Method for accurately and efficiently analyzing chrysanthemum leaf shape characteristics

Abstract

The invention discloses a method for accurately and efficiently analyzing chrysanthemum leaf shape features. The method comprises the following steps: shooting a leaf RGB (Red, Green and Blue) image and preprocessing the image; drawing a chrysanthemum leaf characteristic curve; chrysanthemum leaf key point detection: screening more significant peaks in the distance characteristic curve through peak value prominence, obtaining a key point set, drawing key points in the leaf image, and verifying the accuracy of key point detection; chrysanthemum leaf feature calculation: further calculating leaf features of leaf length-width ratio, leaf stalk length ratio, apical fissure piece included angle and apical fissure piece relative length for leaf images with correct key point detection; characterizing the similarity of the chrysanthemum leaves: calculating the cross correlation of the distance characteristic curves of the left and right sides of the chrysanthemum leaves so as to characterize the contour mirror symmetry of the chrysanthemum leaves. The characteristics of the chrysanthemum leaves are analyzed on the basis of the RGB images of the chrysanthemum leaves, and the method is efficient and convenient; chrysanthemum leaf features are analyzed based on image processing, dependence on a deep learning algorithm and data labeling training is not needed, and the requirement for computing power is low.

Description

Technical Field

[0001] The present invention relates to a method for analyzing the leaf shape characteristics of chrysanthemums, and particularly to a method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums. Background Art

[0002] As one of the world's four major cut flowers, the commercial production of chrysanthemums has extremely high requirements for variety purity. As the expression carrier of genetic traits, the leaf morphological characteristics carry rich morphological fingerprint information and are the core basis for variety identification. At the same time, experienced producers can also obtain the basis for cultivation regulation through the subtle changes in the leaves. Accurately obtaining the key features of the leaf details and establishing an intelligent digital phenotype analysis system are of great significance for guiding the variety morphological fingerprint database and precise decision-making cultivation. Traditional leaf morphological observation methods mostly rely on manual measurement and marking. Professional technicians need to use tools such as vernier calipers to measure each leaf one by one, which has problems such as low efficiency and significant subjective errors, and it is difficult to meet the needs of modern flower industry for batch and rapid phenotype analysis of germplasm resources.

[0003] In recent years, although computer vision technology has made many progress in the field of plant phenotype analysis, there is still no efficient feature analysis technology for the special morphology of chrysanthemum leaves. Summary of the Invention

[0004] Object of the Invention: The object of the present invention is to provide a method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums, which can efficiently, objectively and accurately obtain various phenotype characteristics of chrysanthemum leaves based on the RGB images of chrysanthemum leaves.

[0005] Technical Solution: The present invention includes the following steps:

[0006] Shooting of leaf RGB images and image preprocessing;

[0007] Drawing of chrysanthemum leaf feature curves, specifically including:

[0008] Convex hull contour extraction: Calculate the convex hull of the detected chrysanthemum leaf contour, and the output result is the set of vertices constituting the convex hull;

[0009] Distance calculation: Taking the leaf tip of the chrysanthemum leaf as the starting point, establish the traversal order of the contour points, and record the serial number of the contour point as i. Use the distance formula from a point to a line to calculate its distance to the corresponding convex hull line segment, denoted as d i , and the calculation method is shown in formula (1);

[0010]

[0011] Among them, i represents the serial number of the contour point, (x i , y i ) represents the contour point P iThe coordinates, (x1, y1) and (x2, y2), represent P i The coordinates corresponding to the two endpoints of the convex hull line segment;

[0012] Draw the distance feature curve: Use the contour point number i as the abscissa and the corresponding distance d i as the ordinate to draw the distance feature curve;

[0013] Calculate the number of the bottommost point of the blade contour points. Taking the number as the boundary, divide the distance feature curve into two segments, which respectively correspond to the left and right sides of the blade;

[0014] Chrysanthemum leaf key point detection: Through the peak prominence, screen the relatively prominent peaks in the distance feature curve to obtain the key point set, draw the key points in the leaf image, and verify the accuracy of the key point detection;

[0015] Chrysanthemum leaf feature calculation: For the leaf image with correct key point detection, further calculate the leaf features: leaf length-width ratio, petiole length ratio, apical lobe angle, apical lobe relative length;

[0016] Chrysanthemum leaf similarity characterization: Calculate the cross-correlation of the distance feature curves on the left and right sides of the chrysanthemum leaf to characterize the mirror symmetry of the chrysanthemum leaf contour.

[0017] The point-line correspondence establishment method is as follows: The convex hull points belong to the contour point set and have their corresponding numbers in the blade contour point set. Each blade contour point is located in the corresponding convex hull point number interval through its number. Connect two convex hull points to form a straight line, and calculate the distance from the blade contour point to the straight line. If a blade contour point belongs to the convex hull point, the distance is 0.

[0018] The above-mentioned chrysanthemum leaf key point detection is specifically as follows: Set the significance threshold for peak searching to be not less than 0.45 of the maximum value of the ordinate, and screen all peak points that meet the conditions; if the number of retrieved peak points exceeds 5, only retain the 5 most significant peak points; if the number of peak points is less than 3, further reduce the limit value to search until 3 peak points are found; According to the x values of the screened peak points, correspond to the coordinates of the contour point set to obtain the key point set.

[0019] The calculation method of the above-mentioned leaf length-width ratio is as follows: Retrieve the minimum value x of x in the blade contour point coordinate set min , the maximum value x max , the minimum value y of y min , the maximum value y max , record the leaf length as h leaf , its calculation method is shown in formula (2), record the leaf width as w leaf , its calculation method is shown in formula (3), and record the leaf length-width ratio as LWR, and its calculation method is shown in formula (4);

[0020] w leaf = x max -x min (2)

[0021] h leaf = y max -y min (3)

[0022]

[0023] The calculation method of the petiole length ratio is as follows: in the key point set, the two points with the largest y value are the attachment points of the leaf base at the petiole. Calculate the y average value of these two points, denoted as y petiole , and the petiole length ratio is denoted as PLR. The calculation method is shown in formula (5):

[0024]

[0025] The calculation method of the apical lobe angle is as follows: the deep cleavage points of the apical lobe are the 2 points with the smallest y value in the key point set, and the leaf tip point is the point with the smallest y value in the leaf contour point set. Taking the leaf tip point as the vertex and the deep cleavage points as the two side points, calculate the apical lobe angle:

[0026]

[0027] Let the coordinates of the leaf tip point V be (x v , y v ), and the coordinates of the deep cleavage points A and B be (x a , y a ) and (x b , y b ). First, taking V as the vertex, construct vectors and Calculate the radian value θ of the included angle, and then convert it to the angle value, denoted as ALA. The calculation method is shown in formulas (6) to (12).

[0028] The calculation method of the relative length of the apical lobe is as follows: take the y average value of the coordinates of the 2 deep cleavage points of the apical lobe, denoted as y mean , and the y value of the leaf tip point, y v . Take the difference, and then compare it with the leaf length, which is the relative length of the apical lobe:

[0029]

[0030] The specific image preprocessing is as follows: convert the original RGB image to a grayscale image; convert the grayscale image to a binary image, and finally use the contour detection function to extract the main contour of the chrysanthemum leaf.

[0031] In the characterization of the similarity of chrysanthemum leaves, the distance feature curve of the chrysanthemum leaf is divided into left and right segments, corresponding to the left and right sides of the chrysanthemum leaf respectively, and the feature curve is smoothed.

[0032] When taking the RGB image of the leaf, the posture of the leaf with the leaf tip upward and the leaf stalk downward is adopted.

[0033] Beneficial effects: Based on the RGB image of the chrysanthemum leaf taken by the shooting terminal or scanner, the characteristics of the chrysanthemum leaf are analyzed, which is efficient and convenient; analyzing the characteristics of the chrysanthemum leaf based on the image processing technology, the leaf characteristics can be calculated without relying on deep learning algorithms and data annotation training, and the computing power requirement is low; it is applicable to the calculation of the leaf characteristics of different varieties of chrysanthemums and has universality. Description of the Drawings

[0034] Figure 1 It is the schematic diagram of image preprocessing of the present invention: (a) Grayscale image of the leaf; (b) Binary image of the leaf; (c) Contour image of the leaf;

[0035] Figure 2 It is the schematic diagram of distance calculation of the present invention, d i represents the distance from point P i to its corresponding convex hull line segment;

[0036] Figure 3 It is the distance feature curve of the present invention: (a) Distance feature curve of the whole leaf; (b) Distance feature curve of the left contour of the leaf; (c) Distance feature curve of the right contour of the leaf; The dot in (b) and (c) represents the peak point that meets the condition;

[0037] Figure 4 It is the leaf key points determined by the present invention, and the dot in the figure represents the detected leaf key points;

[0038] Figure 5 It is the schematic diagram of leaf feature calculation of the present invention: (a) Schematic diagram of the length and width of the leaf; (b) Schematic diagram of the attachment point of the leaf base; (c) Schematic diagram of the feature calculation of the apical lobe, V represents the leaf tip point, and A and B represent the two deep cleavage points of the apical lobe; (d) Schematic diagram of the feature calculation of the apical lobe, C represents the midpoint of the two deep cleavage points of the apical lobe, and D represents the midpoint of the attachment point of the leaf base. Detailed Embodiments

[0039] The present invention will be further described below with reference to the accompanying drawings.

[0040] As Figure 1As shown, the method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums according to the present invention is a method that can conveniently and accurately calculate characteristics such as the included angle of the apical lobe of the chrysanthemum leaf, the length of the apical lobe, and the left-right symmetry of the leaf. A photographing terminal or a scanner is used to photograph the chrysanthemum leaf, and combined with image analysis technology, the contour and key points of the chrysanthemum leaf are detected for calculating the characteristics of the chrysanthemum leaf. The specific steps are as follows:

[0041] S1. Photographing the RGB image of the leaf

[0042] In an environment with uniform light, a mobile phone or a camera is used to photograph the chrysanthemum leaf, or a scanner can also be used to scan the leaf to obtain an image. When photographing, it is recommended to use a solid-color background (usually light color), and ensure that the leaf maintains the standard posture with the leaf tip upward and the petiole downward, as Figure 1 shown in a. The obtained picture is saved in JPEG format.

[0043] S2. Image preprocessing

[0044] The OpenCV library in Python is used to convert the RGB picture into a grayscale picture, as Figure 1 shown in a, and then the Otsu's method is used to binarize the picture, as Figure 1 shown in b. The contour finding function is used to extract the contour of the chrysanthemum leaf, as Figure 1 shown in c;

[0045] Specifically, the image preprocessing process is implemented using the Python programming language combined with the OpenCV computer vision library. First, the original RGB image is converted into a grayscale image based on the Excess Green Index (ExG); then the Otsu adaptive threshold method is applied to convert the grayscale image into a binary image, and finally the cv2.findContours contour detection function is used to accurately extract the main contour of the chrysanthemum leaf.

[0046] S3. Drawing the characteristic curve of the chrysanthemum leaf

[0047] S31. Extracting the convex hull contour: The cv2.convexHull function is used to calculate the convex hull of the detected contour of the chrysanthemum leaf, and the output result is the set of vertices constituting the convex hull;

[0048] S32. Distance calculation: Taking the leaf tip of the chrysanthemum leaf as the starting point, a counterclockwise traversal order of the contour points is established, and the serial number of the contour point is denoted as i. The distance from it to the corresponding convex hull line segment is calculated using the point-to-line distance formula, denoted as d i , and the calculation method is shown in formula (1);

[0049]

[0050] where i represents the serial number of the contour point, (xi , y i ) represents the coordinates of the contour point P i . (x1, y1) and (x2, y2) represent the coordinates of the two endpoints of the convex hull line segment corresponding to P i ;

[0051] Specifically, the point-line correspondence is established as follows: The convex hull points belong to the set of contour points and have their corresponding serial numbers in the set of leaf contour points. Each leaf contour point can be located in the corresponding convex hull point serial number interval through its serial number. Connect two convex hull points to form a straight line, and calculate the distance from the leaf contour point to the straight line. For example Figure 2 . If a leaf contour point belongs to a convex hull point, the distance is 0;

[0052] S33. Plot the distance feature curve: Taking the contour point serial number i as the abscissa and the corresponding distance d i as the ordinate, the distance feature curve can be plotted to improve the visualization effect, as shown in Figure 3 a;

[0053] S34. Calculate the serial number of the lowest point of the leaf contour points. Taking the serial number as the boundary, divide the distance feature curve into two segments, which respectively correspond to the left and right sides of the leaf, as shown in Figure 3 b and 3c.

[0054] S4. Key point detection of chrysanthemum leaves

[0055] The edges of chrysanthemum leaves are often deeply lobed and serrated. The distance feature curve can reflect this feature. The distances from the contour points at the deeply lobed parts of the leaf and at the attachment of the leaf base to the petiole to the convex hull line segment are relatively large. The prominent peaks in the distance feature curve can be screened through the prominence parameter in the scipy.signal.find_peaks function, that is, the peak prominence, to obtain the key point set. Draw the key points in the leaf image to verify the accuracy of the key point detection.

[0056] Specifically, taking the unilateral feature curve as an example, set the prominence (significance) threshold for peak searching to be not lower than 0.45 of the maximum value of the ordinate, and screen all peak points that meet the conditions. If the number of retrieved peak points exceeds 5, only retain the 5 most prominent peak points (because chrysanthemum leaves generally have at most 9 lobes, and at most 4 deeply lobed points and 1 petiole point on one side). If the number of peak points is less than 3, further reduce the limit value to search until 3 peak points are found (chrysanthemum leaves generally have 5 lobes, and at least 2 deeply lobed points and 1 petiole point on one side);

[0057] According to the x values of the screened peak points, they can be corresponding to the coordinates of the contour point set to obtain the key point set. These key points can be drawn in the leaf image to verify the accuracy of the key point detection, as shown inFigure 4 。

[0058] S5. Chrysanthemum leaf feature calculation

[0059] For the leaf images with correct key point detection, further calculate the leaf features: leaf length-width ratio (Leaf WidthRatio, LWR), petiole length ratio (Petiole Length Ratio, PLR), apical lobe angle (Apical LobeAngle, ALA), and apical lobe relative length (Apical Lobe Length Ratio, ALLR).

[0060] The leaf contour point P = (x i , y i ), and i is the serial number of the point in the contour point set. For the leaf images with correct key point detection, the following features can be calculated:

[0061] a. Leaf length-width ratio, such as Figure 5 a: Retrieve the minimum value x min of x and the maximum value x max in the leaf contour point coordinate set, the minimum value y min of y, and the maximum value y max . The leaf length is denoted as h leaf , and its calculation method is shown in formula (2). The leaf width is denoted as w leaf , and its calculation method is shown in formula (3). The leaf length-width ratio is denoted as LWR, and its calculation method is shown in formula (4);

[0062] w leaf = x max - x min (2)

[0063] h leaf = y max - y min (3)

[0064]

[0065] b. Petiole length ratio: In the key point set, the two points with the largest y values are the attachment points of the leaf base at the petiole. As shown in Figure 5 b, find the average value of the y coordinates of these two points, denoted as y petiole , and the petiole length ratio is denoted as PLR, and its calculation method is shown in formula (5);

[0066]

[0067] c. Apical lobe angle: The deep cleavage points of the apical lobe are the two points with the smallest y values in the key point set. The leaf tip point is the point with the smallest y value in the leaf contour point set. As shown in Figure 5c. With the tip point as the vertex and the deep-lobed points as the two side points, the apical lobe angle (ALA) can be calculated;

[0068]

[0069]

[0070] Let the coordinates of the tip point V be (x v , y v ), and the coordinates of the deep-lobed points A and B be (x a , y a ) and (x b , y b ). First, with V as the vertex, construct vectors and to calculate the radian value θ of the included angle, and then convert it to the angle value, denoted as ALA. The calculation method is shown in Formulas (6) to (12);

[0071] d. Apical lobe length: Take the average value of the y coordinates of the two deep-lobed points of the apical lobe. For example, Figure 5 d, denoted as y mean . Subtract it from the y value (y v ) of the tip point, and then divide by the leaf length. Then it is the relative length of the apical lobe (Apical Lobe Length Ratio, ALLR) (Formula 13).

[0072]

[0073] S6. Characterization of chrysanthemum leaf similarity

[0074] According to Step S34, divide the distance feature curve of the chrysanthemum leaf into two sections, the left and the right, which correspond to the left and right sides of the chrysanthemum leaf respectively, and use the function scipy.signal.savgol_filter to smooth the feature curve. Use the np.correlate function to calculate the cross-correlation (Leaf Curve Correlation, LCC) of the distance feature curves on the left and right sides of the chrysanthemum leaf to characterize the mirror symmetry of the chrysanthemum leaf contour, mainly to characterize the symmetry of the lobes and serrations.

[0075] Results and analysis

[0076] Detection accuracy of key points of chrysanthemum leaves

[0077] The key points of the leaves detected in this experiment are mainly used to calculate the proportion of petiole length, the included angle of the apical lobe, and the relative length of the apical lobe. Therefore, the detection accuracies of the deep cleavage point of the apical lobe and the leaf attachment point at the petiole were statistically analyzed. The results show that the accuracy of this method for detecting the deep cleavage point of the apical lobe is 0.896, and the accuracy of the leaf attachment point at the petiole is 0.984. This indicates that for most leaves, this method can provide an ideal detection effect.

[0078] Calculation of Chrysanthemum Leaf Characteristics

[0079] Based on the detected key points of the leaves, the characteristics of the leaves were calculated in this experiment, and some of the results are shown in the following table. As can be seen from the table, there are significant differences in the morphological characteristics of chrysanthemum leaves of different varieties, including the length-width ratio of the leaves, the characteristics of the lobes, and the edge morphology, etc. These significant morphological variations indicate that leaf characteristics have the potential for variety identification to a certain extent and can provide a reference for the classification and identification of chrysanthemum varieties.

[0080] Table 1 Leaf Characteristics of Different Varieties

[0081]

[0082]

[0083] In summary, the method for calculating chrysanthemum leaf characteristics of the present invention extracts contour shape information from the RGB image of the leaf, draws a distance feature curve according to the contour information, detects the key points of the leaf, calculates the leaf characteristics, and characterizes the similarity of the leaf contour through the similarity of the distance feature curve. The method of the present invention analyzes chrysanthemum leaf characteristics through RGB images, which is economical, efficient, and accurate, and can reduce manual measurement work. The present invention is applicable to chrysanthemum leaves of different varieties, indicating that the present invention has a wide range of applications and can be popularized and applied.

Claims

1. A method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums, characterized in that, It includes the following steps: Taking the RGB image of the leaf and image preprocessing; Drawing the characteristic curve of the chrysanthemum leaf, specifically including: Convex hull contour extraction: calculating the convex hull of the detected chrysanthemum leaf contour, and the output result is the vertex set that constitutes the convex hull; Distance calculation: Taking the leaf tip of the chrysanthemum leaf as the starting point, establish the traversal order of the contour points, and denote the serial number of the contour point as i. Use the distance formula from a point to a straight line to calculate its distance to the corresponding convex hull line segment, denoted as d i ; Plot the distance feature curve: Use the contour point serial number \(i\) as the abscissa and the corresponding distance \(d\) i as the ordinate to plot the distance feature curve; Calculating the serial number of the bottommost point of the leaf contour points. Taking the serial number as the boundary, the distance feature curve is divided into two segments, corresponding to the left and right sides of the leaf respectively; Detecting the key points of the chrysanthemum leaf: screening the relatively prominent peaks in the distance feature curve through the peak prominence, obtaining the key point set, drawing the key points in the leaf image, and verifying the accuracy of the key point detection; Calculating the characteristics of the chrysanthemum leaf: for the leaf image with correct key point detection, further calculating the leaf characteristics: leaf length-width ratio, petiole length ratio, apical lobe angle, apical lobe relative length; Characterizing the similarity of the chrysanthemum leaf: calculating the cross-correlation of the distance feature curves on the left and right sides of the chrysanthemum leaf to characterize the mirror symmetry of the chrysanthemum leaf contour.

2. The method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums according to claim 1, characterized in that The calculation formula for the distance of the convex hull line segment is: where i represents the serial number of the contour point, (x i , y i ) represents the coordinates of the contour point P i , and (x1, y1) and (x2, y2) represent the coordinates of the two endpoints of the convex hull line segment corresponding to P i .

3. A method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums according to claim 1, characterized in that, For the detection of the key points of the chrysanthemum leaf, specifically: setting the significance threshold for peak searching to be not lower than 0.45 of the maximum value of the ordinate, and screening all peak points that meet the conditions; if the number of retrieved peak points exceeds 5, only the 5 most significant peak points are retained; if the number of peak points is less than 3, the limit value is further reduced to search until 3 peak points are found; according to the x values of the screened peak points, corresponding to the contour point set coordinates, the key point set is obtained.

4. A method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums according to claim 3, characterized in that The calculation method for the aspect ratio of the blade is as follows: retrieve the minimum value x of x in the blade contour point coordinate set min , the maximum value x max , the minimum value y of y min , the maximum value y max , the blade length is denoted as h leaf , and its calculation method is shown in formula (2). The blade width is denoted as w leaf , and its calculation method is shown in formula (3). The blade aspect ratio is denoted as LWR, and its calculation method is shown in formula (4); w leaf = x max -x min (2) h leaf = y max -y min (3) 5. The method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums according to claim 4, characterized in that, The calculation method of the petiole length ratio is as follows: in the key point set, the two points with the largest y value are the attachment points of the leaf base at the petiole, and the y average value of these two points is calculated and denoted as y petiole , and the petiole length ratio is denoted as PLR. The calculation method is shown in formula (5):

6. The method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums according to claim 5, characterized in that, The calculation method for the apical lobe angle is: the deep cleavage points of the apical lobe are the 2 points with the smallest y value in the key point set, and the leaf tip point is the point with the smallest y value in the leaf contour point set. Taking the leaf tip point as the vertex and the deep cleavage points as the two side points, the apical lobe angle is calculated: Let the coordinates of the tip point V be (x v , y v ), and the coordinates of the deeply lobed points A and B be (x a , y a ) and (x b , y b ). First, with V as the vertex, construct vectors and to calculate the radian value θ of the included angle, and then convert it to the angle value, denoted as ALA. The calculation method is shown in Formulas (6) to (12).

7. A method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums according to claim 6, characterized in that, The calculation method for the relative length of the apical lobe is as follows: Take the y - mean value of the coordinates of two deep - split points of the apical lobe, denoted as y mean , and the y - value of the leaf - tip point, y v . Calculate the difference, and then divide it by the leaf length. The result is the relative length of the apical lobe:

8. A method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums according to claim 1, characterized in that, The image preprocessing is specifically: converting the original RGB image into a grayscale image; converting the grayscale image into a binary image, and finally using the contour detection function to extract the main contour of the chrysanthemum leaf.

9. A method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums according to claim 1, characterized in that In the characterization of the similarity of the chrysanthemum leaf, the distance feature curve of the chrysanthemum leaf is divided into two segments on the left and right, corresponding to the left and right sides of the chrysanthemum leaf respectively, and the feature curve is smoothed.

10. A method for accurately and efficiently analyzing the leaf shape characteristics of chrysanthemums according to claim 1, characterized in that, When taking the RGB image of the leaf, the posture of the leaf with the leaf tip upward and the petiole downward is adopted.

Images