Fruit and vegetable picking and screening method and system based on visual image
By receiving and processing basic three-dimensional models of fruits and vegetables, combined with deep learning algorithms and three-dimensional spatial position algorithms, the problem of slow picking and screening speed in traditional technologies is solved, and more efficient and accurate fruit and vegetable screening is achieved.
Patent Information
- Application Number
- CN202411956777.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-29
- Publication Date
- 2025-05-16
AI Technical Summary
The traditional fruit and vegetable picking and screening technology combined with visual image recognition and automation equipment is working at a low speed, making it difficult to accurately determine the location of fruits and vegetables in three-dimensional space.
By receiving the basic three-dimensional model of the collected object, the model is simplified to determine the spatial line diagram, the picture sample library and deep learning algorithm are called, the line features are extracted, and the spatial location of the line features is calculated through the three-dimensional spatial position algorithm, thereby realizing accurate screening of fruits and vegetables.
It improves the efficiency and accuracy of fruit and vegetable picking and screening, can effectively identify and determine the location of fruits and vegetables in three-dimensional space, and improves the working speed.
Smart Images
Figure CN120014038A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of program algorithm technology, and in particular to a method and system for picking and screening fruits and vegetables based on visual images. Background Art
[0002] When fruits and vegetables grow naturally, their branches, leaves and fruits are intertwined, making them difficult to pick by machine. Traditional machine picking of fruits and vegetables often uses visual image recognition and automated equipment to pick fruits and vegetables. Among them, the working speed of the combination of visual image recognition and automated equipment is not high. The reason for the low working speed of the combination of visual image recognition and automated equipment is that when fruits and vegetables grow naturally, their branches, leaves and fruits are intertwined, and it is difficult for automated equipment to determine their position in three-dimensional space through visual image recognition technology. To this end, it is necessary to propose a fruit and vegetable picking and screening method and system based on visual images to address the defect of low working speed of the traditional fruit and vegetable picking and screening technology combined with visual image recognition and automated equipment. Summary of the invention
[0003] Based on this, it is necessary to propose a fruit and vegetable picking and screening method and system based on visual images to address the defect of low working speed of traditional fruit and vegetable picking and screening technology combining visual image recognition and automation equipment.
[0004] The present application provides a fruit and vegetable picking and screening method based on visual images, comprising:
[0005] Receive a basic three-dimensional model of the acquisition object;
[0006] Simplify the basic three-dimensional model of the acquisition object to determine a spatial line drawing of the basic three-dimensional model of the acquisition object;
[0007] Call the image sample library;
[0008] Based on the edge detection algorithm, obtain the line images of all the images in the image sample library;
[0009] Use deep learning algorithms to extract line features of linear images;
[0010] Based on the spatial line drawing of the basic three-dimensional model of the acquisition object, determining the line features that match the basic three-dimensional model of the acquisition object;
[0011] Based on the three-dimensional spatial position algorithm of the spatial line graph and model, the spatial position of each line feature is calculated.
[0012] Furthermore, the basic three-dimensional model of the acquisition object includes one or more of a three-dimensional model of a branch, a three-dimensional model of a leaf, and a three-dimensional data model of a fruit.
[0013] Further, a two-dimensional plane in three-dimensional space is established;
[0014] Determine the moving step size of the two-dimensional plane;
[0015] Generate N replicas using the moving step and the two-dimensional plane; N is a positive integer;
[0016] Add all copied faces to the cutting library.
[0017] Further, according to the direction of any coordinate axis, a copy surface in the cutting library is selected in turn;
[0018] receiving the intersection of the three-dimensional model and the copied surface;
[0019] Based on the edge extraction algorithm, the edge line of the intersection surface is extracted;
[0020] Return to the above step of selecting a copy surface in the cutting library in turn according to the direction of any coordinate axis until all the copy surfaces are selected;
[0021] The gradient algorithm and the direction of the coordinate axis are called to determine the spatial line drawing of the basic three-dimensional model of the acquisition object.
[0022] Further, receiving a standard specification;
[0023] Selecting a spatial line drawing of a basic three-dimensional model of an acquisition object that meets standard specifications to obtain a target spatial line drawing; the spatial line drawing has a three-dimensional dimension of a three-dimensional space;
[0024] In three-dimensional space, select a visual center coordinate;
[0025] Generate a two-dimensional visual map using the visual center coordinates and the target space line map;
[0026] Return to the above and select a visual center coordinate in the three-dimensional space until all visual center coordinates are selected.
[0027] Furthermore, the working radius of the fruit and vegetable picking and screening system using visual images is called;
[0028] Based on the working radius and the three-dimensional center coordinates of the target space line graph, a selection radius of the visual center coordinates is generated;
[0029] Receive the minimum distance length of adjacent visual center coordinates;
[0030] Within the three-dimensional space range of the selection radius of the visual center coordinate, M visual center coordinates to be selected are generated; M is a positive integer.
[0031] Further, a two-dimensional plane visual map is called;
[0032] Using discrete differential operators, the characteristics of the two-dimensional visual image are determined;
[0033] Select a feature of the 2D visual map;
[0034] Select a line feature;
[0035] Determine whether the features of the two-dimensional visual image match the line features;
[0036] If the features of the two-dimensional plane visual image match the line features, a mapping relationship is formed between the features of the two-dimensional plane visual image and the line features, and the method of selecting a line feature is returned until all line features are selected;
[0037] If the features of the two-dimensional plane visual image do not match the line features, return to the step of selecting a line feature until all line features are selected;
[0038] Returning to selecting a feature of the two-dimensional plane visual image until all features are selected;
[0039] The line features with mapping relationships are included in the line feature data set of the spatial position to be calculated.
[0040] Further, a line feature of the line feature data set is selected;
[0041] Using discrete differential operators, the differential operators of line features are obtained;
[0042] Return a line feature of the selected line feature data set until all line features are selected;
[0043] Compute the characteristic equations of all differential operators;
[0044] Map the solution of each characteristic equation to a two-dimensional visual image.
[0045] Further, select a line image;
[0046] Based on the discrete differential operator, the differential operator of the line image is obtained;
[0047] Incorporate the differential operator into the characteristic equation and obtain the value solution of the characteristic equation;
[0048] Select a solution to the characteristic equation;
[0049] Determine whether the value solution of the characteristic equation is consistent with the solution of the characteristic equation;
[0050] If the value solution of the characteristic equation does not match the solution of the characteristic equation, then returning to the step of selecting a solution of the characteristic equation until all solutions of the characteristic equation are selected;
[0051] If the value solution of the characteristic equation matches the solution of the characteristic equation, a one-to-one mapping relationship is formed between the line image and the two-dimensional plane visual image, and the selection of a solution to the characteristic equation is returned until all solutions to the characteristic equation are selected.
[0052] The present application also provides a fruit and vegetable picking and screening system based on visual images, comprising:
[0053] A processor, used to execute the fruit and vegetable picking and screening method based on visual images;
[0054] A camera, communicatively connected to the processor;
[0055] A mechanical claw is communicatively connected to the processor.
[0056] The present application relates to a fruit and vegetable picking and screening method and system based on visual images. By receiving a basic three-dimensional model of a collection object, such as a three-dimensional model of a tender branch to be collected, a three-dimensional model of tender leaves to be collected, and a three-dimensional model of a mature fruit to be collected, the three-dimensional contour lines of the basic three-dimensional model of the collection object can be obtained to obtain the curve shape and curve size data of the basic three-dimensional model of the collection object at different viewing distances and different angles. The image sample library is called, and based on the edge detection algorithm, the line images of all the images in the image sample library are obtained, and the line features of the line images of the line images are extracted using a deep learning algorithm. The curve shape is compared with the line features of the line images, and the viewing distance and angle of the basic three-dimensional model of the collection object corresponding to the line images are determined. BRIEF DESCRIPTION OF THE DRAWINGS
[0057] The drawings constituting a part of this application are used to provide a further understanding of this application, so that other features, purposes and advantages of this application become more obvious. The illustrative embodiment drawings and their descriptions of this application are used to explain this application and do not constitute an improper limitation on this application.
[0058] Figure 1 A method flow chart of a method for picking and screening fruits and vegetables based on visual images provided in one embodiment of the present application.
[0059] Figure 2 A two-dimensional plane and a replicated surface diagram of a method for picking and screening fruits and vegetables based on visual images provided in one embodiment of the present application.
[0060] Figure 3 A basic three-dimensional model and a replicated surface diagram of a collection object for a fruit and vegetable picking and screening method based on visual images provided in one embodiment of the present application.
[0061] Figure 4 A three-dimensional vector differential operator graph of a fruit and vegetable picking and screening method based on visual images provided in one embodiment of the present application.
[0062] Figure 5 A structural connection diagram of a fruit and vegetable picking and screening system based on visual images provided in one embodiment of the present application.
[0063] Reference numerals:
[0064] 100, processor; 200, camera; 300, mechanical claw. DETAILED DESCRIPTION
[0065] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0066] The present application provides a fruit and vegetable picking and screening method based on visual images.
[0067] like Figure 1 As shown, in one embodiment of the present application, a fruit and vegetable picking and screening method based on visual images includes:
[0068] S100, receiving a basic three-dimensional model of an acquisition object.
[0069] S200, simplifying the basic three-dimensional model of the acquisition object to determine a spatial line drawing of the basic three-dimensional model of the acquisition object.
[0070] S300, calling the image sample library.
[0071] S400, based on the edge detection algorithm, obtains the line images of all the images in the image sample library.
[0072] Specifically, we first read an image and convert it into a grayscale image. Then we use the Canny algorithm to detect edges and get a binary image of the lines. Finally, we display the result.
[0073] The specific process of executing the Canny algorithm is to apply a Gaussian filter to smooth the image to remove noise and reduce interference with edge detection. Use the Sobel operator or other gradient operators to calculate the gradient magnitude and direction of each pixel in the image. Perform non-maximum suppression in the gradient direction to refine the edge and retain the pixels with the largest local gradient. Set two thresholds (high threshold and low threshold) and divide the pixels into strong edges, weak edges, and non-edges according to the gradient magnitude. Pixels above the high threshold are considered strong edges, pixels below the low threshold are considered non-edges, and pixels between the two are considered weak edges. By tracing the connection between weak edges and strong edges, the weak edges are connected to the strong edges to form a complete edge contour.
[0074] S500, uses deep learning algorithms to extract line features of linear images.
[0075] Specifically, the light intensity distribution of the light strip cross section is similar to the Gaussian distribution, so the center of the light strip is the vertex of the Gaussian distribution. The image is regarded as a function z(x, y), which is used to represent the grayscale values of different pixels in the image. In Z(x, y), the grayscale distribution of the light strip cross section also has a similar pattern, that is, the lines will appear in the function graph in the form of peaks and gullies. So our basic idea is to find this vertex by Taylor expanding the image function z(x, y) and then approximating the function with a polynomial.
[0076] These vertices can become line features in two-dimensional coordinates, and the line features can be parameterized.
[0077] S600, based on the spatial line graph of the basic three-dimensional model of the acquisition object, determining line features that match the basic three-dimensional model of the acquisition object.
[0078] S700,calculates the spatial position of each line feature based on the three-dimensional spatial position algorithm of the spatial line graph,model.
[0079] This embodiment relates to a fruit and vegetable picking and screening method based on visual images. By receiving a basic three-dimensional model of a collection object, such as a three-dimensional model of a tender branch to be collected, a three-dimensional model of tender leaves to be collected, and a three-dimensional model of a mature fruit to be collected, the three-dimensional contour lines of the basic three-dimensional model of the collection object can be obtained to obtain the curve shape and curve size data of the basic three-dimensional model of the collection object at different viewing distances and different angles. The image sample library is called, and based on the edge detection algorithm, the line images of all the images in the image sample library are obtained, and the line features of the line images are extracted using a deep learning algorithm. The curve shape is compared with the line features of the line images, and the viewing distance and angle of the basic three-dimensional model of the collection object corresponding to the line images are determined.
[0080] In one embodiment of the present application, S100 includes:
[0081] S110, collecting the basic three-dimensional model of the object including one or more of a three-dimensional model of a branch, a three-dimensional model of a leaf, and a three-dimensional data model of a fruit.
[0082] Specifically, the basic three-dimensional model of the collection object can be a three-dimensional model of tender branches to be collected, a three-dimensional model of tender leaves to be collected, or a three-dimensional model of mature fruits to be collected. The three-dimensional contour lines of the basic three-dimensional model of the collection object can be obtained to obtain the curve shape and curve size data of the basic three-dimensional model of the collection object at different viewing distances and angles.
[0083] This greatly improves the efficiency of identifying tender branches to be collected, tender leaves to be collected, and ripe fruits to be collected in the fruit and vegetable picking screening method based on visual images.
[0084] In one embodiment of the present application, S200 includes:
[0085] S211, establish a two-dimensional plane in a three-dimensional space.
[0086] S212, determining the moving step length of the two-dimensional plane.
[0087] S213, using the moving step length and the two-dimensional plane, generate N copy surfaces, where N is a positive integer.
[0088] S214, all the copied surfaces are included in the cutting library.
[0089] Specifically, Figure 2 As shown, a two-dimensional plane is established in a three-dimensional space, such as the three coordinate axes of the three-dimensional space, the X axis, the Y axis and the Z axis. The origins of the three axes are the same point, which is point P. The two-dimensional plane can be the plane (X, Y, O), and the moving step is in the Z axis direction, 0.1 unit length, that is, the copy surface 1 is (X, Y, 0.1), the copy surface 2 is (X, Y, 0.2), until the copy surface N is (X, Y, N / 10).
[0090] Copy surface 1 to copy surface N can be included in the cutting library.
[0091] In one embodiment of the present application, S200 further includes:
[0092] S221, selecting a copy surface in the cutting library in turn according to the direction of any coordinate axis.
[0093] Specifically, in this embodiment, the direction of the coordinate axis is the positive direction of the Z axis.
[0094] S222, receiving the intersection surface of the three-dimensional model and the copy surface.
[0095] Specifically, for the intersection of the 3D model and the copy surface, an intersection image can be read first and converted into a grayscale image. Then the Canny algorithm is used for edge detection to obtain a binary image of the line. Finally, the edge line of the intersection is displayed.
[0096] S223, extracting edge lines of the intersection surfaces based on an edge extraction algorithm.
[0097] S224, returning to the step of selecting one copy surface in the cutting library in turn according to the direction of any coordinate axis, until all copy surfaces are selected.
[0098] S225, calling the gradient algorithm and the orientation of the coordinate axis to determine the spatial line drawing of the basic three-dimensional model of the acquisition object.
[0099] Specifically, Figure 3 As shown, the edge line of each intersection surface is regarded as a (x, y) closed curve.
[0100] Simply, take a closed curve f(x, y) as an example.
[0101]
[0102] That is It is called the (two-dimensional) vector differential operator.
[0103] f x (x, y) is the partial differential of zi at x. y (x, y) is the partial derivative of zi in y.
[0104] The coordinates of the closed curve zi are assigned to x and y of the closed curve zi=f(x, y).
[0105] The two-dimensional matrix Q can be formed:
[0106]
[0107] in, is the kth one corresponding to zi [zi] is a two-dimensional matrix of kxi formed by the z-axis coordinate values of each replica surface, where each column of [zi] contains k z1…k zi.
[0108] Since each (two-dimensional) vector differential operator has an adjacent (two-dimensional) vector differential operator in the plane parallel to the plane zpx, and the vertical distance between these two adjacent (two-dimensional) vector differential operators in the z-axis direction is 0.1, we can calculate The corresponding (one-dimensional) vector differential operator in the z-axis direction.
[0109] In one embodiment of the present application, S200 and subsequent steps include:
[0110] S231, receiving a standard specification.
[0111] Specifically, the standard data format is: single (three-dimensional) vector differential operator ax axis vector differential operator basis + by axis vector differential operator basis + cz axis vector differential operator basis. For example, the x axis vector differential operator basis is
[0112] S232, selecting a space line drawing of a basic three-dimensional model of an acquisition object that meets standard specifications to obtain a target space line drawing. The space line drawing has a three-dimensional size of a three-dimensional space.
[0113] Specifically, by using the vector differential operator of the standard specifications and the spatial line diagram of the basic three-dimensional model of the acquisition object, the spatial line diagram of the basic three-dimensional model of the acquisition object that meets the standard specifications can be calculated relatively quickly.
[0114] S233, select a visual center coordinate in three-dimensional space.
[0115] S234, using the visual center coordinates and the target space line diagram, a two-dimensional plane visual diagram is generated.
[0116] S235a, calling the working radius of the fruit and vegetable picking and screening system based on visual images.
[0117] S235b, based on the working radius and the three-dimensional center coordinates of the target space line graph, generate a selection radius of the visual center coordinates.
[0118] S235c, receiving the minimum distance length of adjacent visual center coordinates.
[0119] S235d, generating M visual center coordinates to be selected within the three-dimensional space range of the selection radius of the visual center coordinate, where M is a positive integer.
[0120] S235, returning to the step of selecting a visual center coordinate in the three-dimensional space until all visual center coordinates are selected.
[0121] Specifically, using the working radius of the robotic claw, the vector origin of the vector differential operator of the spatial line drawing of the basic three-dimensional model of the acquisition object that meets standard specifications and the working radius of the robotic claw can be used to generate a range sphere of the selection radius with the visual center coordinates.
[0122] By cutting the space within the range sphere with any adjacent minimum distance length, M visual center coordinates to be selected can be generated at the intersection of the cuts.
[0123] In one embodiment of the present application, S600 includes:
[0124] S610, calling a two-dimensional plane visual image.
[0125] S620, using a discrete differential operator to determine the characteristics of the two-dimensional plane visual image.
[0126] Specifically, the two-dimensional plane visual image is essentially: the projection of the object on the field of view of the camera's visual center.
[0127] Therefore, the two-dimensional plane visual image can be matched and compared with the line images of all the images in the image sample library.
[0128] The technical means for matching comparison are: S630 to S690.
[0129] S630, selecting a feature of the two-dimensional plane visual image.
[0130] S640, select a line feature.
[0131] S650, determining whether the features of the two-dimensional plane visual image are consistent with the line features.
[0132] Specifically, Figure 4 As shown, a line feature of the line feature data set is selected. A differential operator of the line feature is obtained by using a discrete differential operator. A line feature of the selected line feature data set is returned until all line features are selected. The characteristic equations of all differential operators are calculated. A mapping relationship is formed between the solution of each characteristic equation and the two-dimensional plane visual image.
[0133] S660, if the features of the two-dimensional plane visual image match the line features, a mapping relationship is formed between the features of the two-dimensional plane visual image and the line features, and the process of selecting a line feature is returned until all line features are selected.
[0134] S670, if the features of the two-dimensional plane visual image do not match the line features, return to the step of selecting a line feature until all line features are selected.
[0135] S680, returning to the step of selecting a feature of the two-dimensional plane visual image until all features are selected.
[0136] S690, including the line features with a mapping relationship into a line feature data set of the spatial position to be calculated.
[0137] Specifically, the line features come from: the line-shaped image can be line-shaped based on the Canny algorithm, and the deep learning algorithm is executed on the line-shaped image to determine each vertex of the line-shaped two-dimensional plane visual image. These vertices can form a two-dimensional coordinate point matrix G.
[0138] In one embodiment of the present application, S700 includes:
[0139] S710, select a line drawing.
[0140] S720, based on the discrete differential operator, obtain a differential operator for the line-based image.
[0141] S730, incorporating the differential operator into the characteristic equation to obtain a numerical solution to the characteristic equation.
[0142] S740, select a solution to the characteristic equation.
[0143] S750, determining whether the value solution of the characteristic equation is consistent with the solution of the characteristic equation.
[0144] S760, if the value solution of the characteristic equation does not match the solution of the characteristic equation, then return to the step of selecting a solution of the characteristic equation until all solutions of the characteristic equation are selected.
[0145] S770, if the value solution of the characteristic equation matches the solution of the characteristic equation, a one-to-one mapping relationship is formed between the line image and the two-dimensional plane visual image, and the selection of a solution to the characteristic equation is returned until all solutions to the characteristic equation are selected.
[0146] The present application provides a fruit and vegetable picking and screening system based on visual images.
[0147] like Figure 5 As shown, in one embodiment of the present application, a fruit and vegetable picking and screening system based on visual images includes:
[0148] The processor 100 is used to execute the fruit and vegetable picking and screening method based on visual images.
[0149] The camera 200 is connected to the processor 100 for communication.
[0150] The mechanical claw 300 is communicatively connected with the processor 100 .
[0151] This embodiment relates to a fruit and vegetable picking and screening system based on visual images. The processor 100 can obtain the three-dimensional contour lines of the basic three-dimensional model of the collection object by receiving the basic three-dimensional model of the collection object, such as the three-dimensional model of the tender branches to be collected, the three-dimensional model of the tender leaves to be collected, and the three-dimensional model of the mature fruits to be collected, so as to obtain the curve shape and curve size data of the basic three-dimensional model of the collection object at different viewing distances and different angles. The image sample library is called, and based on the edge detection algorithm, the linearized images of all images in the image sample library are obtained, and the line features of the linearized images are extracted using the deep learning algorithm. The curve shape is compared with the line features of the linearized images to determine the viewing distance and angle of the basic three-dimensional model of the collection object corresponding to the linearized images. When the camera 200 is working, it takes pictures of fruits and vegetables. The processor 100 determines the viewing distance and angle of the fruits and vegetables to be picked based on the fruit and vegetable picking and screening method of the visual image, and the processor 100 drives the mechanical claw 300 to perform fruit and vegetable grabbing.
[0152] The technical features of the above-described embodiments may be arbitrarily combined, and the execution order of the method steps is not limited. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0153] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A fruit and vegetable picking and screening method based on visual images, characterized in that: include: Receive a basic three-dimensional model of the acquisition object; Simplify the basic three-dimensional model of the acquisition object to determine a spatial line drawing of the basic three-dimensional model of the acquisition object; Call the image sample library; Based on the edge detection algorithm, obtain the line images of all the images in the image sample library; Use deep learning algorithms to extract line features of linear images; Based on the spatial line drawing of the basic three-dimensional model of the acquisition object, determining the line features that match the basic three-dimensional model of the acquisition object; Based on the three-dimensional spatial position algorithm of the spatial line graph and model, the spatial position of each line feature is calculated.
2. The method for picking and screening fruits and vegetables based on visual images according to claim 1, characterized in that: The receiving and collecting the basic three-dimensional model of the object includes: The basic three-dimensional model of the acquisition object includes one or more of a three-dimensional model of a branch, a three-dimensional model of a leaf, and a three-dimensional data model of a fruit.
3. The method for picking and screening fruits and vegetables based on visual images according to claim 2, characterized in that: The step of simplifying the basic three-dimensional model of the acquisition object to determine a spatial line drawing of the basic three-dimensional model of the acquisition object includes: Create a two-dimensional plane in three-dimensional space; Determine the moving step size of the two-dimensional plane; Generate N replicas using the moving step and the two-dimensional plane; N is a positive integer; Add all copied faces to the cutting library.
4. The method for picking and screening fruits and vegetables based on visual images according to claim 3, characterized in that: The step of simplifying the basic three-dimensional model of the acquisition object to determine the spatial line drawing of the basic three-dimensional model of the acquisition object further includes: According to the direction of any coordinate axis, select a copy surface in the cutting library in turn; receiving the intersection of the three-dimensional model and the copied surface; Based on the edge extraction algorithm, the edge line of the intersection surface is extracted; Return to the above step of selecting a copy surface in the cutting library in turn according to the direction of any coordinate axis until all the copy surfaces are selected; The gradient algorithm and the direction of the coordinate axis are called to determine the spatial line drawing of the basic three-dimensional model of the acquisition object.
5. The method for picking and screening fruits and vegetables based on visual images according to claim 4, characterized in that: After simplifying the basic three-dimensional model of the acquisition object and determining the spatial line graph of the basic three-dimensional model of the acquisition object, the method includes: Receive a standard specification; Selecting a spatial line drawing of a basic three-dimensional model of an acquisition object that meets standard specifications to obtain a target spatial line drawing; the spatial line drawing has a three-dimensional dimension of a three-dimensional space; In three-dimensional space, select a visual center coordinate; Generate a two-dimensional visual map using the visual center coordinates and the target space line map; Return to the above and select a visual center coordinate in the three-dimensional space until all visual center coordinates are selected.
6. The method for picking and screening fruits and vegetables based on visual images according to claim 5, characterized in that: The returning, selecting a visual center coordinate in the three-dimensional space until all visual center coordinates are selected, includes: The working radius of the fruit and vegetable picking and screening system that uses visual images; Based on the working radius and the three-dimensional center coordinates of the target space line graph, a selection radius of the visual center coordinates is generated; Receive the minimum distance length of adjacent visual center coordinates; Within the three-dimensional space range of the selection radius of the visual center coordinate, M visual center coordinates to be selected are generated; M is a positive integer.
7. The method for picking and screening fruits and vegetables based on visual images according to claim 6, characterized in that: The determining of line features that match the basic three-dimensional model of the acquisition object based on the spatial line graph of the basic three-dimensional model of the acquisition object includes: Call the 2D plane visual map; Using discrete differential operators, the characteristics of the two-dimensional visual image are determined; Select a feature of the 2D visual map; Select a line feature; Determine whether the features of the two-dimensional visual image match the line features; If the features of the two-dimensional plane visual image match the line features, a mapping relationship is formed between the features of the two-dimensional plane visual image and the line features, and the method of selecting a line feature is returned until all line features are selected; If the features of the two-dimensional plane visual image do not match the line features, return to the step of selecting a line feature until all line features are selected; Returning to selecting a feature of the two-dimensional plane visual image until all features are selected; The line features with mapping relationships are included in the line feature data set of the spatial position to be calculated.
8. The method for picking and screening fruits and vegetables based on visual images according to claim 7, characterized in that: The three-dimensional spatial position algorithm based on the spatial line graph and the model calculates the spatial position of each line feature. The three-dimensional spatial position algorithm of the model includes: Select a line feature from the line feature dataset; Using discrete differential operators, the differential operators of line features are obtained; Return a line feature of the selected line feature data set until all line features are selected; Compute the characteristic equations of all differential operators; Map the solution of each characteristic equation to a two-dimensional visual image.
9. The method for picking and screening fruits and vegetables based on visual images according to claim 8, characterized in that: The three-dimensional spatial position algorithm based on the spatial line graph and model calculates the spatial position of each line feature, including: Select a line drawing; Based on the discrete differential operator, the differential operator of the line image is obtained; Incorporate the differential operator into the characteristic equation and obtain the value solution of the characteristic equation; Select a solution to the characteristic equation; Determine whether the value solution of the characteristic equation is consistent with the solution of the characteristic equation; If the value solution of the characteristic equation does not match the solution of the characteristic equation, then returning to the step of selecting a solution of the characteristic equation until all solutions of the characteristic equation are selected; If the value solution of the characteristic equation matches the solution of the characteristic equation, a one-to-one mapping relationship is formed between the line image and the two-dimensional plane visual image, and the selection of a solution to the characteristic equation is returned until all solutions to the characteristic equation are selected.
10. A fruit and vegetable picking and screening system based on visual images, characterized in that: include: A processor, configured to execute the fruit and vegetable picking and screening method based on visual images as described in any one of claims 1 to 9; A camera, communicatively connected to the processor; A mechanical claw is communicatively connected to the processor.