Morphological description method and system for agricultural robot operation target image

Abstract

This invention provides a method and system for image morphological description of agricultural robot operation targets. The method includes: determining a first optimal matching transformation between the target and a preset target matching template; and determining the specific pose information of the target contour based on the first optimal matching transformation. The system executes the method. This invention provides a basis for the precise operation of agricultural robots by quantitatively describing the unique morphological features of agricultural robot operation targets (e.g., fruits, leaves, and inflorescences) for the precise localization of contour feature parts (e.g., pedicels, calyxes, and shoulders) and fitting of occluded contour regions.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an image morphology description method and system for an agricultural robot operation target. Background Art

[0002] In recent years, with the continuous reduction of agricultural workers, the labor cost input in typical labor links such as bagging, picking, pruning and pollination in the fruit and vegetable production process has continued to increase, which has become an important factor restricting the sustainable development of related industries. Manual observation and operation of the color, posture, position and other characteristics of the target fruits, branches and inflorescences that grow randomly and densely in clusters of plants are time-consuming and labor-intensive, and there is a lack of unified standards for the operation. Developing corresponding operation robots to assist and replace manual labor is an effective way to improve production efficiency and reduce labor cost input.

[0003] Taking fruit trees and vegetables that produce fruits as an example, objects such as fruits, leaves and inflorescences are randomly distributed in the image, with different postures, overlapping and occluding each other. In the unstructured agricultural environment, the quantitative description and fitting of the image morphology of objects such as fruits, inflorescences and stems is a necessary prerequisite for robots to operate the objects efficiently and accurately, and has become a research hotspot in the field of agricultural robots.

[0004] In recent years, machine learning models with multi-layer convolutional feature extraction networks as the core have been widely used in agricultural object recognition due to their "end-to-end" model structure, algorithm transferability, and higher recognition accuracy. They have unique advantages in perceiving and fusing multi-dimensional visual information of agricultural objects. The output of target detection and segmentation results is limited to judging whether the target area exists, and lacks a quantitative description of the regional morphology. However, target morphological information is a necessary basis for operations such as fruit picking, pollination, and bagging. Therefore, the current traditional target detection and segmentation methods cannot meet the operational needs of agricultural robots. Summary of the invention

[0005] The image morphological description method and system of the agricultural robot's working targets provided by the present invention are used to solve the above-mentioned problems existing in the prior art. By quantitatively describing the unique morphological features of the agricultural robot's working targets (such as fruits, leaves and inflorescences), the contour feature parts of the targets (such as stalks, calyxes and shoulders) are accurately located, thereby providing a basis for the precise operation of the target production robot.

[0006] The present invention provides an image morphology description method for an agricultural robot operation target, comprising:

[0007] Determining a first best matching transformation between the target and a preset target matching template;

[0008] Determining specific position and posture information of a contour portion of the target according to the first best matching transformation;

[0009] Wherein, the specific posture information is used to describe the image morphology of the target;

[0010] The targets include agricultural robot operation targets.

[0011] According to an image morphology description method for an agricultural robot operation target provided by the present invention, the first best matching transformation between the determined target and a preset target matching template includes:

[0012] Determining a contour feature of the target according to a first edge point gradient distribution feature of the target contour;

[0013] Determining the contour feature of the target according to the gradient distribution feature of the first edge point of the contour part of the target;

[0014] According to the difference between the contour features of the target and the contour features of the preset target matching template, the relative position and posture relationship between the target and the preset target matching template is solved to determine the first best matching transformation.

[0015] According to the image morphology description method of an agricultural robot operation target provided by the present invention, the preset target matching template is obtained in the following manner:

[0016] The contour feature of any target to be matched in the initial sample set is used as an initial matching template;

[0017] Determine a second best matching transformation between the remaining to-be-matched objects in the initial sample set and the initial matching template;

[0018] updating the initial sample set according to the contour features of the remaining to-be-matched targets after the second optimal transformation and the initial matching template;

[0019] Performing a third best matching transformation on the average shape of all to-be-matched objects in the updated initial sample set and the initial matching template to update the average shape;

[0020] Determine a fourth best matching transformation between the second best matching transformation of the remaining objects to be matched and the updated average shape;

[0021] According to the contour features of the remaining to-be-matched targets after the fourth optimal transformation and the initial matching template, the initial sample set is updated until the updated average shape converges, and the converged average shape is used as the preset target matching template.

[0022] According to an image morphology description method for an agricultural robot operation target provided by the present invention, determining specific position information of the target contour according to the first best matching transformation includes:

[0023] Determine the relative translation vector, relative rotation matrix and relative scaling factor between the target and the preset target matching template according to the relative posture relationship;

[0024] The specific posture information is determined according to the relative translation vector, the relative rotation matrix and the relative scaling factor.

[0025] According to an image morphology description method of an agricultural robot operation target provided by the present invention, the gradient distribution feature of the first edge point of the target contour is obtained in the following manner:

[0026] Inputting the image corresponding to the target into a preset segmentation model to obtain the pixel area of ​​the target in the image;

[0027] Perform edge detection on the pixel area to obtain the first edge point gradient distribution feature.

[0028] According to the present invention, a method for describing the image morphology of an agricultural robot operation target further comprises:

[0029] According to the edge point gradient distribution characteristics of the visible contour in the target and the preset target matching template, a fifth best matching transformation of the edge point gradient distribution characteristics of the visible contour is determined, and the third edge point gradient distribution characteristics of the obscured contour of the target are fitted.

[0030] The present invention also provides a target image description system, comprising: a data acquisition module and a morphology description module;

[0031] The data acquisition module is used to determine a first best matching transformation between the target and a preset target matching template;

[0032] The morphological description module is used to determine specific position and posture information of the target contour according to the first best matching transformation;

[0033] Wherein, the specific posture information is used to describe the image morphology of the target;

[0034] The targets include agricultural robot operation targets.

[0035] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, an image morphological description method for an agricultural robot operation target as described in any one of the above-mentioned methods is implemented.

[0036] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the image morphological description method for the agricultural robot's operating target as described in any one of the above-mentioned methods.

[0037] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements the image morphological description method of the agricultural robot's operating target as described in any one of the above-mentioned methods.

[0038] The image morphological description method and system of the agricultural robot's working targets provided by the present invention can be used to accurately locate the contour feature parts of the targets (such as stalks, calyxes and shoulders) by quantitatively describing the unique morphological features of the agricultural robot's working targets (such as fruits, leaves and inflorescences), thereby providing a basis for the precise operation of the target production robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 It is a flow chart of the image morphology description method of the agricultural robot operation target provided by the present invention;

[0041] Figure 2 It is a schematic diagram of key edge points and gradients of a fruit provided by the present invention;

[0042] Figure 3 It is a schematic diagram of the dominant gradient direction provided by the present invention;

[0043] Figure 4 It is a schematic diagram of quantitative description of fruit morphology based on segmentation model matching provided by the present invention;

[0044] Figure 5 It is a fruit contour fitting schematic diagram provided by the present invention;

[0045] Figure 6 It is a structural schematic diagram of the image morphology description system of the agricultural robot operation target provided by the present invention;

[0046] Figure 7 It is a schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] The purpose of the present invention is to provide an image morphological description method for an agricultural robot's operating target, provide a basis for the agricultural robot to obtain the morphological description of the operating target, and improve the recognition accuracy of randomly distributed and mutually occluded targets in images of different scales in unstructured environments. In particular, taking fruit targets as an example, a general description method for the contour curve of cone images is proposed, which is used for contour morphological fitting of the occluded area of ​​the fruit, and realizes the segmentation and recognition of occluded and adhered fruits. The specific implementation is as follows:

[0049] Figure 1 FIG. 1 is a flow chart of the image morphology description method of the agricultural robot operation target provided by the present invention, such as Figure 1 As shown, the method includes:

[0050] S1, determining a first best matching transformation between a target and a preset target matching template;

[0051] S2. Determine specific position and posture information of the target contour part according to the first best matching transformation;

[0052] Among them, the specific posture information is used to describe the image morphology of the target;

[0053] The objectives include agricultural robot operation objectives.

[0054] It should be noted that the execution subject of the above method may be a computer device.

[0055] Optionally, the target in the present invention may be specifically an agricultural robot operation target such as fruit, leaf and inflorescence. The following takes the target as fruit as an example to describe the image morphology of the agricultural robot operation target provided by the present invention in detail, specifically:

[0056] Under natural growth conditions, fruit trees have dense branches and leaves, and the fruits have different postures and are mutually adhered and blocked. In order to achieve precise manipulation of the fruit, it is necessary not only to obtain its visible image area, but also to fit its invisible area to locate its overall shape. However, the current typical semantic segmentation deep learning model is limited to segmenting the visible area of ​​the target and lacks a quantitative description of the target morphology. Therefore, it is necessary to improve the existing semantic segmentation deep learning model so that it has the ability to output fruit morphological feature parameters "end-to-end".

[0057] Based on the obtained fruit pixel area, the morphological description of its contour is a necessary basis for the production robot to accurately implement operations. For example, the "similar shape" feature of the apple fruit contour is manifested as: a closed contour composed of discrete edge points, and the distribution law of its edge gradient direction is similar. Therefore, it is believed that edge points with the same (or similar) gradient direction on different fruit contours are matching feature points. In order to accurately characterize the gradient direction characteristics of edge points, the dominant gradient direction of the specific neighborhood of the point is used as the gradient direction of the point (such as Figure 2 , Figure 3 shown).

[0058] Assume that the preset target (fruit i) matching template is L, and the matching transformation of fruit i to the preset fruit matching template L is T i , using mathematical methods, find the best matching transformation of fruit i to the preset fruit matching template L (ie the first best matching transform).

[0059] According to the obtained fruit i, the first best match transformation to the preset fruit matching template L is performed. Determine the edge point gradient distribution characteristics of the contour part of fruit i, such as the edge point position and the edge point gradient direction. According to the edge point position and the edge point gradient direction of the contour part of fruit i, the edge point position and the edge point gradient direction of the contour part of fruit i (that is, the specific position information of the contour part of fruit i) can be determined, thereby realizing the relative position positioning of the specific contour part of fruit i (such as the stalk, calyx and shoulder) to realize the morphological description of the contour part of fruit i, and providing a basis for the precise operation of the production robot.

[0060] It should be noted that different preset matching templates are required when locating specific contour parts of different targets such as fruits, leaves and inflorescences.

[0061] The image morphological description method of the agricultural robot's working target provided by the present invention, through quantitative description of the singular morphological features of the agricultural robot's working target (such as fruits, leaves and inflorescences), is used for accurate positioning of the target's contour feature parts (such as stalks, calyxes and shoulders), as well as fitting of obscured contour areas, thereby providing a basis for the precise operation of the target production robot.

[0062] Furthermore, in one embodiment, step S1 may specifically include:

[0063] S11, determining the contour feature of the target according to the gradient distribution feature of the first edge point of the target contour;

[0064] S12. According to the difference between the contour features of the target and the contour features of the preset target matching template, a relative position and posture relationship between the target and the preset target matching template is solved to determine a first best matching transformation.

[0065] Furthermore, in one embodiment, step S2 may specifically include:

[0066] S21, determining the relative translation vector, relative rotation matrix and relative scaling factor between the target and the preset target matching template according to the relative posture relationship;

[0067] S22. Determine specific posture information according to the relative translation vector, relative rotation matrix and relative scaling factor.

[0068] Optionally, suppose the outline of fruit i is composed of N edge point gradients (i.e., the first edge point gradient distribution features, including edge point positions and edge point gradient directions) (x n ,y n ), so the template of fruit i is l i , expressed as Let the preset fruit matching template be L, and the matching transformation of fruit i to the preset fruit matching template L be T i , the relative position relationship between the fruit i and the preset fruit matching template L (including the relative translation vector t i , relative rotation matrix R i (θ) and the relative scaling factor s i ) is expressed as follows:

[0069]

[0070] The difference in posture between fruit i and the preset fruit matching template L is E, which can be expressed as E = [l i -l] T [l i -l], so that it is unknown variables s i ,θ i The partial differential is zero, and the relative translation vector t can be obtained i , relative rotation matrix R i (θ) and the relative scaling factor s i The specific value of the relative translation vector t between the obtained fruit and the preset fruit matching module i , relative rotation matrix R i (θ) and the relative scaling factor s i Substituting the specific value of into the above formula, we can get the first best matching transformation of fruit i to the preset fruit matching template L

[0071] And transform according to the first best match obtained The relative translation vector t between the fruit in the image and the preset fruit matching module i , relative rotation matrix R i (θ) and the relative scaling factor s i , determine the specific posture information of the fruit.

[0072] The image morphological description method of the agricultural robot's working target provided by the present invention realizes a quantitative description of the target's singular morphological features by transforming the agricultural robot's working target (such as a fruit) into the best matching transformation between the preset target matching templates, thereby laying a foundation for the subsequent precise positioning of the contour parts of the fruit (such as the stalk, calyx and shoulder) and providing a basis for the precise operation of the agricultural target production robot.

[0073] Furthermore, in one embodiment, the preset target matching template in step S1 is obtained in the following manner:

[0074] Step A: taking the contour feature of any target to be matched in the initial sample set as the initial matching template;

[0075] Step B, determining the second best matching transformation between the remaining to-be-matched targets in the initial sample set and the initial matching template;

[0076] Step C, updating the initial sample set according to the contour features of the remaining to-be-matched targets after the second best transformation and the initial matching template;

[0077] Step D: performing a third best matching transformation on the average shape of all the objects to be matched in the updated initial sample set and the initial matching template to update the average shape;

[0078] Step E, determining the fourth best matching transformation between the second best matching transformation of the remaining objects to be matched and the updated average shape;

[0079] Step F: updating the initial sample set according to the contour features of the remaining targets to be matched after the fourth optimal transformation and the initial matching template until the updated average shape converges, and using the converged average shape as the preset target matching template.

[0080] Optionally, the preset target matching template is obtained by training through the following steps, specifically:

[0081] Step 1: Set the contour features of M targets to be matched (if real) {l 1 ,l 2 ,...,l M} to form an initial sample set, and select any contour feature of the target to be matched from the initial sample set as the initial matching template, for example, l1 , thus we can get the remaining targets to be matched {l 2 ,l 3 ,...,l M} and the initial matching template l 1 The best matching transformation (ie the second best matching transform).

[0082] Step 2: According to the contour features of the remaining targets to be matched after the second best transformation and the initial matching template l 1 , for the initial sample set {l 1 ,l 2 ,...,l M} is updated to obtain a new sample set:

[0083] Step 3: Calculate the updated sample set The shape average The average shape is obtained by adding the edge point gradient distribution features of the contour features of each target to be matched in the updated sample set and averaging them. And the average shape With the initial matching template l 1 Take the best matching transformation to get the third best matching transformation and use it as the new average shape The average shape originally calculated to update.

[0084] Step 4: Set the remaining targets to be matched {l 2 ,l 3 ,...,l M The second best matching transformation of The updated average shape is the new average shape Perform the best matching transformation to obtain the second best matching transformation of the remaining targets to be matched With the updated average shape The best matching transformation between (ie, the fourth best matching transformation).

[0085] Step 5: According to the contour features of the remaining objects to be matched after the fourth optimal transformation and the initial matching template l 1 , for the initial sample set {l 1 ,l 2 ,...,l M} is updated until the average shape after update Convergence, i.e. when the average shape after the update When the coordinate changes of the gradient distribution characteristics of the edge points of the corresponding fruit tend to be stable, the updated average shape is determined Converge, and use the converged average shape as the preset target matching template L.

[0086] Step 6: If the updated average shape If it does not converge, repeat steps 3-5 until the updated average shape is convergence.

[0087] The image morphological description method of the agricultural robot operation target provided by the present invention obtains a preset target (if real) matching template through training, which lays a foundation for subsequently determining the best matching transformation of the target based on the preset target matching template and realizing the quantitative description of the target's singular morphological features.

[0088] Further, in one embodiment, the first edge point gradient distribution feature of the contour part of the target in step S11 is obtained by:

[0089] Step 1: Input the image corresponding to the target into a preset segmentation model to obtain the pixel area of ​​the target in the image;

[0090] Step 2: Perform edge detection on the pixel area to obtain the first edge point gradient distribution feature.

[0091] Alternatively, see Figure 4 , the image corresponding to fruit i is input into the preset segmentation model (such as Mask RCNN) to obtain the pixel area of ​​fruit i in the image. By performing edge detection on the pixel area, a pixel point has different gradients in different directions. In order to improve the computational efficiency, only the largest gradient of the edge pixel point, that is, the dominant gradient, is used as the first edge point gradient distribution feature Then according to the gradient distribution feature of the first edge point and the first best matching transformation The quantitative image morphological description parameters such as the scale scaling ratio of target i (determined by the obtained relative scaling factor) and the center point (determined by the obtained relative rotation matrix and relative translation vector) are obtained to determine the specific position information of the outline (fruit stalk, calyx and shoulder) of fruit i.

[0092] The image morphological description method of the agricultural robot working target provided by the present invention aims at the singular morphological features of the agricultural robot working target (such as fruit), realizes the quantitative description of the target contour part through image recognition of the target and the gradient distribution characteristics of the edge points of the contour part of the identified target.

[0093] Furthermore, in one embodiment, the method further comprises:

[0094] S3. Determine a fifth best matching transformation of the edge point gradient distribution characteristics of the visible contour according to the edge point gradient distribution characteristics of the visible contour in the target and a preset target matching template, and fit the third edge point gradient distribution characteristics of the obscured contour of the target.

[0095] Optionally, according to the fifth best matching transformation of the edge point gradient distribution feature of the visible contour in the fruit i and the preset target matching template L, the third edge point gradient distribution feature of the obscured contour of the target i is determined, specifically:

[0096] If the edge point gradient distribution characteristics of the visible contour of the known target i are transformed to the fifth best match of the preset target matching template L, the image morphological parameters of the target i (i.e., the first edge point gradient distribution characteristics, which include the visible and invisible edge point positions and edge point gradient directions) can be obtained. For the invisible area of ​​overlapping fruits, the transformation relationship relative to the fifth best match transformation can be solved based on the edge point position coordinates of the known limited visible contour parts, and then the complete edge point positions of the fifth best match transformation are inversely transformed to estimate the unmatched edge point positions, thereby achieving fitting of the invisible contour area (see Figure 5 ).

[0097] The image morphological description method of the agricultural robot's working target provided by the present invention, based on the existing target recognition model, further realizes the quantitative description of the singular morphological features of the agricultural robot's working target and the fitting of the occluded contour area, thereby providing a basis for the precise positioning of targets such as fruits, leaves and inflorescences that are clustered, disorderly, mutually occluded and have different postures, and providing a basis for the intelligent and precise operation of the production robot.

[0098] The image morphology description system of the agricultural robot working target provided by the present invention is described below. The image morphology description system of the agricultural robot working target described below and the image morphology description method of the agricultural robot working target described above can be referenced to each other.

[0099] Figure 6 FIG. 1 is a schematic diagram of the structure of the image morphology description system for the agricultural robot operation target provided by the present invention. Figure 6 As shown, including:

[0100] Data acquisition module 610 and morphology description module 611;

[0101] A data acquisition module 610 is used to determine a first best matching transformation between a target and a preset target matching template;

[0102] A morphological description module 611 is used to determine specific position and posture information of the target contour according to the first best matching transformation;

[0103] Among them, the specific posture information is used to describe the image morphology of the target;

[0104] The objectives include agricultural robot operation objectives.

[0105] The image morphological description system of the agricultural robot's operating targets provided by the present invention can be used for the precise positioning of the target's contour feature parts (such as stalks, calyxes and shoulders) and the fitting of obscured contour areas through quantitative description of the unique morphological features of the agricultural robot's operating targets (such as fruits, leaves and inflorescences), thereby providing a basis for the precise operation of the target production robot.

[0106] Figure 7 is a schematic diagram of the physical structure of an electronic device provided by the present invention, such as Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 711, a memory 712 and a bus 713, wherein the processor 710, the communication interface 711 and the memory 712 communicate with each other through the bus 713. The processor 710 may call the logic instructions in the memory 712 to execute the following method:

[0107] Determining a first best matching transformation between the target and a preset target matching template;

[0108] Determine specific position and posture information of the target contour according to the first best matching transformation;

[0109] Among them, the specific posture information is used to describe the image morphology of the target;

[0110] The objectives include agricultural robot operation objectives.

[0111] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer power screen (which can be a personal computer, a server, or a network power screen, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0112] Furthermore, the present invention discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the image morphology description method of the agricultural robot operation target provided by the above-mentioned method embodiments, for example, including:

[0113] Determining a first best matching transformation between the target and a preset target matching template;

[0114] Determine specific position and posture information of the target contour according to the first best matching transformation;

[0115] Among them, the specific position and posture information is used to describe the image morphology of the target;

[0116] The objectives include agricultural robot operation objectives.

[0117] On the other hand, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the image morphology description method of the agricultural robot operation target provided in the above embodiments, for example, including:

[0118] Determining a first best matching transformation between the target and a preset target matching template;

[0119] Determine specific position and posture information of the target contour according to the first best matching transformation;

[0120] Among them, the specific posture information is used to describe the image morphology of the target;

[0121] The objectives include agricultural robot operation objectives.

[0122] The system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0123] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer power screen (which can be a personal computer, a server, or a network power screen, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for describing the image morphology of an agricultural robot operation target, characterized in that: include: Determining a first best matching transformation between the target and a preset target matching template; Determining specific position and posture information of the target contour according to the first best matching transformation; Wherein, the specific posture information is used to describe the image morphology of the target; The objectives include agricultural robot operation objectives; The preset target matching template is obtained in the following manner: The contour feature of any target to be matched in the initial sample set is used as an initial matching template; Determine a second best matching transformation between the remaining to-be-matched objects in the initial sample set and the initial matching template; updating the initial sample set according to the contour features of the remaining to-be-matched targets after the second optimal transformation and the initial matching template; Performing a third best matching transformation on the average shape of all the objects to be matched in the updated initial sample set and the initial matching template to update the average shape, wherein the average shape is obtained by adding and averaging the edge point gradient distribution features of the contour features of each object to be matched in the updated sample set; Determine a fourth best matching transformation between the remaining to-be-matched objects after the second best matching transformation and the updated average shape; According to the contour features of the remaining to-be-matched targets after the fourth optimal transformation and the initial matching template, the initial sample set is updated until the updated average shape converges, and the converged average shape is used as the preset target matching template.

2. The image morphology description method of the agricultural robot operation target according to claim 1 is characterized in that: The determining of a first best matching transformation between the target and a preset target matching template comprises: Determining the contour feature of the target according to the gradient distribution feature of the first edge point of the target contour; According to the difference between the contour features of the target and the contour features of the preset target matching template, the relative position and posture relationship between the target and the preset target matching template is solved to determine the first best matching transformation.

3. The image morphology description method of the agricultural robot operation target according to claim 2 is characterized in that: Determining specific position and posture information of the target contour according to the first best matching transformation includes: Determine the relative translation vector, relative rotation matrix and relative scaling factor between the target and the preset target matching template according to the relative posture relationship; The specific posture information is determined according to the relative translation vector, the relative rotation matrix and the relative scaling factor.

4. The image morphology description method of the agricultural robot operation target according to claim 2 is characterized in that: The first edge point gradient distribution feature of the target contour is obtained by: Inputting the image corresponding to the target into a preset segmentation model to obtain the pixel area of ​​the target in the image; Perform edge detection on the pixel area to obtain the first edge point gradient distribution feature.

5. The image morphology description method of the agricultural robot operation target according to any one of claims 1 to 4, characterized in that: Also includes: According to the edge point gradient distribution characteristics of the visible contour in the target and the preset target matching template, a fifth best matching transformation of the edge point gradient distribution characteristics of the visible contour is determined, and the third edge point gradient distribution characteristics of the obscured contour of the target are fitted.

6. An image morphology description system for agricultural robot operation targets, characterized in that: include: Data acquisition module and morphological description module; The data acquisition module is used to determine a first best matching transformation between the target and a preset target matching template; The morphological description module is used to determine specific position and posture information of the target contour according to the first best matching transformation; Wherein, the specific posture information is used to describe the image morphology of the target; The objectives include agricultural robot operation objectives; The preset target matching template is obtained in the following manner: The contour feature of any target to be matched in the initial sample set is used as an initial matching template; Determine a second best matching transformation between the remaining to-be-matched objects in the initial sample set and the initial matching template; updating the initial sample set according to the contour features of the remaining to-be-matched targets after the second optimal transformation and the initial matching template; Performing a third best matching transformation on the average shape of all the objects to be matched in the updated initial sample set and the initial matching template to update the average shape, wherein the average shape is obtained by adding and averaging the edge point gradient distribution features of the contour features of each object to be matched in the updated sample set; Determine a fourth best matching transformation between the remaining to-be-matched objects after the second best matching transformation and the updated average shape; According to the contour features of the remaining to-be-matched targets after the fourth optimal transformation and the initial matching template, the initial sample set is updated until the updated average shape converges, and the converged average shape is used as the preset target matching template.

7. An electronic device comprising a processor and a memory storing a computer program, characterized in that: When the processor executes the computer program, the image morphological description method of the agricultural robot operation target described in any one of claims 1 to 5 is implemented.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the image morphological description method of the agricultural robot operation target as described in any one of claims 1 to 5 is implemented.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, it implements the morphological description method of the agricultural robot operation target image as described in any one of claims 1 to 5.

Images