Image detection method for grafted rootstock

Abstract

The present invention discloses an image detection method for grafted rootstocks. The method can comprehensively detect and judge information such as surface damage, curvature, bud nodes, coarseness, and clamping positions of grafted rootstocks through edge detection based on an improved Canny operator, surface damage detection based on a morphology and area method, curvature detection based on a Hough transform linear method, and identification of rootstock coarseness and grasping points. The method can meet the selection requirements of an automatic grafting machine for grafted rootstocks. Compared with traditional manual detection methods, the method has higher detection efficiency and more accurate and stable detection results.

Description

Technical Field

[0001] The invention relates to the technical field of seedling image recognition, in particular to an image detection method for grafted rootstocks. Background Art

[0002] my country is a major grape-growing country, with the second-largest grape-planting area in the world. Grafting, which enriches varieties, improves quality, and enhances ecological resilience, is currently the primary method of grape propagation. Grafting can be divided into two methods: manual and mechanical. Mechanical grafting overcomes many of the challenges of traditional manual grafting, including low production efficiency, high labor intensity, and inconsistent grafting quality. It holds significant practical significance for achieving factory-scale production of seedlings and promoting the healthy development of the grape-growing industry.

[0003] Currently, the only reported grafting machines for mechanical grape grafting are the German Wahler Omega-Uno semi-automatic grape grafting machine and the PJJ-50 semi-automatic grape grafting machine developed by the Yantai Agricultural Machinery Science Research Institute in China. While these semi-automatic grafting machines significantly improve grafting efficiency compared to manual grafting, they also suffer from low automation levels, dependence on operator skill, and difficulties in further improving grafting efficiency. With the loss of skilled workers and rising labor costs, the development of automatic grafting machines has become an urgent need for the development of the grape seedling industry.

[0004] Before grafting, quickly and accurately selecting rootstocks and scion trees that meet the operation requirements is the premise and basis for ensuring the success rate and grafting survival rate of the automatic grafting machine. Generally, the selection requirements of the grafting rootstock for the automatic grafting machine include: (1) no surface damage that affects the grafting survival rate, such as cracks; (2) it should be as straight as possible with a small curvature to meet the requirements of its rapid and stable placement in the cutting module; (3) cutting should be performed 4 cm above the bud node, and its thickness at the cutting point should be close to that of the scion tree to ensure that the cambium is aligned after the rootstock and scion are joined, thereby ensuring the survival rate of the grafted seedlings; (4) the clamping position of the seedling feeding robot on the rootstock should be relatively fixed with its cutting position to ensure that the rootstock is cut at the set position after it is placed in the cutting module. Therefore, the automatic grafting machine needs to complete the above-mentioned detection and judgment of the rootstock's surface damage, curvature, bud node, thickness, clamping position and other information before grafting.

[0005] Due to the advantages of machine vision, such as large amount of information, non-contact, low failure rate, and low cost, its application in seedling information detection has received increasing attention. For example, Chinese invention patent 201910610333.9 proposes a method for identifying cutting parameters of tomato seedlings using computer vision technology. The method includes the following steps: (1) image acquisition and image contour chain extraction, (2) determination of the center slope of the stem, (3) determination of the root position of the seedling, and (4) determination of the position of the true leaves and cotyledons. For example, Chinese invention patent 201210417635.2 proposes a method for measuring the external features of grafted seedlings based on machine vision, which includes the following steps: (1) capturing an image of the entire tray of grafted seedlings in a top-down direction; (2) using an image processing algorithm to calculate the cotyledon parameter information of all grafted seedlings in the captured image of the entire tray of grafted seedlings; (3) capturing an image of a row of grafted seedlings in a front-facing direction, and using an image processing algorithm to calculate the plant height and the major and minor axis parameter information of the seedling diameter ellipse of each grafted seedling in the row; (4) capturing images of each row of grafted seedlings in a front-facing direction, and repeating step C until all grafted seedlings have been measured. Both of the above technologies are used to detect the external features of fruit and vegetable seedlings and are key technologies for vegetable grafting machines. However, fruit and vegetable seedlings differ greatly from grape seedlings in terms of color, external morphology, grafting method, and detection object, and the above detection methods cannot be used for grape rootstocks.

[0006] In order to solve the problem of low efficiency of existing manual seedling selection, Chinese invention patent 202110714343.4 discloses a seedling diameter measurement system and method based on machine vision. The measurement system consists of a mechanical transmission mechanism, a color sensor, an image capture mechanism and a computer image analysis module. The color sensor collects the color of the seedling and sends a voltage signal. The computer image analysis module receives the voltage signal to control the image capture mechanism to take a photo to obtain the seedling image. The computer image analysis module processes the image and obtains the seedling diameter data. In addition, Chinese invention patent 202110714346.8 discloses a system and method for measuring the curvature of seedlings suitable for mechanical grafting based on machine vision. The system consists of a mechanical transmission platform, a seedling image recognition and processing module and a seedling curvature measurement module. The mechanical transmission platform is used to transfer the seedlings to the position of the seedling image recognition and processing module. The seedling image recognition and processing module is used for collecting and preprocessing the seedling image. The seedling curvature measurement module extracts features from the image and calculates the curvature information of the seedling. The above two technical solutions respectively solve the problem of extracting single information of seedling thickness and curvature, but neither of the above two methods can directly meet or simply combine the two to meet the requirements of the automatic grafting machine for comprehensive detection and judgment of information such as surface damage, curvature, bud nodes, thickness, and clamping position of the grafted rootstock, so as to quickly and accurately select rootstocks and scion wood that meet the operation requirements, and ensure the success rate of the automatic grafting machine operation and the grafting survival rate. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide an image detection method for grafted rootstocks, which can comprehensively detect and judge information such as surface damage, curvature, bud nodes, thickness, clamping position, etc. of the grafted rootstock, so as to quickly and accurately select rootstocks and scion woods that meet the operation requirements, thereby ensuring the operation success rate and grafting survival rate of the automatic grafting machine.

[0008] In order to solve the above technical problems, the present invention adopts the following technical means:

[0009] To solve the above technical problems, the present invention provides an image detection method for grafted rootstocks, comprising the following steps:

[0010] S1, collect images of the rootstock and the detection platform, grayscale, filter and binarize the images, and use the improved Canny operator to extract the edge contours of each rootstock. The specific steps are as follows:

[0011] S1.1 Collect images of the test platform before and after stock placement using the same parameters;

[0012] S1.2 uses the image with the rootstock placed as the original image and the image without the rootstock placed as the background image to obtain the rootstock image using the background difference method. Compared with other commonly used background segmentation algorithms, the background difference method is more sensitive to scene changes, but has the advantages of high operating efficiency, good stability, and no high requirements on the target. However, since the automatic grafting machine operates indoors, the image acquisition scene and acquisition parameters are controllable, so scene changes can be easily and effectively controlled by taking measures.

[0013] S1.3 grayscales the rootstock image using the 3r-b operator to obtain a grayscale image;

[0014] S1.4 performs Gaussian filtering on the grayscale image and performs binarization processing using the Ostu algorithm to obtain a binary image;

[0015] S1.5 inverts the binary image and marks the connected domains of each rootstock placement area;

[0016] S1.6 uses the improved Canny method to perform edge detection on the image, extracts the edge contour of each rootstock, and inverts the image again to obtain the rootstock contour image;

[0017] Step S2: Detecting the surface damage of each rootstock using morphology and planimetry. Surface damage to grape grafted rootstocks is usually manifested as cracks and cutting marks formed during the shearing process. When the damage reaches a certain level, it will make it difficult for the grafted seedlings to survive. Therefore, it is necessary to detect the degree of rootstock damage. The specific steps of the detection are as follows:

[0018] S2.1 performs a closing operation on the binary image and subtracts the result from the original image to obtain an image of the damaged area. This operation can reveal narrow, discontinuous, and elongated gullies and small holes in the connected domain.

[0019] S2.1 Use the area method to remove small, isolated areas;

[0020] S2.3 Mark the connected domains of each rootstock placement area;

[0021] S2.4 Calculate the length, width, and area parameters of each damaged location; these parameters serve as the basis for determining the damage type and its severity. When the parameters exceed a set threshold, the rootstock is deemed unsuitable for grafting.

[0022] The principle and process of graft survival is as follows: After the scion is grafted onto the rootstock, a brown film forms on the surface of the wound between the rootstock and scion due to the remnants of dead cells, covering the wound. Subsequently, stimulated by callus hormones, cells surrounding the wound and cambium cells divide vigorously, breaking down the brown film and forming callus tissue. As the callus tissue continues to grow, filling the gap between the scion and rootstock, the parenchyma cells of the callus tissue of the rootstock and scion connect with each other, connecting the cambium layers of the two. The callus tissue continues to differentiate, forming new xylem inwardly and new phloem outwardly, further connecting the vessels and sieve tubes, thus uniting the rootstock and scion into a unified entity. If the rootstock surface is not inspected for damage, any damage to the rootstock surface will require wound healing. This increases the consumption of nutrients and water needed for plant healing, hindering wound healing and healthy growth of the rootstock and scion after healing. It also increases the risk of bacterial infection and even necrosis at the damaged rootstock surface. This technical solution detects damage on the surface of the rootstock and calculates parameters such as the length, width and area of ​​each detected damage location through detection; further uses the parameters to judge the type of damage and the degree of its impact on grafting. If the judgment result exceeds the set threshold, it can be determined that the rootstock is not suitable for grafting operations, thereby excluding unsuitable rootstocks, effectively improving the grafting survival rate, and ensuring that good rootstocks have better healing ability at the grafted joint, facilitating healing and growth after grafting.

[0023] Step S3: Use the Hough transform linear method to detect the curvature of each rootstock. If the curvature of the selected rootstock is too large, it will not be able to be smoothly placed in the corresponding position of the cutting module. Therefore, it is necessary to detect whether the curvature of the selected rootstock meets the requirements. The specific steps of the detection are:

[0024] S3.1 performs Hough linear transform on the rootstock outline image;

[0025] S3.2 filters the detected line segments, deletes isolated line segments and small line segments, and selects two line segments that can represent the bending state of the rootstock in each rootstock placement area;

[0026] S3.3 Calculate the slope of a line segment using the coordinates of two points on the same line segment;

[0027] S3.4 uses the slope of the line segment to calculate the angle between the two line segments, and uses the angle to represent the curvature of the rootstock; when the value exceeds a set threshold, it can be determined that the rootstock cannot be used for subsequent cutting and joining operations;

[0028] The curvature detection of this technical solution is completed on the basis of the classic Hough linear transform, and the detection algorithm is highly robust. In addition, according to the operating requirements of the cutting module, it is only necessary to detect two straight line segments that can represent the bending state of the rootstock and calculate the angle between the two. Therefore, the detection algorithm has a small amount of data processing, a fast detection processing speed, and high efficiency, and is easy to use to make a fast, accurate, and effective choice of whether the rootstock meets the requirements based on the curvature.

[0029] Step S4, identifying the thickness of each rootstock; if the stock and scion have large differences in thickness, the joint will be unstable, and the cambium will be difficult to align and callus tissue will not form, which will affect the survival rate of the grafted seedlings. Selecting stock and scion with similar thickness for grafting is a prerequisite for ensuring the survival rate of the grafted seedlings; therefore, before matching with the scion, the thickness of the rootstock should be identified, and the specific identification steps are:

[0030] S4.1 traverse each V-direction rootstock placement area of ​​the rootstock contour image, use the difference in U coordinates of two contour points with the same V coordinate as a single thickness, and use the average of the single thicknesses of the same rootstock as the average thickness of the rootstock;

[0031] S4.2 traverse each U-direction rootstock placement area, taking the difference in V coordinates between two contour points with the same U coordinate as a single thickness, and taking the average of the single thicknesses as the average thickness of the rootstock;

[0032] S4.3 rotate the image so that the remaining rootstock placement areas are in the U or V direction, and calculate the single girth and average girth of each rootstock according to S4.1 or S4.2;

[0033] S4.4 Identify the upper and lower ends of each rootstock based on the characteristics that the single thickness at the upper end does not change much and the single thickness at the lower end gradually decreases; there is a bud node at the upper end of the rootstock, which is used for joining with the scion wood, and the lower end is used for rooting. The placement direction of the rootstock in the cutting module cannot be reversed, so it needs to be identified.

[0034] S4.5 identifies the single maximum thickness in the upper region of the rootstock as the bud node;

[0035] S4.6 The average value of the single thickness in the upper 1 / 3-2 / 3 area of ​​the bud node is used to identify the thickness of the rootstock;

[0036] The present invention has the following advantages for detecting the stock thickness: (1) The detection result is the thickness information of the cutting part of the stock (i.e., the grafting joint), and this information is used as the basis for the stock-scion thickness matching, which has higher matching accuracy and is more conducive to the alignment of the stock-scion cambium and the survival of the grafted seedlings. (2) This detection system can be applied to the detection of stock, and can also be applied to the detection of scion. During the detection, the scion and the stock share the same image acquisition system, which has the best detection effect. The stock (scion) thickness information can be described in pixels, without the need to obtain the actual thickness through coordinate transformation. Therefore, the detection algorithm has higher accuracy and efficiency, and is easy to implement online detection.

[0037] Step S5: Based on the conversion relationship between the image coordinate system and the workbench coordinate system, a certain position (such as 4 cm) below the bud node of the rootstock to be taken is identified as the rootstock grasping point.

[0038] Further preferred technical solutions are as follows:

[0039] Preferably, the detection table is evenly provided with 8 stock placement positions, one of which is a seedling removal position. The system can control the detection table to rotate precisely and send the stock to be taken to the seedling removal position. When the system is working, it will judge the stock that best matches the selected scion wood based on the surface damage, curvature and roughness information of the 8 detected stock, and rotate it to the seedling removal position; then, it will identify the gripping point of the stock and wait for the seedling removal manipulator to send it to the cutting module. Through the above setting, the roughness of multiple stock can be detected at one time, and the detection efficiency is higher.

[0040] Preferably, the improved Canny method adds two diagonal gradient operators on the basis of the original Canny method, and the gradient amplitude is expressed as:

[0041]

[0042]

[0043]

[0044] Among them, G is the gradient amplitude, θ is the gradient amplitude, G x , G y G is the horizontal and vertical gradient operator used by the original Canny method. 45 , G 135 To improve the Canny method, two diagonal gradient operators are added, P 11 -P 33is the 8-bit area of ​​the current pixel. Experiments show that the combination of this operator and the Gaussian filter has better results in stock edge detection than the original Canny method.

[0045] Preferably, the length, width and area parameters of the damaged position are all measured in pixels.

[0046] The present invention has the following advantages and positive effects:

[0047] (1) The system can detect parameters such as the length, width, and area of ​​cracks and cutting marks on the surface of grafted rootstocks. Based on these detection parameters, the system can easily determine the type of damage, the degree of impact, and whether it is suitable for grafting operations.

[0048] (2) The curvature of the grafted rootstock can be detected, and the system can determine whether the rootstock can meet the placement requirements in the cutting module based on this parameter.

[0049] (3) The system can identify the upper and lower ends, bud node position, thickness and clamping position of the grafted rootstock. Based on this information, the system can control the seedling-taking robot to achieve accurate seedling supply operations.

[0050] The present invention can realize automatic detection of grafted rootstocks. Compared with the existing technology, it can comprehensively detect and judge information such as surface damage, curvature, bud nodes, thickness, clamping position, etc. of the grafted rootstock, so as to quickly and accurately select rootstocks and scion woods that meet the operation requirements, thereby ensuring the operation success rate and grafting survival rate of the automatic grafting machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0052] Figure 1 This is a flow chart of the grafted rootstock image detection method of the present invention;

[0053] Figure 2 It is a schematic structural diagram of the detection platform of the present invention;

[0054] Figure 3 This is a rendering of the rootstock edge detection method of the present invention;

[0055] Figure 4 This is a rendering of the method for detecting damage to the rootstock surface of the present invention;

[0056] Figure 5 This is a rendering of the rootstock curvature detection method of the present invention.

[0057] Explanation of the accompanying numbers: 1-detection table, 2-camera, 3-seedling removal robot, 4-cutting module, 5-workbench. DETAILED DESCRIPTION

[0058] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0059] According to an embodiment of the present invention, a method for detecting an image of a grafted rootstock is provided. Figure 1-Figure 5 As shown, this embodiment includes the following steps:

[0060] Step S1, collect the images of the rootstock and the detection platform 1, grayscale, filter and binarize the images, and use the improved Canny operator to extract the edge contours of each rootstock. The specific steps are as follows:

[0061] S1.1 Collect images of the test platform 1 before and after the stock is placed using the same parameters;

[0062] S1.2 uses the rootstock image as the original image and the unrootstock image as the background image to obtain the rootstock image using the background difference method. While this method is more sensitive to scene changes than other commonly used background segmentation algorithms, it offers advantages such as high efficiency, good stability, and minimal target requirements. Because the automatic grafting machine operates indoors, the image acquisition scene and acquisition parameters are controllable, making scene changes a simple problem.

[0063] S1.3 grayscales the rootstock image using the 3r-b operator to obtain a grayscale image;

[0064] S1.4 performs Gaussian filtering on the grayscale image and performs binarization processing using the Ostu algorithm to obtain a binary image;

[0065] S1.5 inverts the binary image and marks the connected domains of each rootstock placement area;

[0066] S1.6 uses the improved Canny method to perform edge detection on the image, extracts the edge contour of each rootstock, and inverts the image again to obtain the rootstock contour image;

[0067] Step S2: Detecting the surface damage of each rootstock using morphology and planimetry. Surface damage to grape grafted rootstocks is usually manifested as cracks and cutting marks formed during the shearing process. When the damage reaches a certain level, it will make it difficult for the grafted seedlings to survive. Therefore, it is necessary to detect the degree of rootstock damage. The specific steps of the detection are as follows:

[0068] S2.1 performs a closing operation on the binary image and subtracts the result from the original image to obtain an image of the damaged area. This operation can reveal narrow, discontinuous, and elongated gullies and small holes in the connected domain.

[0069] S2.1 Use the area method to remove small, isolated areas;

[0070] S2.3 Mark the connected domains of each rootstock placement area;

[0071] S2.4 Calculate the length, width, and area parameters of each damaged location; these parameters serve as the basis for determining the damage type and its severity. When the parameters exceed a set threshold, the rootstock is deemed unsuitable for grafting.

[0072] Step S3, using the Hough transform linear method to detect the curvature of each rootstock; if the curvature of the selected rootstock is too large, it will not be able to be smoothly placed in the corresponding position of the cutting module 4; therefore, it is necessary to detect whether the curvature of the selected rootstock meets the requirements, and the specific steps of the detection are:

[0073] S3.1 performs Hough linear transform on the rootstock outline image;

[0074] S3.2 filters the detected line segments, deletes isolated line segments and small line segments, and selects two line segments that can represent the bending state of the rootstock in each rootstock placement area;

[0075] S3.3 Calculate the slope of a line segment using the coordinates of two points on the same line segment;

[0076] S3.4 uses the slope of the line segment to calculate the angle between the two line segments, and uses the angle to represent the curvature of the rootstock; when the value exceeds a set threshold, it can be determined that the rootstock cannot be used for subsequent cutting and joining operations;

[0077] Step S4, identifying the thickness of each rootstock; if the stock and scion have large differences in thickness, the joint will be unstable, and the cambium will be difficult to align and callus tissue will not form, which will affect the survival rate of the grafted seedlings. Selecting stock and scion with similar thickness for grafting is a prerequisite for ensuring the survival rate of the grafted seedlings; therefore, before matching with the scion, the thickness of the rootstock should be identified, and the specific identification steps are:

[0078] S4.1 traverse each V-direction rootstock placement area of ​​the rootstock contour image, use the difference in U coordinates of two contour points with the same V coordinate as a single thickness, and use the average of the single thicknesses of the same rootstock as the average thickness of the rootstock;

[0079] S4.2 traverse each U-direction rootstock placement area, taking the difference in V coordinates between two contour points with the same U coordinate as a single thickness, and taking the average of the single thicknesses as the average thickness of the rootstock;

[0080] S4.3 rotate the image so that the remaining rootstock placement areas are in the U or V direction, and calculate the single girth and average girth of each rootstock according to S4.1 or S4.2;

[0081] S4.4 Identify the upper and lower ends of each rootstock based on the characteristics that the single thickness at the upper end does not change much and the single thickness at the lower end gradually decreases; there is a bud node at the upper end of the rootstock, which is used for joining with the scion wood, and the lower end is used for rooting. The placement direction of the rootstock in the cutting module cannot be reversed, so it needs to be identified.

[0082] S4.5 identifies the single maximum thickness in the upper region of the rootstock as the bud node;

[0083] S4.6 The average value of the single thickness in the upper 1 / 3-2 / 3 area of ​​the bud node is used to identify the thickness of the rootstock.

[0084] Step S5: Based on the conversion relationship between the image coordinate system and the workbench coordinate system, a certain position (such as 4 cm) below the bud node of the rootstock to be taken is identified as the rootstock grasping point.

[0085] Preferably, the detection platform 1 is evenly provided with 8 stock placement positions, one of which is a seedling removal position. The system can control the detection platform to rotate accurately and send the stock to be removed to the seedling removal position. When the system is working, it will determine the stock that best matches the selected scion based on the surface damage, curvature and roughness information of the 8 detected stock, and rotate it to the seedling removal position; then, it identifies the gripping point of the stock and waits for the seedling removal robot 3 to send it to the cutting module.

[0086] Preferably, the improved Canny method adds two diagonal gradient operators on the basis of the original Canny method, and the gradient amplitude is expressed as:

[0087]

[0088]

[0089]

[0090] Among them, G is the gradient amplitude, θ is the gradient amplitude, G x , G y G is the horizontal and vertical gradient operator used by the original Canny method. 45 , G 135 To improve the Canny method, two diagonal gradient operators are added, P 11 -P 33 is the 8-bit area of ​​the current pixel. Experiments show that the combination of this operator and the Gaussian filter has better results in stock edge detection than the original Canny method.

[0091] Preferably, the length, width and area parameters of the damaged position are all measured in pixels.

[0092] This embodiment has the following advantages and positive effects:

[0093] (1) The system can detect parameters such as the length, width, and area of ​​cracks and cutting marks on the surface of grafted rootstocks. Based on these detection parameters, the system can easily determine the type of damage, the degree of impact, and whether it is suitable for grafting operations.

[0094] (2) The curvature of the grafted rootstock can be detected, and the system can determine whether the rootstock can meet the placement requirements in the cutting module based on this parameter.

[0095] (3) The system can identify the upper and lower ends, bud node position, thickness and clamping position of the grafted rootstock. Based on this information, the system can control the seedling-taking robot to achieve accurate seedling supply operations.

[0096] (4) It can realize automatic detection of grafted rootstocks, which is more intelligent, more efficient, and has more accurate and stable detection results than traditional manual methods.

[0097] The above-mentioned embodiment of the present invention can realize automatic detection of grafted rootstocks. Compared with the existing technology, it can comprehensively detect and judge the surface damage, curvature, bud nodes, thickness, clamping position and other information of the grafted rootstock, so as to quickly and accurately select the rootstock and scion wood that meet the operation requirements, thereby ensuring the operation success rate and grafting survival rate of the automatic grafting machine.

[0098] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for detecting an image of a grafted rootstock, characterized in that: The following steps are involved: Step S1: collect images of the rootstock and the detection platform, grayscale, filter and binarize the images, and use the improved Canny operator to extract the edge contours of each rootstock. The specific steps are as follows: S1.1 Collect images of the test platform before and after stock placement using the same parameters; S1.2 uses the image with the stock placed as the original image and the image without the stock placed as the background image, and obtains the stock image using the background difference method; S1.3 grayscales the rootstock image using the 3r-b operator to obtain a grayscale image; S1.4 performs Gaussian filtering on the grayscale image and performs binarization processing using the Ostu algorithm to obtain a binary image; S1.5 inverts the binary image and marks the connected domains of each rootstock placement area; S1.6 uses the improved Canny method to perform edge detection on the image, extracts the edge contour of each rootstock, and inverts the image again to obtain the rootstock contour image; Step S2, using morphology and planimetry to detect surface damage of each rootstock, specifically the following steps: S2.1 performs a closing operation on the binary image and calculates the difference between the operation result and the original image to obtain an image of the damaged area; S2.1 Use the area method to remove small, isolated areas; S2.3 Mark the connected domains of each rootstock placement area; S2.4 Calculate the length, width, and area parameters of each damaged location; Step S3, using the Hough transform linear method to detect the curvature of each rootstock, the specific steps are as follows: S3.1 performs Hough linear transform on the rootstock outline image; S3.2 filters the detected line segments, deletes isolated line segments and small line segments, and selects two line segments that can represent the bending state of the rootstock in each rootstock placement area; S3.3 Calculate the slope of a line segment using the coordinates of two points on the same line segment; S3.4 Use the slope of a line segment to calculate the angle between two line segments and use this angle to represent the curvature of the rootstock; Step S4, identifying the thickness of each rootstock, specifically the following steps: S4.1 traverse each V-direction rootstock placement area of ​​the rootstock contour image, use the difference in U coordinates of two contour points with the same V coordinate as a single thickness, and use the average of the single thicknesses of the same rootstock as the average thickness of the rootstock; S4.2 traverse each U-direction rootstock placement area, taking the difference in V coordinates between two contour points with the same U coordinate as a single thickness, and taking the average of the single thicknesses as the average thickness of the rootstock; S4.3 rotate the image so that the remaining rootstock placement areas are in the U or V direction, and calculate the single girth and average girth of each rootstock according to S4.1 or S4.2; S4.4 Identify the upper and lower ends of each rootstock based on the characteristics that the single thickness of the upper end does not change much and the single thickness of the lower end gradually decreases; S4.5 identifies the single maximum thickness in the upper region of the rootstock as the bud node; S4.6 The average value of the single thickness in the upper 1 / 3-2 / 3 area of ​​the bud node is used to identify the thickness of the rootstock; Step S5: Based on the conversion relationship between the image coordinate system and the workbench coordinate system, the lower side of the bud node of the rootstock to be taken is identified as the rootstock grasping point.

2. The image detection method of a grafted rootstock according to claim 1, characterized in that: The detection table is evenly provided with 8 rootstock placement positions, one of which is a seedling removal position. The system can control the detection table to rotate accurately and send the rootstock to be removed to the seedling removal position.

3. The image detection method of a grafted rootstock according to claim 1, characterized in that: The improved Canny method adds two diagonal gradient operators on the basis of the original Canny method, and its gradient amplitude is expressed as: Among them, G is the gradient amplitude, θ is the gradient amplitude, G x , G y G is the horizontal and vertical gradient operator used by the original Canny method. 45 , G 135 To improve the Canny method, two diagonal gradient operators are added, P 11 -P 33 The 8-bit area of ​​the current pixel.

4. The image detection method for grafted rootstock according to claim 1, wherein: The length, width and area parameters of the damaged position are all measured in pixels.

Images