obtaining a leaf image by an imaging device; extracting a distance parameter using the distance between the central point and the outline of the leaf in the leaf image; extracting a shape parameter using morphological features of the leaf in the leaf image; and determining the plant classification of the inputted leaf image by comparing the leaf image with recognition candidate leaf images based on the distance parameter and shape parameter. The distance parameter extraction step includes the following steps: detecting a leaf area in the leaf image; detecting the central point and the outline of the leaf in the leaf area; calculating distances from the central point of the leaf to all the pixels on the outline of the leaf; and extracting a distance parameter based on the calculated distances. [Reference numerals] (AA) Start;(BB) Input image;(CC) Detect a leaf area;(DD) Extract the outline and the central point of the leaf;(EE) Calculate the distance between the outline and the central point;(FF) Extract a feature parameter based on the distance;(GG) Extract a feature parameter based on morphological and geographical features;(HH) Recognize the plant (18 steps) using the leaf features;(II) End

Description

PLANT CLASSIFICATION METHOD AND SYSTEM USING AUTORECOGNITION OF LEAF IMAGE}

The present invention relates to a method and system for classifying plants using leaves images.

The method of classifying plants by analyzing the leaves of plants is a field studied in classical plant taxonomy. Recently, researches on how to classify plants by analyzing the leaves image have been conducted.

Searching for the shape of the leaves and comparing the similarity with the sample stored in the database, or as the Korean Patent Application Publication No. 2008-0022257 to extract the feature points of the leaf vein is trying to classify through the leaf vein.

However, in the prior art, classification is performed based only on the leaves appearance, or classification based only on the leaf veins, and thus, when the plant classification is not accurate and the leaf veins are not clearly detected, the classification is difficult.

The present invention seeks to provide a method and system for classifying plants using various parameters to overcome the problems of the prior art.

In the present invention, plants are classified through sequence-based feature parameters and frequency-domain feature parameters based on the distance between the center point and the outline in the leaves image. Furthermore, we will classify plants accurately using additional feature parameters based on the morphological features of leaves.

The present invention aims to improve the performance of plant recognition using various feature parameters, and to provide a classification method or system in which feature extraction of leaves and classification of plant types are automatically performed without user intervention.

The solution of the present invention is not limited to those mentioned above, and other solutions not mentioned can be clearly understood by those skilled in the art from the following description.

In order to solve the above problems, a plant classification method using automatic leaf image recognition according to an embodiment of the present invention includes obtaining an input leaf image through an image device, and extracting a distance parameter using a distance from a leaf center point to a leaf outline in an input leaf image. Becoming; Extracting a shape parameter from the input leaf image using the shape features of the leaves; And determining a plant classification of the input leaves image by comparing the recognition candidate leaves image based on a distance parameter and a shape parameter.

 The step of extracting a distance parameter may include detecting a leaf region in an input leaf image, detecting a leaf center point and a leaf outline in the leaf region, calculating a distance from the leaf center point to all pixels constituting the leaf outline, and A distance parameter is extracted based on the distance.

According to another aspect of the present invention, a plant classification system using automatic leaf image recognition may include: an image device unit obtaining an input leaf image, a database unit storing a recognition candidate leaf image, and a leaf region from an input leaf image obtained from the image device unit. Parameter extraction unit for extracting feature parameters using the leaf region detection unit for detecting, the distance from the leaf center point to the leaf outline and the morphological features of the leaves in the leaf region detected by the leaf region detection unit, and the feature parameter extracted in the parameter extraction unit And a classification unit for determining a plant classification of the input leaf image by comparing with the recognition candidate leaf image.

The distance parameter is a feature parameter that includes the mean of the distance, the standard deviation of the distance, and the ZCR of the distance, and the mean of the FFT values, the standard deviation of the FFT values, the number of samples with an FFT value above the mean, and the priority of the top N samples with large FFT values. Frequency characteristic parameters including rank, average of FFT phases, standard deviation of FFT phases, and ZCR of FFT phases, wherein the FFT is characterized by performing a Fast Fourier Transform (FFT) over distance.

The shape parameters include length-to-width ratio (Ratio = L / W), circle similarity (FormFactor = (4πA) / P2), rectangular shape similarity (Rectangularity = (LW) / A), length and circumference ratio (PLratio = L / P), length width and circumference ratio (PLWratio = (L + W) / P) and leaf vein width ratio (Vein ratio = A / VA).

The classification step consists of the mean of the distance, the standard deviation of the distance, the ZCR of the distance, the mean of the FFT values, the standard deviation of the FFT values, the number of samples with an FFT value above the mean, the average of the FFT phases, the standard deviation of the FFT phases, and the ZCR of the FFT phases. In order, and further compares the length and width ratio, circular similarity, rectangular similarity, length and circumference ratio, length width and circumference ratio, and leaf vein width ratio, and compares them to the input leaf image. Only one of these remains and when the classification is complete, no further comparison is performed.

The process of finally classifying the plant classification of the input leaf image may further include comparing the top N samples having a large FFT value for the recognition candidate leaf image with the top N samples having a large FFT value for the input leaf image. have. By comparing the order and rank of the N samples of the recognition candidate leaves image and the N samples of the input leaves image, the recognition candidate leaves image having the highest similarity is selected as the final plant classification.

The plant classification method and system using the automatic leaf image recognition according to the present invention can automatically process 16 leaf feature parameter extraction processes using the outline and center point of the leaf and the same from the acquired image, and derive one classification result. Provide convenience for classification of plant resources.

In addition, by using the feature parameter of the frequency domain together with the sequence-based feature parameter of the leaf image, it is possible to obtain an effect of improving the performance of the plant recognition result.

The effects of the present invention are not limited to those mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the following description.

1 is a schematic flowchart of a plant classification method using leaf image automatic recognition according to the present invention.
2 is a screen illustrating a step of detecting a leaf region in the present invention.
FIG. 3 is a graph of an example of a distance from a leaf center point to an outline and a value (Magnitude) of performing a Fourier transform (FFT).
4 is a diagram illustrating a method of detecting a leaf length and a leaf width.
5 is a screen illustrating a method of detecting leaf veins in leaves.
FIG. 6 is a diagram illustrating a process of comparing the top 10 samples having a large FFT value for a recognition candidate leaf image with the top 10 samples having a large FFT value for an input leaf image.
7 is a diagram illustrating an example of an entire process of determining a plant classification of an input leaves image.
8 is a block diagram showing a schematic configuration of a plant classification system using automatic leaf image recognition according to the present invention.
9 is a graph showing the results of performing classification on the leaves image in order to test the effect of the present invention.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.

The terms first, second, A, B, etc., may be used to describe various components, but the components are not limited by the terms, but may be used to distinguish one component from another . For example, without departing from the scope of the present invention, 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. And / or < / RTI > includes any combination of a plurality of related listed items or any of a plurality of related listed items.

As used herein, the singular " include "should be understood to include a plurality of representations unless the context clearly dictates otherwise, and the terms" comprises & , Parts or combinations thereof, and does not preclude the presence or addition of one or more other features, integers, steps, components, components, or combinations thereof.

Before describing the drawings in detail, it is to be clarified that the division of constituent parts in this specification is merely a division by main functions of each constituent part. That is, two or more constituent parts to be described below may be combined into one constituent part, or one constituent part may be divided into two or more functions according to functions that are more subdivided. Each of the components to be described below may additionally perform some or all of the functions of other components in addition to the main functions of the components, and some of the main functions of each of the components are different. Of course, it may be carried out exclusively by. Therefore, the existence of each component described through this specification should be interpreted functionally, and for this reason, the configuration of the components according to the plant classification system 100 using the leaf image automatic recognition of the present invention is an object of the present invention. It should be clear that it may be different from FIG. 8 to the extent that can be achieved.

Hereinafter, a plant classification method and system 100 using leaf image automatic recognition will be described in detail with reference to the accompanying drawings.

First, a plant classification method using leaf image automatic recognition according to the present invention will be described. In the plant classification method using the leaf image automatic recognition will be described the main configuration of the present invention.

1 is a schematic flowchart of a plant classification method using leaf image automatic recognition according to the present invention. The plant classification method using the leaf image automatic recognition according to the present invention is performed in a computer device, a portable device such as a smartphone, or a dedicated device for plant classification. Or it may be provided in the form of a program for execution in a computer device.

According to an embodiment of the present invention, a method of classifying plants using leaf image automatic recognition includes: obtaining an input leaf image through an image device, extracting a distance parameter using a distance from a leaf center point to a leaf outline in the input leaf image, and input leaf The method may include extracting a shape parameter from the image using the morphological features of the leaves and determining a plant classification of the input leaf image by comparing the recognition parameter with the recognition parameter based on the distance parameter and the shape parameter.

The step of extracting a distance parameter may include detecting a leaf region in an input leaf image, detecting a leaf center point and a leaf outline in the leaf region, calculating a distance from the leaf center point to all pixels constituting the leaf outline, and A distance parameter is extracted based on the distance.

Although the step of detecting the leaf region is included in the step of extracting the distance parameter, it can be described as a series of steps as shown in FIG. 1.

An imaging device from which an input leaves image is obtained corresponds to a camera device. A camera connected to a computer, a camera mounted on a portable device such as a smartphone, or a detachable camera device can be used. That is, a variety of devices capable of capturing an image and data can be used for the imaging device.

On the other hand, the image obtained by the imaging device is preferably photographed on a single color background, such as white to minimize the effect of the background when detecting the leaf area.

The image acquired by the imaging apparatus is then subjected to a constant transformation for parameter extraction. 2 is a screen illustrating a step of detecting a leaf region in the present invention.

In order to detect the leaf region from the input image, it is preferable to convert the input image into a gray scale image using Equation (1) below.

Here, RGB means red, green, and blue color values, respectively.

Thereafter, the converted gray scale image is converted into a binary image using Equation (2) below.

Here, f (x, y) represents the coordinates of the gray scale image, and T represents a threshold for binary image conversion. By the binary image conversion, the leaf region may be determined in the input image.

Furthermore, we extract the leaf center and outline for parameter extraction. In the present invention, in order to extract the feature parameters of the leaves, the leaves outline and center point (weight center) are first extracted. This is the preparation process to calculate the distance from the center point to the outline.

The leaf center point may be extracted by Equation (3) below using the detected binary image of the leaf.

Here, N is the number of pixels in the leaf area (width), and (x _n , y _n ) represents coordinate values of the pixel.

Meanwhile, various methods may be used to detect the outline in the leaf region image. For example, a leaf outline may be extracted using an edge detection technique such as Laplacian or Sobel.

The step of calculating the distance from the leaves center point to all the pixels constituting the leaves outline may be obtained by using Equation (4) below.

Where D (i) is the distance between the leaves center point and the i-th leaves outline pixel, C _x is the x coordinate of the leaves center point, C _y is the y coordinate of the leaves center point, and E (i) _x is the i-th leaves outline pixel Is the x coordinate of, and E (i) _y is the y coordinate of the ith leaf outline pixel.

When the distance from the leaf center point to each pixel constituting the outline is calculated, a feature parameter based on this distance is extracted. Since the distance means the distance from the center of the leaves to the outline. Feature parameters based on distance are called distance parameters.

The distance parameter includes three feature parameters including the mean of the distance, the standard deviation of the distance, and the ZCR of the distance. It is simply a mathematically manipulated parameter of distance. ZCR refers to the number of times each distance crosses the average value based on the average value.

In addition, the distance parameters are the mean of the FFT values, the standard deviation of the FFT values, the number of samples with an FFT value above the mean, the priority of the top N samples with the larger FFT values, the average of the FFT phases, the standard deviation of the FFT phases, and the ZCR of the FFT phases. It includes seven frequency characteristic parameters including. It is a parameter that processed the result of FFT (Fast Fourier Transform) on distance in various forms.

The priority of the top N samples in the FFT value (Magnitude) is to compare the values of the respective samples according to the FFT result, extract the N samples from the highest value, sort them in descending order, and then rank the samples according to the positions of the samples. it means.

FFT transforms sequence-based data into the frequency domain. By using the feature parameters in the frequency domain, it is possible to analyze the features of the leaves in more detail than when using only the sequence-based feature parameters. FIG. 3 is a graph of an example of a distance from a leaf center point to an outline and a value (Magnitude) of performing a Fourier transform (FFT).

Furthermore, the shape parameters are five feature parameters extracted using the length, width, width, and circumference of the leaf image, and include one feature parameter extracted using the feature of the leaf vein.

4 is a diagram illustrating a method of detecting a leaf length and a leaf width. After extracting the length of the leaves based on the longest distance between the center point of the leaves and all the pixels of the outline, the width of the leaves is extracted by extracting the pixels of the outline perpendicular to the length of the leaves based on the center point (weight center). Extract The width of the leaves is also the longest of the straight lines perpendicular to the length.

Specifically, the shape parameters include the length-to-width ratio, the form-factor similarity, the rectangular-type similarity, the length-to-circumference ratio, the length-to-circumference ratio, and the vein ratio. Are six parameters.

The length-to-width ratio is defined as in Equation 5 below. In the variables used below, L is the longest length of the leaves, W is the longest width orthogonal to L, A is the width of the leaves, P is the width of the leaves, and VA is the width of the leaf veins.

FormFactor and Rectangularity determine the degree to which the planar shape of the leaves is close to a circle or a right angle, respectively. Respective equations (6) and (7) are defined, respectively.

Further, the length and circumference ratio PLratio and the length width and circumference ratio PLWratio are defined by Equations (8) and (9), respectively.

Finally, the vein ratio is the ratio of the leaf vein width to the leaf vein width. The leaf vein area ratio is defined as in Equation (10) below.

5 is a screen illustrating a method of detecting leaf veins in leaves. In order to detect the leaf vein, the leaf image is converted into a gray scale image and an open operation is performed. Thereafter, a difference image between the gray scale image and the image on which the open operation is performed is converted into a binary image.

Now, it is necessary to compare the recognition candidate leaf image and the input leaf image stored in the database based on the extracted distance parameter and shape parameter. To be exact, the parameters of the recognition candidate leaves image are compared with those of the input leaves image. For this purpose, the parameter extraction performed on the input leaf image should be identically performed on the recognition candidate leaf image stored in the database. Alternatively, parameter extraction is performed and stored in the database.

The determining of the plant classification of the input leaf image may include determining the first classification step and classifying the shape parameter of the recognition candidate leaf image and the input leaf image by comparing the distance parameter of the recognition candidate leaf image stored in the database with the distance parameter of the input leaf image. The shape parameters may be classified into second classification stages which are classified by comparison.

The first classification step is the mean of the distance, the standard deviation of the distance, the ZCR of the distance, the mean of the FFT values, the standard deviation of the FFT values, the number of samples with an FFT value greater than the mean, the priority of the top N samples with large FFT values, and the FFT Compare the mean of the phases, the standard deviation of the FFT phase, and the ZCR of the FFT phase in that order.

The second classification step compares length-to-width ratio, circular similarity, rectangular similarity, length-to-circumference ratio, length-to-circumference ratio, and leaf vein width ratio in order.

Among the 16 parameters, the first classification step using the remaining parameters except the priority of the upper N samples having the large FFT value is the first classification step and the second classification step.

Overall 1) mean of distance => 2) standard deviation of distance => 3) ZCR of distance => 4) mean of FFT values => 5) standard deviation of FFT values => 6) number of samples with FFT values above average => 7) Average of FFT Phase => 8) Standard Deviation of FFT Phase => 9) ZCR of FFT Phase => 10) Length and Width Ratio => 11) Circle Shape Similarity => 12) Rectangular Shape Similarity => 13 ) Length and girth ratio => 14) Length and girth ratio => 15) Leaf vein width ratio.

In each comparison process, plants corresponding to the leaves image are classified, and only the types of plants classified at each stage perform the classification process of the next stage. If only one recognition candidate leaf image remains during this process, the input leaf image is classified as one remaining candidate leaf image.

Furthermore, if the classification is not performed in step 15, the remaining classification is performed using the priority of the upper N samples having the large FFT value.

That is, the determining of the plant classification of the input leaf image may further include comparing the top N samples having a large FFT value for the recognition candidate leaf image with the top N samples having a large FFT value for the input leaf image. have. The basic idea is to compare the order and rank of the N samples of the recognition candidate leaves image and the N samples of the input leaves image to select the recognition candidate leaves image having the highest similarity as the final plant classification.

Specifically, the final plant classification is performed in three steps described below.

1) Compare the order and rank of the top N samples of recognition candidate leaf images and the N samples extracted from the input leaf image and calculate a score. FIG. 6 illustrates a process of comparing the top 10 samples having a large FFT value (when N = 10) with respect to the recognition candidate leaves image as an example.

As shown in FIG. 6, for each of the 10 samples, 10 points for the same sequence number and rank, 9 points for the first rank difference, 8 points for the second rank difference, and 6 points for the fourth rank difference, The final score is calculated by summing up the 10 scores given and finally classified as the plant having the highest score.

2) When two or more plants have the highest score, the recognition result is obtained by using the number of samples having the same order and rank of the top 10 samples of the FFT value of the recognition candidate leaf images and the top 10 samples of the FFT value of the input leaf image. To derive. In other words, the number of samples having the same order and rank is ultimately classified as plants.

3) When there are two or more plants with the same order and rank of the top 10 samples, the number of samples having the same FFT value of the recognition candidate leaf images and the order of the top N samples of the FFT value of the input leaf image Deduce the recognition result by using. That is, the number of samples having the same number of samples is finally classified into plants.

7 is a diagram illustrating an example of an entire process of determining a plant classification of an input leaves image. FIG. 7 illustrates all three steps of comparing the above-described steps of step 15 with the upper N samples of the FFT value.

As each step is performed, candidate leaves that are determined to be less relevant are excluded, and finally, final classification is performed among candidates having a high compared score of the top N samples of the FFT value.

Another embodiment of the present invention is a computer-readable recording medium having recorded thereon a program for realizing the method for classifying plants using the automatic leaf image recognition. This corresponds to a medium on which a program executed on a computer device is recorded.

Yet another embodiment of the present invention is a plant classification system 100 using leaf image automatic recognition. 8 is a block diagram showing a schematic configuration of a plant classification system 100 using automatic leaf image recognition according to the present invention. The description of the plant classification method using the leaf image automatic recognition will be briefly described.

The plant classification system 100 using the automatic leaf image recognition according to the present invention includes an image device unit 110 for obtaining an input leaf image, a database unit 140 storing feature parameters of a recognition candidate leaf image, and an image device unit ( The leaf region detection unit 120 that detects the leaf region from the input leaf image acquired by 110, the distance from the leaf center point to the leaf outline in the leaf region detected by the leaf region detector 120, and the morphological characteristics of the leaf A parameter extractor 130 for extracting feature parameters and a classifier 150 for determining a plant classification of the input leaf image by comparing the feature parameter extracted from the parameter extractor 130 with feature parameters of the recognition candidate leaf image Include.

The plant classification system 100 using the leaf image automatic recognition according to the present invention may be implemented in the form of a computer device, or may be implemented through a smartphone and an application running on the smartphone.

The leaf region detector 120 detects the leaf region by converting the input leaf image acquired through the imaging apparatus into a gray scale and converting the image into a binary image.

The parameter extractor 130 detects the leaf center point and the leaf outline in the leaf area, calculates the distance from the leaf center point to all the pixels constituting the leaf outline, and extracts the distance parameter based on the distance. 131) and a shape parameter unit 132 for extracting shape parameters based on the length of the leaves, the width of the leaves, the width of the leaves, the circumference of the leaves, or the width of the leaf veins in the leaf region.

The distance parameter is a feature parameter that includes the mean of the distance, the standard deviation of the distance, the ZCR of the distance, and the mean of the FFT values, the standard deviation of the FFT values, the number of samples with an FFT value above the mean, and the priority of the top N samples with large FFT values. Frequency feature parameters including rank, average of FFT phases, standard deviation of FFT phases, and ZCR of FFT phases, where FFT is a Fast Fourier Transform (FFT) over distance.

The shape parameters include length-to-width ratio (Ratio = L / W), circle similarity (FormFactor = (4πA) / P2), rectangular shape similarity (Rectangularity = (LW) / A), length and circumference ratio (PLratio = L / P), length width and circumference ratio (PLWratio = (L + W) / P) and leaf vein width ratio (Vein ratio = A / VA).

The classifying unit 150 may determine the distance parameter as the mean of the distance, the standard deviation of the distance, the ZCR of the distance, the mean of the FFT values, the standard deviation of the FFT values, the number of samples having an FFT value greater than the mean, and the top N samples having a large FFT value. Priority, average of FFT phases, standard deviation of FFT phases, and ZCR of FFT phases are compared in order, and shape parameters are compared to length and width ratios, circle shape similarity, rectangular shape similarity, length and circumference ratio, length width and circumference ratio. And the leaf vein area ratio is compared in order, but only one recognition candidate leaf image compared to the input leaf image remains when the classification is completed, the comparison is not performed.

Meanwhile, the classification unit 150 may further include comparing the upper N samples having a large FFT value for the recognition candidate leaves image with the upper N samples having a large FFT value for the input leaves image. The classification unit 150 compares the order and rank of N samples of the recognition candidate leaves image and N samples of the input leaves image to select the recognition candidate leaves image having the highest similarity as the final plant classification.

In FIG. 9, the feature parameter of the recognition candidate leaf image stored in the classification unit and the database unit 140 is compared with the parameter of the input leaf image. It is preferable that the database 140 stores the feature parameter of the recognition candidate leaves image in advance. However, technically, the recognition candidate leaves image is stored in the database unit 140, and it is of course possible to separately extract and compare the parameters of the image.

9 is a graph showing the results of performing classification on the leaves image in order to test the effect of the present invention.

In order to evaluate the effectiveness and performance of the present invention, the researchers performed a plant classification method and recognition experiment using automatic leaf image recognition. A total of 1907 leaves images were collected over the Internet for the experiment.

 Extracting the features of the 16 leaves images using the collected leaves images and setting the range to the minimum-maximum value of the features of the leaves using the distribution of the features, the plant classification method and the database model using the leaves leaf image automatic recognition Generated.

Using the collected 1907 images, 18 kinds of plant classification and recognition experiments were performed. As a result of the experiment, 1820 leaves images were correctly recognized and showed 95.44% of plant recognition performance.

It is to be understood that both the foregoing general description and the following detailed description of the present invention are exemplary and explanatory and are intended to provide further explanation of the invention as claimed. It will be understood that variations and specific embodiments which may occur to those skilled in the art are included within the scope of the present invention.

100: Plant classification system using automatic leaf image recognition
110: image device unit 120: leaves region detection unit
130: parameter extraction unit 140: database unit
150: classification unit

Claims (18)

Obtaining an input leaf image through an image device;
Extracting a distance parameter using the distance from a leaf center point to a leaf outline in the input leaf image;
Extracting form parameters from the input leaf image using the shape features of the leaves; And
And a plant classification of the input leaf image is determined based on the distance parameter and the shape parameter based on the recognition candidate leaf image. The method of claim 1,
Extracting the distance parameter
Detecting a leaf region from the input leaf image;
Detecting the leaves center point and the leaves outline in the leaves area;
Calculating a distance from the leaf center point to all pixels constituting the leaf outline; And
And a distance parameter is extracted based on the distance. Plant classification method using automatic leaf image recognition. 3. The method of claim 2,
The step of detecting the leaf region is
The method of classifying plants using leaves image automatic recognition, characterized in that the input leaves image obtained through the imaging device is converted to a gray scale and then converted to a binary image. 3. The method of claim 2,
The step of calculating the distance is a plant classification method using automatic leaf image recognition, characterized in that performed using the following equation.

Where D (i) is the distance between the leaves center point and the i-th leaves outline pixel, C _x is the x coordinate of the leaves center point, C _y is the y coordinate of the leaves center point, and E (i) _x is the i-th leaves outline The x coordinate of the pixel, and E (i) _y is the y coordinate of the i-th leaf outline pixel) 3. The method of claim 2,
The distance parameter is
Feature parameters including the mean of the distance, the standard deviation of the distance and the ZCR of the distance;
Frequency including average of FFT values, standard deviation of FFT values, number of samples with FFT values above average, priority of top N samples with large FFT values, average of FFT phases, standard deviation of FFT phases, and ZCR of FFT phases And a feature parameter, wherein the FFT performs Fast Fourier Transform (FFT) on the distance. The method of claim 1,
The shape parameter is
Length to width ratio (Ratio = L / W), circle similarity (FormFactor = (4πA) / P2), rectangular shape similarity (Rectangularity = (LW) / A), length to circumference ratio (PLratio = L / P), A method for classifying plants using leaves image automatic recognition, comprising length width and circumference ratio (PLWratio = (L + W) / P) and leaf vein width ratio (Vein ratio = A / VA).
(Where L is the longest length of the leaves, W is the longest width orthogonal to L, A is the width of the leaves, P is the width of the leaves, and VA is the width of the leaf veins) The method of claim 1,
The distance parameters are the mean of the distance, the standard deviation of the distance, the ZCR of the distance, the average of the FFT values, the standard deviation of the FFT values, the number of samples with an FFT value greater than the average, the priority of the top N samples with large FFT values, and the FFT phase A frequency characteristic parameter comprising an average of, a standard deviation of the FFT phase, and a ZCR of the FFT phase, wherein the FFT is a Fast Fourier Transform (FFT) at the distance;
The shape parameters include length to width ratio (Ratio = L / W), circle similarity (FormFactor = (4πA) / P2), rectangular shape similarity (Rectangularity = (LW) / A), length and circumference ratio (PLratio = L / P), length width and circumference ratio (PLWratio = (L + W) / P) and leaf vein width ratio (Vein ratio = A / VA), where L is the longest length of the leaves and W is orthogonal to L The longest width, A is the width of the leaves, P is the circumference of the leaves, VA is the width of the leaf vein plant classification method using automatic leaf image recognition. The method of claim 7, wherein
The step of determining the plant classification of the input leaves image
A first classification step of classifying by comparing the distance parameter of the recognition candidate leaf image stored in a database with the distance parameter of the input leaf image; And
And a second classification step of classifying by comparing the shape parameter of the recognition candidate leaf image with the shape parameter of the input leaf image. 9. The method of claim 8,
The first classification step includes the mean of the distance, the standard deviation of the distance, the ZCR of the distance, the mean of the FFT values, the standard deviation of the FFT values, the number of samples with an FFT value greater than the mean, the mean of the FFT phases, the standard deviation of the FFT phases, and the FFT. Phase ZCR is compared in order, and the second classification step compares length and width ratios, circle similarity, rectangular shape similarity, length and circumference ratio, length width and circumference ratio, and leaf vein width ratio in order. 1. There is only one recognition candidate leaf image compared to the leaf image, and when the classification is completed, the comparison is not performed after the plant classification method using automatic leaf image recognition. 9. The method of claim 8,
The step of determining the plant classification of the input leaves image
Comparing the upper N samples having a large FFT value for the recognition candidate leaves image with the upper N samples having a large FFT value for the input leaves image, the N samples of the recognition candidate leaves image and the A method for classifying plants using leaves image recognition, characterized in that the recognition candidate leaves image having the highest similarity is selected as the final plant classification by comparing the order and rank of N samples of the input leaves image. The computer-readable recording medium which recorded the program for implementing the plant classification method using the leaf image automatic recognition of any one of Claims 1-10. An imaging device to obtain an input leaf image;
A database unit for storing the recognition candidate leaves image;
A leaf region detector for detecting a leaf region from the input leaf image acquired by the imaging apparatus;
A parameter extraction unit for extracting feature parameters using the distance from the center of the leaves to the leaves outline and the morphological features of the leaves in the leaves area detected by the leaves area detection unit;
And a classifier configured to determine a plant classification of the input leaves image based on the feature parameter extracted by the parameter extractor based on the feature parameter extracted from the parameter extractor. The method of claim 12,
And the leaf region detection unit detects the leaf region by converting the input leaf image obtained through the imaging apparatus into a gray scale and converting the image into a binary image. The method of claim 12,
The parameter extraction unit
A distance parameter unit for detecting the leaf center point and the leaf outline in the leaf area, calculating a distance from the leaf center point to all pixels constituting the leaf outline, and extracting a distance parameter based on the distance; And
And a shape parameter unit extracting shape parameters based on the length of the leaves, the width of the leaves, the width of the leaves, the circumference of the leaves, or the width of the leaf veins in the leaves region. 15. The method of claim 14,
The distance parameter is a feature parameter including the mean of the distance, the standard deviation of the distance, the ZCR of the distance and the mean of the FFT value, the standard deviation of the FFT value, the number of samples with an FFT value above the average, and the top N samples with a large FFT value. A frequency characteristic parameter including the priority of the FFT phase, the mean of the FFT phase, the standard deviation of the FFT phase, and the ZCR of the FFT phase, wherein the FFT is a Fast Fourier Transform (FFT) at the distance,
The shape parameters include length to width ratio (Ratio = L / W), circle similarity (FormFactor = (4πA) / P2), rectangular shape similarity (Rectangularity = (LW) / A), length and circumference ratio (PLratio = L / P), length width and circumference ratio (PLWratio = (L + W) / P) and leaf vein area ratio (Vein ratio = A / VA), characterized in that the plant classification system using automatic leaf image recognition.
(Where L is the longest length of the leaves, W is the longest width orthogonal to L, A is the width of the leaves, P is the width of the leaves, and VA is the width of the leaf veins) 16. The method of claim 15,
The classifier
The distance parameter of the recognition candidate leaves image stored in the database and the distance parameter of the input leaves image is compared, and the shape parameter of the recognition candidate leaves image and the shape parameter of the input leaves image is classified and classified automatically Plant classification system using recognition. 17. The method of claim 16,
The classifier
The distance parameter may be obtained by averaging distance, standard deviation of distance, ZCR of distance, average of FFT values, standard deviation of FFT values, number of samples with FFT values above average, average of FFT phases, standard deviation of FFT phases, and FFT phases. Compare the ZCRs in order,
The shape parameters are compared in order of length and width ratio, circle shape similarity, rectangular shape similarity, length and circumference ratio, length width and circumference ratio, and leaf vein width ratio, and only one recognition candidate leaf image is compared with the input leaves image. When the classification of the boy is completed, the comparison is not performed after the plant classification system using automatic leaf image recognition. 17. The method of claim 16,
The classifier
Comparing the upper N samples having a large FFT value for the recognition candidate leaves image and the upper N samples having a large FFT value for the input leaves image, the N samples of the recognition candidate leaves image and the A plant classification system using leaf image automatic recognition, characterized in that the recognition candidate leaf image having the highest similarity is selected as the final plant classification by comparing the order and rank of N samples of the input leaf image.

Images