CN103632157B - Individual plant wheat fringe portion kernal number method of counting - Google Patents

Abstract

The invention discloses a method for counting the number of grains on ears of wheat. The method comprises the steps of: S1. collecting a front view of wheat ears by using a CCD camera, and preprocessing the collected front views; S2. mathematical morphology analysis: The opening operation in the mathematical morphology processing is used to reasonably remove the small parts of the structural elements; the filling operation is performed on each cavity, and all objects smaller than the area P are removed by the area method; Correlation between the relevant characteristic parameters of the ear and the number of grains, extracting the area characteristics and cob length characteristics of the wheat ear; S4. Establishing a mathematical model: using the least square method to establish the characteristic parameters of the ear area and cob length and the number of grains of the wheat ear Mathematical model among them; S5. Mathematical model test: To verify the correlation between wheat ear area characteristics and long axis characteristics and its grain number, to ensure the practicability of these two models. The method of the present invention can establish a mathematical model between the area feature of the ear image, the length feature of the major axis and the number of grains by extracting the area feature and the length feature of the major axis of the wheat ear image, thereby being able to count the ears of wheat more accurately, rapidly and objectively The number of grains, compared with manual operation, greatly improves the efficiency of ear grain counting.

Description

Counting Method of Grain Number in Spike of Single Wheat Plant

technical field

The invention relates to a method for counting the number of grains at the panicle of a single wheat plant by using digital image processing technology.

Background technique

The number of grains in the ear of wheat is the number of full grains in the ear when the wheat is mature.

The final yield in the wheat production process is always the most concerned thing for agricultural producers. In fact, in order to facilitate the post-harvest processing and management of wheat, it is always an urgent need for producers and agricultural management departments to know the wheat yield as early as possible and estimate the yield. Therefore, the estimation of wheat field yield needs to accurately know the specific number of grains on the ear when the wheat grains are mature. When agronomic breeding and cultivation experts conduct laboratory planting tests, they also hope to quickly know the number of wheat spikes and accelerate wheat breeding. On the basis of counting the number of grains on the ears of wheat, according to the thousand-grain weight of this variety and other relevant parameters, the yield of field wheat can be estimated quickly and accurately.

The agricultural production measurement methods at this stage are as follows: sample area and survey content. When measuring yield in theory, take 1 square meter at each point to investigate the number of spikes per mu; randomly pick 20 spikes from each sampling point to investigate the number of grains per spike; the thousand-grain weight is calculated according to the annual thousand-grain weight of the variety.

The theoretical yield is calculated according to the following formula: Yield per mu (kg/mu) = number of ears per mu × number of grains per ear × thousand grain weight × 0.85.

It can be seen that the number of grains on the ear of wheat is one of the important basic steps in the measurement of theoretical yield. At present, the number of grains on the ear of wheat is obtained by manual counting, so as to obtain the yield of wheat. This method is time-consuming and laborious. There are also certain errors in the counting process, and different production measurement technicians may obtain different results, which is subject to greater influence and poor accuracy.

Therefore it is of practical significance to develop a kind of fast and accurate machine counting wheat ear grain number, and to reach the machine counting wheat ear grain number, at first need to determine the method that wheat ear grain number carries out automatic counting, in obtaining wheat ear grain number accurately On the basis of the number of wheat grains, the yield per mu of wheat is obtained according to the thousand-grain weight of wheat grains and the number of ears per mu. The number of grains on the ears of wheat is related to characteristic parameters such as cob length and ear area, which are characteristic of wheat ears. Therefore, the present invention has a theoretical basis for obtaining the number of grains on ears of wheat through digital image processing and counting.

Contents of the invention

(1) Technical problems to be solved

The technical problem to be solved in the present invention is: provide a kind of method that utilizes digital image processing technology to count out the grain number of ear of wheat, calculate the area and cob length of the area where grain of wheat ear is located by the method of digital image processing, establish ear ear Correlation between grain area and cob length with its number of grains in the panicle.

(2) Technical solution

To achieve the above object, the invention provides a method for counting the number of grains in ears of wheat, the method comprising steps:

S1. Utilize the CCD camera to collect the front view of the ear of wheat, and preprocess the collected front view;

The specific steps of S1 include: image acquisition, grayscale processing, and binarization processing, wherein the preprocessing includes two steps of grayscale processing and binarization processing.

S1.1 The image acquisition mainly uses the CCD camera to collect the front view of the ear of wheat. The devices for collecting the image of the ear of wheat include a CCD camera, a stage, a computer, etc., and the lighting conditions are natural light. Wherein the table top of the stage is a white bottom plate, and the CCD camera is suspended above the stage through the frame, so that the frame connects the stage and the CCD camera into one body. Under natural light, a single ear of wheat was placed on the stage according to the front side, and the table top of the stage was a white plate with uniform color, and the CCD camera was taken from the upper part at the same angle to take pictures, and the images were collected at the same light intensity and at the same angle ( Vertical), the same focal length, the same height.

S1.2 Grayscale processing: grayscale processing is performed on the color image to obtain a grayscale image of wheat ears, and a histogram of the image is obtained from the grayscale image.

S1.3 Binarization processing: Determine the threshold value of the gray distribution of the background area and the wheat ear area through the histogram, and use the Otsu method to select the threshold that maximizes the variance between classes, so as to separate the wheat ear area from the background area and obtain the wheat ear area. An image in which the spike is completely separated from the background area.

S2. Mathematical Morphological Analysis: Perform mathematical morphological processing on the preprocessed front view image, and use the open operation in the mathematical morphological analysis to reasonably remove the small parts of the structural elements; perform filling operations on each cavity, and use the area method Remove all objects smaller than the area P;

The specific steps of S2 include: combined operations of dilation and erosion, image filling and desmearing

S2.1 Combination operation of dilation and erosion: The opening operation is used in the image processing process (the operation first erodes the image, and then dilates the eroded image, and the two operations use the same structural element) to process the image.

After the mathematical morphology processing (opening operation), the awns or noise points in the image are significantly reduced, and the parts smaller than the structural elements have been reasonably removed.

S2.2 Image filling and decontamination processing: perform filling operation on the binary image. For binary images, change the connected background pixel 0 value to foreground pixel 1 value until the boundary of the object is reached. An important application of the filling operation is to perform a hole-filling operation on an image, and perform a filling operation on each hole. At the same time, use the area method to remove redundant objects, and remove all objects smaller than the area P. In this image processing process, the binary image after the opening operation is filled and decontaminated.

The processed image is clear and clear. Only the part of the wheat ear, that is, the wheat awns and miscellaneous points, have been reasonably removed, which provides favorable conditions for the extraction of subsequent characteristic parameters.

S3. Wheat spike characteristic parameter extraction: Wheat spike image characteristic parameter is the basic description of the basic characteristics of wheat spike image, and is the theoretical basis for establishing a mathematical model between wheat spike-related characteristic parameters and grain number. The present invention is only for The area feature of wheat spike and the length feature of cob are extracted.

The specific steps for extracting characteristic parameters of wheat spikes in S3 include: extraction of characteristic parameters of wheat spike area and feature extraction of wheat spike cob length

S3.1 Extraction of area feature parameters: the filled and de-impured binary image is the area where only the wheat ear part has a pixel value of 1, and the area of the wheat ear part is the number of pixels with a pixel value of 1 calculated by image processing technology .

S3.2 Feature extraction of cob length: the cob length of the wheat ear image is the longest line segment that can be obtained by connecting two points on the ellipse.

S4. Establishing a mathematical model: using the least square method to establish a mathematical model between the characteristic parameters of the ear of wheat and the number of grains.

S4 establishes the mathematical model between the characteristic parameters of wheat ears and the number of grains, including steps: the mathematical model between the characteristic parameters of wheat ear area and the number of grains on the actual ear, the characteristic parameter of the length of the cob of wheat ear and the number of grains on the actual ear between mathematical models.

S4.1 Establish the correlation between the area characteristic parameters and the number of grains: establish a mathematical model between the area characteristic parameters obtained by image processing calculation and the actual number of grains on the ears of wheat by the least square method.

S4.2 Establish the correlation between the cob length parameter and its grain number: establish a mathematical model between the cob length characteristic parameter obtained by image processing calculation and the actual number of grains on the ear by the least square method.

S5. Mathematical model test: verify the mathematical model between wheat ear area characteristics and cob length characteristics and its grain number, ensuring the practicability of these two models.

The verification steps of the correlation mathematical model between the number of grains on the ear of wheat and its characteristics of S5 include: the verification of the mathematical model between the characteristics of the ear area of wheat and its number of grains, and the verification of the mathematical model between the length of the cob length of the ear of wheat and its number of grains.

S5.1 Test the model between the wheat ear area characteristic parameters and the number of grains: use the number of grains of the test sample as a standard to measure the number of grains in the ear of wheat obtained through the mathematical model. A correlation curve was established between the actual number of grains in the verification wheat sample set obtained as the actual value and the predicted value of the number of grains in the ear of wheat obtained through the mathematical model, and the degree of difference between the two was compared.

S5.2 Test the model between the characteristic parameters of the cob length of wheat cobs and the number of grains: use the same method as in S5.1 to test the model between the cob length of wheat cobs and the number of grains.

(3) Beneficial effects

The present invention grayscales and binarizes the front view image of the ear of wheat, and uses the mathematical morphology method to separate the awn and the ear of wheat, so as to obtain only the main part of the ear of wheat, so that it can continue to use the image processing method to extract Spike area characteristics and cob length characteristics. Through the extracted wheat ear area characteristics and cob length characteristic parameters, the correlation mathematical model between them and the number of grains obtained by manual counting was established, and the model verification method was further proved that the obtained mathematical model has high precision and can effectively replace manual counting .

Description of drawings

Figure 1 The working process of the counting system of wheat ears and grains

Figure 2 Collected front view image of wheat ears

Figure 3 Grayscale processing of color images

Figure 4 grayscale histogram

Figure 5 Grayscale image to binary image

Figure 6 The binary image after the combined operation of dilation and erosion

Figure 7 The image after filling and descrambling

detailed description

The wheat panicle grain counting method that the present invention proposes is described in detail as follows in conjunction with accompanying drawing and embodiment.

In order to quickly obtain the characteristic parameters of wheat spike area and cob length, establish the correlation between the area characteristic parameters, cob length characteristic parameters and the number of grains, so there are two key points, one is to quickly obtain the wheat spike area The other is to establish a mathematical model for these two characteristic parameters and the number of grains in the ear, so as to establish a foundation for quickly obtaining the number of grains in the ear of wheat.

According to shown in Fig. 1, according to the method that a kind of fast acquisition wheat panicle grain number counting of the present invention comprises steps:

S1. Utilize the CCD camera to collect the front view of the ear of wheat, and preprocess the collected front view;

The specific steps of S1 include: image acquisition, grayscale processing, and binarization processing

S1.1 Image acquisition: The image acquisition mainly uses the CCD camera to collect the front view of the ear of wheat. The devices for collecting the image of the ear of wheat include a CCD camera, a stage, a computer, etc., and the lighting conditions are natural light. Wherein the table top of the stage is a white bottom plate, and the CCD camera is suspended above the stage through the frame, so that the frame connects the stage and the CCD camera into one body. Under natural light, a single ear of wheat was placed on the stage according to the front side, and the table top of the stage was a white plate with uniform color, and the CCD camera was taken from the upper part at the same angle to take pictures, and the images were collected at the same light intensity and at the same angle ( Vertical), the same focal length, the same height. According to the above method, the front view image of wheat ear is obtained as shown in Fig. 2 .

The preprocessing of the front view image of wheat ears includes two steps of grayscale processing and binarization processing.

S1.2 Gray-scale processing: Gray-scale processing is performed on the color image to obtain a gray-scale image of wheat ears, and the front view image of wheat ears is processed by gray-scale processing to obtain the image shown in Figure 3, as can be seen from the figure The grayscale image effectively preserves the details of the ear area with clear edges. Get the histogram of the grayscale image from the grayscale image. The histogram of the grayscale image 3 is shown in Figure 4. In the histogram, the gray distribution characteristics of the wheat ear area and the background area are clearly distinguished and easy to separate.

S1.3 Binarization processing: Determine the threshold value of the gray distribution of the background area and the wheat ear area through the histogram, and use the Otsu method to select the threshold that maximizes the variance between classes, so as to separate the wheat ear area from the background area and obtain the wheat ear area. An image in which the spike is completely separated from the background area. The binarized image obtained after binarization processing is shown in Figure 5.

S2. Mathematical Morphological Analysis: Perform mathematical morphological processing on the preprocessed front view image, and use the open operation in the mathematical morphological analysis to reasonably remove the small parts of the structural elements; perform filling operations on each cavity, and use the area method All objects smaller than the area P are decontaminated; the edge information extraction processing based on mathematical morphology is better than the edge extraction algorithm based on differential operation, and the sensitivity to noise is low, and the extracted edge is relatively smooth. The morphological filter based on mathematical morphology can effectively filter out noise by using morphological operators with the help of prior geometric feature information, and can also retain the original information in the image.

The specific steps of S2 include: combined operations of dilation and erosion, image filling and decluttering

S2.1 Combination operations of dilation and erosion

In the process of image processing, open operation is used (the operation first corrodes the image, and then expands the corroded image, and the two operations use the same structural element) to process the image. The circular structural element (disk) is used in this processing, and the radius R is 7. The processing result is shown in Figure 6.

After the mathematical morphology processing (opening operation), the awns or miscellaneous points in the image are significantly reduced, and the parts smaller than the structural elements have been reasonably removed. At the same time, the ears and stalks of the ears of wheat are also removed after this operation.

S2.2 Image padding and decluttering

Perform padding operations on binary images. For binary images, change the connected background pixel 0 value to foreground pixel 1 value until the boundary of the object is reached. An important application of the filling operation is to perform a hole-filling operation on an image, and perform a filling operation on each hole. At the same time, use the area method to remove redundant objects, and remove all objects smaller than the area P. In this image processing process, the binary image after the opening operation is filled and decontaminated, and the processing results are shown in Figure 7.

The processed image is clear and clear. Only the part of the wheat ear, that is, the wheat awns and miscellaneous points, have been reasonably removed, which provides favorable conditions for the extraction of subsequent characteristic parameters.

S3. Wheat spike characteristic parameter extraction: Wheat spike image characteristic parameter is the basic description of the basic characteristics of wheat spike image, and is the theoretical basis for establishing a mathematical model between wheat spike-related characteristic parameters and grain number. The present invention is only for The area feature of wheat spike and the length feature of cob are extracted.

The specific steps for extracting characteristic parameters of wheat spikes in S3 include: calculation of wheat spike area feature quantities and extraction of wheat spike cob length features

S3.1 Extraction of area feature parameters: the filled and de-impured binary image is the area where only the wheat ear part has a pixel value of 1, and the area of the wheat ear part is the number of pixels with a pixel value of 1 calculated by image processing technology . Calculate the number of pixels inside the boundary as the area of the grain of its image.

S3.2 Extraction of cob length feature parameters: the cob length of the wheat ear image is the longest line segment that can be obtained by connecting two points on the ellipse.

S4. Establishing a mathematical model: establishing a mathematical model to establish a mathematical model between the characteristic parameters of the wheat ear image and the number of grains at the time using the least square method.

S4.1 Establishing the correlation between the area characteristic parameters and the number of grains: establishing a mathematical model between the area characteristic parameters calculated by image processing and the actual number of grains on the ears of wheat by the least square method.

In order to count the number of grains on ears of wheat, this method recognizes the size of the image area of ears of wheat of the same variety, establishes the relationship between the number of ears of ears of wheat and the area of the image, and uses 100 wheat ears as samples for establishing a mathematical model. Through this method, the correlation linear model between the characteristic parameters of the wheat ear image area and the number of wheat grains is established as follows: y=-1.7871+0.0007x, R=0.9583.

S4.2 Establish the correlation between the cob length parameter and its grain number: establish a mathematical model between the cob length characteristic parameter obtained by image processing calculation and the actual number of grains on the ear by the least square method.

In order to count the number of grains on the ears of wheat, this method uses the identification of the cob length of the image of the same variety of wheat ears, and establishes the relationship between the number of grains of wheat ears and the length of the cob of the image, and uses 100 wheat ears as samples for establishing a mathematical model . Through this method, the correlation coefficient R between the characteristic parameters of the cob length of wheat cobs and the number of grains is 0.9474, and the mathematical model expression is Y=0.1089X-24.958, which shows that the length of the cob cobs of wheat has a very significant linear relationship with the number of grains .

S5. Mathematical model test: verify the mathematical model between wheat ear area characteristics and cob length characteristics and its grain number, ensuring the practicability of these two models.

S5.1 The verification steps of the mathematical model of the correlation between the number of wheat spikes and its characteristics include: the verification of the mathematical model between the characteristics of the wheat spike area and its number of kernels, and the mathematical model between the characteristics of the cob length of the wheat spike and its number of kernels verify. In this invention, 40 verification samples (more than one-third of the test samples) are used to test the above mathematical model. The 40 samples and the test samples belong to the same variety, and the same image acquisition system is used. The method adopted for the model is the same.

S5.1 Test the model between wheat ear area and grain number

The correlation coefficient R between the actual value and the predicted value of wheat spikes is 0.9781, which is close to 1, indicating that the model can obtain the number of wheat spikes through digital image processing.

Wheat samples with errors between 12% and 16% only accounted for 2.50% of the total verification samples (that is, one sample), and most of the error rates of the 40 verification samples were concentrated in 0-4%, 4%-8%, and 8%. -12%, it can be seen from the figure that the wheat samples with an error rate below 12% account for 97.5% of the total verification samples.

S4.2 Test the model between the cob length of wheat ears and the number of grains

The same method was used to test the model between cob length and grain number in wheat.

The correlation coefficient R between the actual value and the predicted value of wheat spikes is 0.9501, which is close to 1, indicating that the model can obtain the number of wheat spikes through digital image processing.

Wheat samples with errors between 12% and 16% accounted for 12.50% of the total verification samples (that is, 5 samples), and most of the error rates of the 40 verification samples were concentrated in 0-4%, 4%-8%, and 8%. -12%, the analysis is available, and the wheat samples with an error rate below 12% account for 87.5% of the total verification samples. The percentage detected is lower relative to the previous model.

From the test results of the above two models, it can be seen that the model test results between wheat ear area and grain number are better than the model test results between wheat cob length and grain number.

The above method for counting the number of grains on spikes of wheat is aimed at the spikes of wheat with awns, and the method for counting the number of grains on spikes of wheat without awns is also applicable.

Claims (4)

1. A method for counting the number of grains in ears of wheat, the method comprising steps: S1. extracting the wheat ear image grain area characteristic parameter and cob length characteristic parameter after image processing to the front view image of the wheat ear part collected; S2. Establishing the mathematical model between the area of the ear image and the number of grains in the ear of wheat is: y=-1.7871+0.0007x, the correlation coefficient R=0.9583 between the characteristic parameters of the area of the ear image and the number of grains in the ear of wheat; The mathematical model of the length characteristic parameter and its grain quantity is y=0.1089x-24.958, and the correlation coefficient between the wheat cob length characteristic parameter and its grain quantity is R=0.9474; S3. The mathematical model between the wheat ear area characteristics and cob length characteristics and the number of grains was verified, and the correlation coefficient R between the actual value and the predicted value of the wheat ear area characteristic mathematical model was 0.9781 , the wheat samples with an error rate of less than 12% accounted for 97.5% of the total verification samples; the correlation coefficient R between the actual value and the predicted value of the wheat ear grain number in the mathematical model of wheat ear cob length characteristics was 0.9501, and the error rate was 12% Wheat samples below % accounted for 87.5% of the total verification samples. 2. wheat spike grain number counting method according to claim 1, is characterized in that, described spike image area feature parameter and cob length feature parameter extraction in step S1 comprises steps: S1.1 Extraction of area feature parameters of wheat ear image: after image processing of the collected front view image of wheat ear, a binary image with only part of the wheat ear grain area and the background area is obtained, and the number of pixels in the wheat ear grain area is calculated That is, the area feature quantity of the ear image; S1.2 Extraction of cob length characteristic parameters: after image processing, the front view image of wheat ears is processed to obtain a binary image with only part of the wheat ear grain area and the background area, and the cob length is connected to the ellipse through the wheat ear grain part area The length of the longest line segment that can be obtained from two points of . 3. wheat spike grain number counting method according to claim 1, is characterized in that, described in step S2 establishes the mathematical model between wheat spike characteristic parameter and its grain number and comprises steps: S2.1 Establish the correlation mathematical model between the characteristic parameters of the wheat ear image area and the number of grains: use the least squares method to establish a mathematical model between the area characteristic quantity calculated by image processing and the actual number of grains in the ear of wheat: y=-1.7871 +0.0007x, the correlation coefficient R=0.9583 between the characteristic parameters of the wheat ear image area and the number of wheat ear grains; S2.2 Establish the correlation mathematical model between the cob length characteristic parameters of the wheat ear image and the number of grains: use the least square method to establish a mathematical model between the cob length characteristic quantity obtained by image processing calculation and the actual wheat ear grain number: y=0.1089x-24.958, the correlation coefficient R=0.9474 between the characteristic parameter of cob length of wheat and its grain number. 4. the method for counting the number of grains on ears of wheat according to claim 1, is characterized in that, the verification mathematical model in step S3 comprises steps: S3.1 Test the model between the wheat ear area and the number of grains: establish a correlation curve between the actual number of grains of the wheat sample obtained and verified as the actual value and the number of wheat ear grains obtained through the mathematical model as the predicted value , and comparing the degree of difference between the two, the correlation coefficient R between the actual value and the predicted value of wheat panicle grain number is 0.9781, and the wheat samples with an error rate below 12% account for 97.5% of the total verification samples; S3.2 Test the model between the cob length of wheat cob and the number of grains: test the model between the cob length of wheat cob and the number of grains, and the correlation coefficient between the actual value and the predicted value of the number of grains in the ear of wheat R is 0.9501, and the wheat samples whose error rate is less than 12% account for 87.5% of the total verification samples. The verification ensures the practicability of the two models.