CN103093456A - Corn ear character index computing method based on images - Google Patents

Abstract

The invention relates to image processing technology, and specifically discloses a calculation method for corn ear trait indexes. The method includes: acquiring color images CI _i (i=1, 2, 3, 4) of corn ears, and converting them into grayscale images SI _i and binary images BI _i ; and correcting axial and radial distortions to obtain Corrected color image CI2 _i , grayscale image HI _i and binary image BI2 _i ; image HI _i is segmented to obtain a binary image BI3 _i containing only corn kernels, and image BI3 _i is stitched to obtain a panoramic view of the grain BI4: Carry out morphological operations on the image BI3 _i to calculate the bald tip area and the length of the bald tip of the corn ear, calculate the number of rows of corn ears according to the image BI3 _i , and calculate the number of rows of corn ears according to the image BI4. The technical scheme of the invention restores the true shape of ears and grains by correcting axial and radial distortions, ensures the accuracy of parameter calculation, and provides an integrated solution for the calculation of various trait indexes of corn ears.

Description

Image-Based Calculation Method of Corn Ear Traits Index

technical field

The invention relates to the technical field of image processing, in particular to a method for calculating trait indexes of corn ears.

Background technique

The agronomic traits of corn ears are routine measurement items in scientific research and production, which have different degrees of influence on corn yield. In order to fully understand the degree of impact of quantitative traits on yield and to clarify the composition of various traits on yield, it is necessary to accurately measure a large number of ear samples. At present, the number of grains per ear, the number of grains per row, the number of rows of ears and the length of bald tip of corn ears are generally obtained by manual measurement and statistical methods. The process is cumbersome and the error is large, especially the number of And other statistically significant indicators, the measurement results are easily affected by the experience of the measurement personnel.

Precise corn ear seed testing based on images is a technical means that integrates computer vision technology and computer image processing technology to detect and analyze corn ears, and then quickly obtain characteristic data of ear traits. The method of seed testing based on ear images is simple, intuitive, convenient and fast, which can greatly reduce labor costs, improve the efficiency of corn seed testing, promote the precision and standardization of corn seed test procedures, and realize the manageability, traceability and visualization of the test process. The existing image-based method for calculating the trait characteristic parameters of corn ears is mainly aimed at a single trait index, and lacks a unified and integrated solution in the implementation method. For example, "Machine Vision Research on Counting the Number of Corn Ear Rows" (Han Zhongzhi et al., Corn Science, 2010) uses machine vision methods to automatically detect the ear row number index, and calculates the distribution characteristics of the corn ear row number by collecting cross-sectional images of corn ears. The disadvantage of the method is that destructive sampling is required on the ears, which destroys valuable samples and affects the determination of other ear trait parameters.

With the wide application of computer and information technology in the field of agriculture, automatic seed testing of corn ears based on image processing technology has become a trend. At present, there is still no complete calculation method for the important trait indicators of maize ears, including the number of kernels per ear, row number of ear, number of kernels per row, and bald tip rate. The existing image-based calculation methods for corn ear traits are often based on a single index, and there is a lack of an integrated solution that can simultaneously calculate the number of grains per ear, the number of rows of ears, the number of grains per row, and the rate of baldness.

Contents of the invention

(1) Technical problems to be solved

The purpose of the present invention is to provide an image-based calculation method for corn ear trait indicators, so as to solve the problems that the existing ear trait indicator calculation methods are single, only one of the indicators is calculated, and the sampling method is destructive.

(2) Technical solutions

In order to solve the problems of the technologies described above, the present invention proposes a method for calculating an image-based corn ear trait index, said method comprising the following steps:

S1. Obtain the color image CI _i of the ear of corn by taking an image every 90 degrees, and i=1,2,3,4,

converting the color image CI _i into a grayscale image SI _i , and converting the grayscale image SI _i into a binary image BI _i containing only ears of corn;

S2. Perform axial distortion correction and radial distortion correction respectively on the color image CI _i , the grayscale image SI _i and the binary image BI _i to obtain the corrected color image CI2 _i and the corrected grayscale image Degree image HI _i and corrected binary image BI2 _i ;

S3. Segment the grayscale image HI _i to obtain a binary image BI3 _i containing only corn kernels,

Splicing the binary images BI3 _i to obtain a grain panorama BI4, the grain panorama BI4 includes all grains on the three-dimensional surface of the corn ear;

S4. Performing morphological operations on the binary image BI3 _i , calculating the bald area of the corn ear, and calculating the bald length of the corn ear,

According to the binary image BI3 _i , calculate the number of rows of corn ears,

According to the grain panorama BI4, the ear row number of the corn ear is calculated.

Optionally, in step S1, converting the grayscale image SI _i into a binary image BI _i containing only ears of corn specifically includes:

Determine the intensity of the boundary pixel between the corn ear and the background in the grayscale image SI _i and binarize the grayscale image SI _i , fill the holes in the image using the dilation operation, delete the islands in the image and restore the corn ear by using the erosion operation , so as to obtain the binary image BI _i .

Optionally, in step S2, the axial distortion correction and the radial distortion correction specifically include:

Based on the binary image BI _i , establish the OBB bounding box of the ear of corn in each image, and establish the local coordinate system of the ear according to the OBB bounding box, calculate the axial distortion correction parameter and the radial distortion correction parameter, according to the The axial distortion correction parameter and the radial distortion correction parameter respectively correct the color image CI _i , the grayscale image SI _i and the binary image BI _i .

Optionally, step S3 further includes:

After the grayscale image HI _i is segmented, the segmented corn kernels are screened using the color and shape features of the corn kernels, so as to obtain the binary image BI3 _i , wherein,

The color feature of the corn kernel is obtained from the color image CI2 _i , and the shape feature of the corn kernel is obtained from the binary image BI2 _i .

Optionally, in step S3, splicing the binary image BI3 _i specifically includes:

Expand the binary image BI3 _i according to the three-dimensional surface of the ear, establish the mapping relationship between the same corn kernels on each image, and extract the corn kernels with complete structure from each image as the corn kernels on the kernel panorama BI4 .

Optionally, in step S4, performing morphological operations on the binary image BI3 _i specifically includes:

Perform an expansion operation on the binary image BI3 _i until all the corn grains in the image are fused to form a single-connected region, record the number of expanded pixels, and then restore the boundaries of the corn grains by using the corrosion operation, and then combine the obtained image with the The difference operation is performed on the binary image BI2 _i to obtain the area not containing corn kernels, and further obtain the bald area.

Optionally, in step S4, calculating the number of rows of corn ears specifically includes:

According to the coordinates of each corn kernel in the local coordinate system of the ear in the binary image BI3 _i , set up a point set S _i , set the starting point A _i and the end point B _i of counting the number of grains in a row, and set the starting point A _i and the end point B _i are added to the point set S _i ,

Connect the set of points S _i into a plane triangular grid T _i , and calculate the shortest path _{P i} from the starting point A _i to the end point B _i in the plane triangular grid T _i , the shortest path The number of grid nodes other than the starting point A _i and the end point B _i in P _i is the number of row grains of the binary image BI3 _i , and the average value of the number of row grains of the binary image BI3 _i That is the number of rows of grains of the corn ear.

Optionally, in step S4, calculating the number of ear rows of the corn ear specifically includes:

According to the grain panorama BI4, the number of grains per ear of the corn ear is calculated, and the quotient obtained by dividing the number of grains per ear by the number of grains per row is the number of rows per ear of the corn ear.

Optionally, in step S4, calculating the number of ear rows of the corn ear specifically includes:

According to the grain panorama BI4, calculate the circumference of the corn ear and the average width of the corn kernel, the quotient obtained by dividing the circumference by the average width of the corn kernel is the ear row number of the corn ear .

Optionally, in step S4, calculating the number of ear rows of the corn ear specifically includes:

According to the position of the center point of each corn kernel in the grain panorama BI4, establish a point set S', set the starting point A' and the end point B' of ear row number counting, and set the starting point A' and the end point B 'Added to the point set S',

Connect the point set S' into a plane triangular grid T', and calculate the shortest path P' from the starting point A' to the terminal point B' in the plane triangular grid T', the shortest path The number of grid nodes in P' except the start point A' and the end point B' is the number of ear rows of the corn ear.

(3) Beneficial effects

Compared with the prior art, the technical solution proposed by the present invention has the following advantages:

1. Based on the axial/radial distortion correction, the morphological characteristics of ears and kernels are restored, and the accuracy of the calculation method based on the color and shape of corn kernels is improved.

2. Based on the results of grain segmentation and splicing, it provides a new calculation method for the number of grains per ear, the length of the bald tip, the number of grains in a row and the number of rows in an ear, and provides an integrated solution for the calculation of the trait parameters of corn ears. The calculation of the parameters of the various traits of the ears is simple and clear, which provides a solution for the automatic test of corn ears.

Description of drawings

Fig. 1 is the basic flow chart of the image-based calculation method of corn ear trait index proposed by the present invention.

Fig. 2 is a schematic diagram of an image of an ear of corn in an embodiment of the present invention.

Fig. 3 is a schematic diagram of a method for calculating the number of row grains in an embodiment of the present invention.

Fig. 4 is a schematic diagram of a method for calculating the number of ear rows in one embodiment of the present invention.

Detailed ways

The specific implementation manners of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

The present invention proposes a kind of image-based corn ear trait index calculation method, as shown in Figure 1, described method comprises the following steps:

S1. Obtain the color image CI _i of the ear of corn by taking an image every 90 degrees, and i=1,2,3,4,

converting the color image CI _i into a grayscale image SI _i , and converting the grayscale image SI _i into a binary image BI _i containing only ears of corn;

S2. Perform axial distortion correction and radial distortion correction respectively on the color image CI _i , the grayscale image SI _i and the binary image BI _i to obtain the corrected color image CI2 _i and the corrected grayscale image Degree image HI _i and corrected binary image BI2 _i ;

S3. Segment the grayscale image HI _i to obtain a binary image BI3 _i containing only corn kernels,

Splicing the binary images BI3 _i to obtain a grain panorama BI4, the grain panorama BI4 includes all grains on the three-dimensional surface of the corn ear;

S4. Performing morphological operations on the binary image BI3 _i , calculating the bald area of the corn ear, and calculating the bald length of the corn ear,

According to the binary image BI3 _i , calculate the number of rows of corn ears,

According to the grain panorama BI4, the ear row number of the corn ear is calculated.

The preferred implementation of each step will be described in detail below through specific embodiments.

In step S1, the image of the ear of corn should be collected and preprocessed first.

Under a certain light condition, an image is taken every 90 degrees along the direction perpendicular to the central axis of the corn ear on a solid-color background, and the four color images obtained are represented as CI _i (respectively CI ₁ , CI ₂ , CI ₃ and CI ₄ ), transform it into a grayscale image SI _i with a value range of 255, and further combine morphological operations to obtain a binary image BI _i containing only corn ears.

Among them, the image shooting should try to ensure that the lighting conditions are the same, the shooting direction should be as perpendicular as possible to the central axis of the corn ear, and the shooting angle of adjacent images should be guaranteed to be 90 degrees.

Convert the color image to a grayscale image, the conversion method is: grayscale = (R×30+G×59+B×11)/100.

The binary image of the ear of corn refers to firstly determining the boundary pixel intensity of the ear and the background, which is used to binarize the grayscale image, and then uses the expansion operation to fill the hole in the ear image, and uses the erosion operation to delete the island in the image and restore the ear. Boundary, the final binary image contains only corn ears.

Among them, the ear threshold can be obtained using the OSTU method; using morphological operations, it should be ensured that the number of cumulative expansion and erosion pixels is equal to restore the outer contour edge of the ear; the method for judging that only the ear is included in the binary image is to use the calculated The size of the ear area, combined with prior knowledge to determine the ear area, delete other noise areas.

In step S2, the image needs to be corrected for axial and radial distortion.

Based on the binary image BI _i of the fruit ear, the OBB bounding box of each fruit ear is established, and the local coordinate system of the fruit ear is established based on this, and the axial and radial distortion correction parameters are calculated, so as to obtain the corrected ear color image CI2 _i , grayscale Image HI _i and binary image BI2 _i .

Among them, the OBB bounding box refers to the rectangle closest to the boundary of the ear, whose height in the long axis direction is the ear height, and its width in the short axis direction is the ear width.

The ear local coordinate system refers to the long axis direction of the bounding box as the central axis direction Y of the corn ear, and a short axis of the bounding box as the X axis to establish a new coordinate system. The origin of the coordinate system is located at the center point of the short axis.

Axial distortion correction is to calculate the angle between the Y-axis direction of the ear bounding box and the image height direction, rotate the ear to the vertical direction, then calculate the reference height of the ear, and shrink/enlarge each ear area proportionally to the specified height. The reference height is obtained by directly calculating the average height of the ear on the 4 images; proportional scaling means that when the height is scaled to the reference height, the ear width is scaled year-on-year, and the scaling factor is the ratio of the reference height to the height of the ear before scaling. The new scaled image is generated from the original image using a linear interpolation method.

Radial distortion correction is to use the three-dimensional shape characteristics of the ear to establish the corresponding relationship between each pixel on the ear image and each point on the three-dimensional ear surface, so that the ear image is expanded in the radial direction, so that each pixel on the image and its represented Ear surface nodes have the same physical scale. Among them, the ear radial direction refers to the radial direction (along the width direction of the image) of the cross-section passing through the center of the ear; the radial expansion method is that the current ear image is considered to be a projection image of a three-dimensional ear, and the image after expansion is a tiled image of a three-dimensional ear surface The resulting two-dimensional image. Among them, the ear center is determined by the edge pixel features in the ear binary image BI _i .

The distortion correction parameters are calculated by using the binary image, so as to obtain the corrected color image CI2 _i , grayscale image HI _i and binary image BI2 _i .

In step S3, grain segmentation and splicing are performed on the image.

The image after distortion correction restores the real size of corn kernels on the ear to a certain extent, and combines the color and shape features of the kernels to segment the gray image HI _i of the ear to obtain a binary image BI3 _i containing only corn kernels.

Among them, the image segmentation method can use any effective corn kernel segmentation method; the segmented kernel result is screened by its color feature and shape feature, the color feature is obtained from the image CI2 _i , and the kernel shape feature is obtained from the image BI2 _i . Among them, through the above distortion correction, the grain shape feature on the image can reflect the real shape of the grain, making the screening based on the grain shape more accurate.

Grain stitching is to unfold the segmented grain image BI3 _i according to the three-dimensional surface of the ear, and extract the most complete grain information from each grain image to stitch together a grain panorama BI4. The grain panorama contains all the grain information on the three-dimensional surface of the ear. The splicing method is to establish the mapping relationship between the same grains on different segmented images, and extract the grains with the least grain distortion and the most complete structure from each image as the grains on the panorama BI4.

In step S4, the obtained image is used to calculate the trait indicators such as the length of the bald tip of the corn ear, the number of grains in a row, the number of grains in an ear, and the number of rows in an ear.

The bald area is defined as the area where no corn kernels grow on the top of the ear, and the bald length is the vertical height of this area. At present, the method of calculating the length of the bald tip is generally to separately segment the bald tip area in the image. Due to the large variation in the shape, color, and size of the bald tip, there has been a lack of effective methods to segment the bald tips of various ear types. Based on the segmented grain image BI3 _i , the present invention uses morphological operations to fill the holes between the grains, merges the corn grain area into a single connected area, and then calculates the difference with the ear image BI2 _i to obtain the area that does not contain the grain. The part located at the top of the fruit ear in this area is judged as the bald area, and parameters such as the area of the bald point and the length of the bald point are calculated.

Among them, the process of morphological operation specifically includes: applying expansion operation to image BI3 _i until all grains are fused together to form a single connected area, recording the number of expansion pixels; then applying erosion operation to restore grain boundaries; and then combining the obtained image with The original contour images are subtracted to obtain the bald area.

The difference operation refers to subtracting the area of corn kernels from the ear binary image BI2 _i to obtain the area of ungrown corn kernels.

Figure 2 shows the extracted bald regions from 4 input images.

Row kernel count is defined as the average number of corn kernels per row on the ear image. Due to the variety of grain arrangements, manual counting is mainly based on experience and judgment, and there is a lack of automated and standardized calculation methods. Based on the segmented grain image BI3 _i , the present invention establishes a row grain number counting method that conforms to human visual habits.

First calculate the shape center coordinates of each grain on the image BI3 _i , and input it into the point set S _i ; set the starting point and end point of the counting of the number of grains in a row (generally take the two intersection points of the OBB bounding box of the corrected image and the Y axis) , and add it to the central point set S _i . Then, the point set S _i is connected into a planar triangular grid, and the calculation of the number of row grains is transformed into the calculation of the shortest path of the nodes in the given grid, the grid nodes passing through the path from the starting point to the end point (ie , not including the start and end points) is the number of row grains of the ear.

Among them, the triangular mesh is generated from the point set S _i using the Delaunay method. The Delaunay method can divide a set of scattered points into a triangular mesh, where the triangles satisfy two criteria: the empty circle feature and the maximized minimum corner feature.

Computes the shortest path between specified two points in a planar grid, using the BellmanFord method. The BellmanFord method is a dynamic programming algorithm for solving the single-source shortest path. First, an initial two-dimensional matrix is constructed. The size of the matrix is equal to the number of nodes. The value of each element in the matrix is initially assigned a maximum value, indicating that the path is unreachable. Then extract the vertex index of each edge in the grid, and calculate the weight of the connection between the vertices, where the connection weight is mainly based on the distance between two points, and a custom weight adjustment factor is added. The weight adjustment factor is to make the path search comply with user-defined features. In the present invention, an initial straight line is generated from the starting point to the end point, and the angle between each edge and the initial straight line and each point on the edge to the initial straight line are calculated. The vertical distance of , the angle, vertical distance, and distance between two points on the edge are used as the weight of the edge, and assigned to the corresponding position of the matrix. Then use the BellmanFord method to calculate all the node numbers on the path from the start point to the end point.

Since the number of ear rows with the least distortion and the most complete grain information on a single ear image is generally 3-4 rows, at least 2 shortest paths can be calculated for each image, which is used to count the average number of rows of grains. The method of calculating the number of rows of multiple groups of grains for each ear image is: record the grid nodes (except the start point and end point) obtained by each path calculation, and assign all the grid edges connected to these nodes to the maximum value in the array , block the path search, and calculate the shortest path between the starting point and the ending point again, that is, get the other row grain number indicators in the image.

Figure 3 is a schematic diagram of calculating the number of rows of kernels from 4 input images respectively, and the number of rows of kernels of corn ears finally obtained is the average number of row kernels of these images.

In the image-based ear species test, the number of ear rows is the most difficult index to determine accurately. Ear row number is defined as the number of independent rows of kernels on the surface of the ear. Due to the variety of ear varieties and individual differences, the arrangement of grains on the ear is diverse, which makes it very difficult to determine the number of rows in the ear. The invention establishes a decision-making model for calculating the number of ear rows, which uses three methods to calculate the number of ear rows in combination with the characteristics of grains, and provides applicable conditions.

The first method: the number of ear rows 1 = the number of grains per ear/the number of grains in a row. The key to this method is to accurately calculate the number of grains per ear and the number of rows of grains. The number of grains per ear is calculated from the grain panorama BI4, and the number of rows of grains is obtained from the image _BI3i .

The second method: the number of ear rows 2 = the circumference of the ear/the average width of the grain. Among them, the ear circumference is directly calculated from the image BI4; the calculation method of the average grain width is to first calculate the average area of the grain on the ear, that is, the average area of the grain = (ear area - bald area - invalid area) / ear grain Then calculate the aspect ratio of the grain from the image BI4, and combine the above two parameters to calculate the average width of the grain. Among them, the invalid area refers to the area on the ear except for the grain and the bald tip, which is obtained by using a calculation method similar to the bald tip area.

The third method, which is similar to the calculation method of the number of grains in a row, calculates the center point of all grains in the grain panorama, and then sets the starting point and end point on the panorama (generally take the two sides of the OBB bounding box of the corrected image along the Y axis) The midpoint of each side), these nodes are used to generate a triangular mesh according to the Delaunay criterion, and the BellmanFord method is used again to calculate the number of grid nodes that the shortest path from the starting point to the ending point passes through, and take it as the number of ear rows 3. Fig. 4 is a schematic diagram of the third method for calculating the number of ear rows.

The above three methods calculate the number of ear rows from different angles. Among them, the first method uses the global statistical information of the ear grains, and the applicable condition is that the number of ear grains is not much different; the second method considers the shape characteristics of the grains, The applicable condition is that the grain size in the ear is basically the same; the third method is suitable for any grain type, and the weight factor of the grid connection needs to be adjusted. Ultimately, which ear row number calculation method to use needs to be determined during the image processing process according to the characteristics of the current ear and its kernels.

The above is only a preferred embodiment of the present invention, it should be pointed out that for those skilled in the art, without departing from the technical principle of the present invention, some improvements and replacements can also be made, and these improvements and replacements are also It should be regarded as the protection scope of the present invention.

Claims (10)

1. an image-based corn ear character index calculating method, is characterized in that, said method comprising the steps of:

S1, the mode by taking an image every 90 degree are obtained the coloured image CI of corn ear _i, and i=1,2,3,4,

With described coloured image CI _iBe converted into gray level image SI _i, and with described gray level image SI _iBe converted into the bianry image BI that only comprises corn ear _i

S2, respectively to described coloured image CI _i, described gray level image SI _iAnd described bianry image BI _iCarry out axial distortion correction and Lens Distortion Correction, the coloured image CI2 after being proofreaied and correct _i, the gray level image HI after correction _iAnd the bianry image BI2 after proofreading and correct _i

S3, to described gray level image HI _iCut apart, obtain only comprising the bianry image BI3 of corn kernel _i,

With described bianry image BI3 _iSplice, obtain seed panorama sketch BI4, described seed panorama sketch BI4 comprises all seeds on described corn ear three-dimensional surface;

S4, to described bianry image BI3 _iCarry out morphology operations, calculate the bald point zone of described corn ear, and calculate the bald sharp length of described corn ear,

According to described bianry image BI3 _i, calculate the row grain number of described corn ear,

According to described seed panorama sketch BI4, calculate the tassel row number of described corn ear.

2. image-based corn ear character index calculating method according to claim 1, is characterized in that, in step S1, with described gray level image SI _iBe converted into the bianry image BI that only comprises corn ear _iSpecifically comprise:

Determine described gray level image SI _iThe boundary pixel intensity of middle corn ear and background and to described gray level image SI _iCarry out binaryzation, use dilation operation to fill vacancy, use the isolated island in corrosion operation deleted image and the border of recovering corn ear, thereby obtain described bianry image BI _i

3. image-based corn ear character index calculating method according to claim 1, is characterized in that, in step S2, described axial distortion correction and described Lens Distortion Correction specifically comprise:

Based on described bianry image BI _iSet up the OBB bounding box of corn ear in each image, and set up the fruit ear local coordinate system according to described OBB bounding box, calculate axial distortion correction parameter and Lens Distortion Correction parameter, according to described axial distortion correction parameter and described Lens Distortion Correction parameter respectively to described coloured image CI _i, described gray level image SI _iAnd described bianry image BI _iProofread and correct.

4. image-based corn ear character index calculating method according to claim 1, is characterized in that, step S3 further comprises:

To described gray level image HI _iAfter cutting apart, utilize the color characteristic of corn kernel and shape facility that the corn kernel that is partitioned into is screened, thereby obtain described bianry image BI3 _i, wherein,

The color characteristic of described corn kernel is by described coloured image CI2 _iObtain, the shape facility of described corn kernel is by described bianry image BI2 _iObtain.

5. one of according to claim 1-4 described image-based corn ear character index calculating methods, is characterized in that, in step S3, with described bianry image BI3 _iSplice specifically and comprise:

Described bianry image BI3i is launched according to the fruit ear three-dimensional surface, set up the mapping relations between identical corn kernel on each image, and extract the corn kernel of structural integrity as the corn kernel on described seed panorama sketch BI4 from each image.

6. image-based corn ear character index calculating method according to claim 1, is characterized in that, in step S4, described bianry image BI3i carried out morphology operations specifically comprise:

To described bianry image BI3 _iCarry out dilation operation, form a simply connected region until all corn kernels in image merge, record the expansion number of pixels, the border that then utilizes erosion operation to recover corn kernel, then with the image and the described bianry image BI2 that obtain _iCarry out difference operation, obtain not comprising the zone of corn kernel, and it is regional further to obtain described bald point.

7. image-based corn ear character index calculating method according to claim 3, is characterized in that, in step S4, the row grain number that calculates described corn ear specifically comprises:

According to described bianry image BI3 _iIn the coordinate of each corn kernel in described fruit ear local coordinate system, set up point set S _i, the starting point A of setting row grain counting number _iWith terminal point B _i, and with described starting point A _iWith described terminal point B _iJoin described point set S _iIn,

With described point set S _iConnect into planar triangulations T _i, and calculate described planar triangulations T _iIn from described starting point A _iTo described terminal point B _iShortest path P _i, described shortest path P _iIn except described starting point A _iWith described terminal point B _iThe number of grid node in addition is described bianry image BI3 _iA row grain number, described bianry image BI3 _iThe mean value of row grain number be the row grain number of described corn ear.

8. image-based corn ear character index calculating method according to claim 1, is characterized in that, in step S4, the tassel row number that calculates described corn ear specifically comprises:

According to described seed panorama sketch BI4, calculate the grain number per spike of described corn ear, the business that described grain number per spike obtains divided by described row grain number is the tassel row number of described corn ear.

9. image-based corn ear character index calculating method according to claim 1, is characterized in that, in step S4, the tassel row number that calculates described corn ear specifically comprises:

According to described seed panorama sketch BI4, calculate the girth of described corn ear and the mean breadth of corn kernel, the business that described girth obtains divided by the mean breadth of described corn kernel is the tassel row number of described corn ear.

10. image-based corn ear character index calculating method according to claim 1, is characterized in that, in step S4, the tassel row number that calculates described corn ear specifically comprises:

According to the center position of each corn kernel in described seed panorama sketch BI4, set up point set S ', set starting point A' and the terminal point B ' of tassel row number counting, and described starting point A' and described terminal point B ' are joined in described point set S ',

Described point set S ' is connected into planar triangulations T ', and calculating the shortest path P ' from described starting point A' to described terminal point B ' in described planar triangulations T ', the number of the grid node in described shortest path P ' except described starting point A' and described terminal point B ' is the tassel row number of described corn ear.

Images