Picking manipulator based on artificial intelligence visual identification

Abstract

The invention relates to the technical field of picking machinery, in particular to a picking manipulator based on artificial intelligence visual identification. According to the method, the picking state evaluation index of each picked fruit corresponding to the picking region is obtained through comprehensive analysis and calculation of the overlapping area of the picking region corresponding to the appearance contour of each picked fruit, the overlapping number of feature points in the skin texture of each picked fruit and the illumination intensity of each picking region corresponding to each picked fruit; then, the picking state evaluation index of each picked fruit corresponding to the picking area is compared with a set picking state evaluation index threshold value to obtain a corresponding picking signal, and a picking manipulator adjusts control parameters according to the corresponding picking signal to determine whether picking operation needs to be carried out on the fruits or not; according to the picking manipulator, the picking position of the picked fruits and the recovery position of the picked fruits, the picking efficiency of the picking manipulator on the fruits is remarkably improved, the picked fruits can be automatically classified, and the picking effect on the fruits is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of picking machinery, and in particular to a picking manipulator based on artificial intelligence visual recognition. Background Art

[0002] Automated production and processing is an inevitable trend, and agricultural production is no exception. The Chinese government has long determined the strategy of promoting agriculture through science and education. Popularizing automated equipment in agricultural production, improving production efficiency and reducing labor costs are the future development direction of agricultural production. Fruit planting occupies a pivotal position in agricultural production, and fruit picking in my country is almost entirely achieved by human labor. With the development of social economy and the acceleration of population aging, there will be a serious shortage of agricultural labor, which will lead to an increase in fruit production costs. Therefore, if the agricultural fruit industry wants to develop rapidly, it is urgent to improve labor efficiency and realize the automation of fruit picking.

[0003] At present, agricultural harvesting robots are still in the research stage and there is still a long way to go before commercialization and practical application. For these reasons, it is of long-term significance to develop a high-efficiency, low-cost agricultural harvesting robot based on machine vision. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a picking robot based on artificial intelligence visual recognition to solve the above-mentioned technical defects.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: A picking manipulator based on artificial intelligence visual recognition, comprising a mounting base, characterized in that: a rotating chassis is rotatably provided on the top of the mounting base, a fixed frame is fixedly provided on the top of the rotating chassis, and a first movable frame is rotatably provided on the top of the fixed frame, an adjusting cylinder is rotatably provided on one side of the fixed frame, and a driving end of the adjusting cylinder is movably connected to the bottom of the first movable frame;

[0006] A first mounting frame is rotatably provided on one side of the top of the first movable frame, a servo linear slide is fixedly provided on one side of the first mounting frame, and a second movable frame is slidably provided on the surface of the servo linear slide, a second motor is fixedly provided on one side of the top of the second movable frame, and a second mounting frame is rotatably provided on the other side of the top of the second movable frame, one end of the output shaft of the second motor is fixedly connected to one side of the second mounting frame, a third motor is fixedly provided on one side of the second mounting frame, and a driving screw is fixedly provided on one end of the output shaft of the third motor, a picking frame is fixedly provided on one side of the second mounting frame, and three picking arms are movably provided on one side of the picking frame;

[0007] The picking arm is internally provided with a visual recognition module, a fruit state analysis module, a fruit picking evaluation module, an execution terminal and a database; the visual recognition module is respectively connected to the fruit state analysis module and the database; the fruit state analysis module is respectively connected to the fruit picking evaluation module and the database; and the fruit picking evaluation module is respectively connected to the execution terminal and the database.

[0008] Preferably, a driving motor is fixedly provided on the top of the inner wall of the mounting base, and a driving gear is fixedly provided on the output shaft of the driving motor, an inner gear ring is fixedly provided on the inner wall of the rotating chassis, and the surface of the driving gear meshes with the inner wall of the inner gear ring for transmission.

[0009] Preferably, a first motor is fixedly disposed on the other side of the top of the first movable frame, and an output shaft of the first motor is fixedly connected to one side of the first mounting frame.

[0010] Preferably, an adjustment frame is threadedly provided on the surface of the driving screw, and three adjustment rods are movably provided on the outer circumferential surface of the adjustment frame, and one end of the three adjustment rods is movably connected to one end of three picking arms respectively.

[0011] Preferably, three limiting rods are fixedly provided on one side of the second mounting frame, and one end of the three limiting rods are slidably connected to the inside of the adjustment frame.

[0012] Preferably, the visual recognition module is used to collect data on the picking position, shape image set, and light intensity of each picked fruit in each picking area corresponding to the picking robot during picking.

[0013] Preferably, the fruit state analysis module is used to analyze the picking state of each picked fruit corresponding to each picking area of ​​the picking manipulator during picking, and the specific analysis process is as follows:

[0014] Extracting the surface image of each picked fruit in the picking area from the shape image set of each picked fruit in each picking area when the picking robot is picking, obtaining the surface image of each picked fruit in the picking area, and obtaining the shape contour and skin texture of each picked fruit in the picking area based on the surface image of each picked fruit in the picking area;

[0015] Compare the outlines of each picked fruit corresponding to the picking area with the outlines of reference picked fruits stored in the database to obtain the overlapping area of ​​the outlines of each picked fruit corresponding to the picking area and the reference picked fruit, which is recorded as the overlapping area of ​​the outlines of each picked fruit corresponding to the picking area;

[0016] The reference skin texture of each picked fruit is obtained from the database, and a number of feature points are extracted on the reference skin texture. The skin texture of each picked fruit in the picking area corresponding to the fruit is compared with the skin texture of the reference picked fruit stored in the database for overlap of feature points, and the number of overlaps of feature points in the skin texture of each picked fruit in the picking area corresponding to the fruit and the skin texture of the reference picked fruit is obtained, which is recorded as the number of overlaps of feature points in the skin texture of each picked fruit in the picking area corresponding to the fruit.

[0017] Preferably, the fruit picking evaluation module is used to evaluate the picking status of each picking area corresponding to each picked fruit when the picking manipulator is picking, and the specific evaluation method is as follows:

[0018] The fruit picking evaluation parameters of each picking area corresponding to each picked fruit are formed by the overlapping area of ​​the picking area corresponding to the outer contour of each picked fruit, the overlapping number of feature points in the skin texture of each picked fruit, and the light intensity of each picking area corresponding to each picked fruit;

[0019] From the fruit picking evaluation parameters of each picking area corresponding to each picked fruit, the values ​​of the overlap area, overlap number and light intensity of each picking area corresponding to each picked fruit are extracted and recorded as S j i , Q j i and L j i , i represents the number of each picked fruit, i=1,2,...,n, n represents the sum of the numbers of each picked fruit, j represents the number of each picking area, j=1,2,...,m, m represents the sum of the numbers of each picking area;

[0020] According to the formula The picking state evaluation index of each picked fruit in the picking area is calculated. ΔS is the difference in the allowable overlap area of ​​each picked fruit in the picking area. ΔQ is the difference in the allowable overlap quantity of each picked fruit in the picking area. ΔL is the difference in the allowable light intensity of each picked fruit in the picking area. S j i-1 It is expressed as the overlapping area of ​​the outline of the i-1th picked fruit in the picking area, Q j i-1 It is expressed as the number of overlapping feature points in the skin texture of the i-1th picked fruit in the picking area, L j i-1 It is represented as the light intensity of the picking area corresponding to the i-1th picked fruit, and a1, a2, and a3 represent the weight factors corresponding to the set overlapping area, overlapping number, and light intensity, respectively.

[0021] Preferably, the picking state evaluation index of each picked fruit in the picking area is compared with a set picking state evaluation index threshold, and the specific comparison results are as follows:

[0022] If γ i ≥q1, it means that the picking status of each fruit in the picking area is normal, and a normal signal is generated and sent to the execution terminal;

[0023] If q1>γ i >q2, it means that the picking status of each fruit in the picking area is poor quality, and a poor quality signal is generated and sent to the execution terminal;

[0024] If γ i ≤q2, it means that the picking status of each fruit in the picking area is abnormal, and an abnormal signal is generated and sent to the execution terminal.

[0025] Among them, q1 and q2 both represent the picking state evaluation index thresholds of each picked fruit corresponding to the set picking area, and q1>q2>0.

[0026] Preferably, the execution terminal performs corresponding operations based on the signal generated by the fruit picking evaluation module, and the specific operation method is as follows:

[0027] After receiving a normal signal, the execution terminal controls the picking robot to pick the fruit and put it into a normal collection frame. After receiving a poor quality signal, the execution terminal controls the picking robot to pick the fruit and put it into a diseased fruit collection frame. After receiving an abnormal signal, the execution terminal controls the picking robot not to pick the fruit.

[0028] Compared with the prior art, it has the following beneficial effects:

[0029] 1. In the present invention, the position of a picking arm of a picking robot is flexibly controlled through the first movable frame, the second movable frame and the second mounting frame, so that the three picking arms on one side of the second mounting frame are moved to the side of the fruit to be picked, and finally the output shaft of the third motor is used to control the driving screw to rotate clockwise, and the driving screw is used to drive the adjusting frame to slide to one side along the limit rod, so that the three picking arms are spread outward, and then the second movable frame is driven by the servo linear slide to move to one side of the fruit to be picked, so that the picked fruit is between the three picking arms, and the output shaft of the third motor is controlled to rotate counterclockwise to allow the three picking arms to cover the picked fruit, and finally the second movable frame is controlled to reset along the first mounting frame to complete the fruit picking operation, thereby significantly improving the fruit picking effect of the picking robot.

[0030] 2. In the present invention, a visual recognition module is set on the picking robot to collect data on the picking position, fruit image and light intensity of the picked fruits in each picking area during the fruit picking process. The picking robot is quickly positioned based on the picking position of the picked fruits in each picking area, thereby improving the picking efficiency of the picking robot for the fruits. Data is collected through the fruit image and light intensity of the picked fruits in each picking area, which provides strong data support for the subsequent control of the picking parameters of the picked fruits by the picking robot.

[0031] 3. In the present invention, a comprehensive analysis is performed on the overlapping areas of the picking areas corresponding to the outer contours of each picked fruit, the overlapping number of feature points in the skin texture of each picked fruit, and the light intensity of each picking area corresponding to each picked fruit to calculate the picking state evaluation index of the picking area corresponding to each picked fruit, and then the picking state evaluation index of the picking area corresponding to each picked fruit is compared with the set picking state evaluation index threshold to obtain a corresponding picking signal, so that the picking robot adjusts the control parameters according to the corresponding picking signal to decide whether picking operation is needed for the fruit, as well as the recovery position of the fruit after picking, thereby significantly improving the picking efficiency of the picking robot for the fruit, and being able to realize automatic classification processing of the picked fruits, thereby improving the picking effect of the fruit. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a schematic diagram of a picking robot structure based on artificial intelligence visual recognition according to an embodiment of the present invention;

[0033] Figure 2 A schematic diagram of a driving motor and a driving gear structure according to an embodiment of the present invention;

[0034] Figure 3 It is a schematic diagram of the first motor and the second movable frame structure according to an embodiment of the present invention;

[0035] Figure 4 It is a schematic diagram of the servo linear slide and the first mounting frame structure according to an embodiment of the present invention;

[0036] Figure 5 It is a schematic diagram of the picking frame and picking arm structure of an embodiment of the present invention;

[0037] Figure 6 This is a system principle block diagram of a picking robot based on artificial intelligence visual recognition according to an embodiment of the present invention.

[0038] In the figure, 1. mounting base; 2. rotating chassis; 3. fixed frame; 4. first movable frame; 5. adjusting cylinder; 6. driving motor; 7. driving gear; 8. inner ring gear; 9. first mounting frame; 10. servo linear slide; 11. second movable frame; 12. first motor; 13. second motor; 14. second mounting frame; 15. third motor; 16. driving screw; 17. picking frame; 18. picking arm; 19. adjusting frame; 20. adjusting rod; 21. limiting rod. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0040] Embodiment 1:

[0041] See also Figures 1 to 6 As shown, a picking manipulator based on artificial intelligence visual recognition comprises a mounting base 1, a rotating chassis 2 is rotatably provided on the top of the mounting base 1, a fixed frame 3 is fixedly provided on the top of the rotating chassis 2, and a first movable frame 4 is rotatably provided on the top of the fixed frame 3, an adjusting cylinder 5 is rotatably provided on one side of the fixed frame 3, and the driving end of the adjusting cylinder 5 is movably connected to the bottom of the first movable frame 4, wherein a driving motor 6 is fixedly provided on the top of the inner wall of the mounting base 1, and a driving gear 7 is fixedly provided on the output shaft of the driving motor 6, an inner gear ring 8 is fixedly provided on the inner wall of the rotating chassis 2, and the surface of the driving gear 7 is meshed with the inner wall of the inner gear ring 8 for transmission;

[0042] Further, a first mounting frame 9 is rotatably provided on one side of the top of the first movable frame 4, a servo linear slide 10 is fixedly provided on one side of the first mounting frame 9, and a second movable frame 11 is slidably provided on the surface of the servo linear slide 10, a first motor 12 is also fixedly provided on the other side of the top of the first movable frame 4, and the output shaft of the first motor 12 is fixedly connected to one side of the first mounting frame 9, a second motor 13 is fixedly provided on one side of the top of the second movable frame 11, and a second mounting frame 14 is rotatably provided on the other side of the top of the second movable frame 11, and one end of the output shaft of the second motor 13 is fixedly connected to one side of the second mounting frame 14;

[0043] Furthermore, a third motor 15 is fixedly provided on one side of the second mounting frame 14, and a driving screw 16 is fixedly provided on one end of the output shaft of the third motor 15; a picking frame 17 is fixedly provided on one side of the second mounting frame 14, and three picking arms 18 are movably provided on one side of the picking frame 17, wherein the three picking arms 18 are distributed at equal angles about the central axis of the picking frame 17, and the picking arms 18 are composed of a number of movable connecting rods and pins; an adjusting frame 19 is threadedly provided on the surface of the driving screw 16, and three adjusting rods 20 are movably provided on the outer peripheral surface of the adjusting frame 19, and one end of the three adjusting rods 20 is movably connected to one end of the three picking arms 18, respectively.

[0044] Furthermore, three limiting rods 21 are fixedly provided on one side of the second mounting frame 14 , and one end of the three limiting rods 21 are slidably connected to the inside of the adjustment frame 19 .

[0045] It should be noted that when picking fruits, the output shaft of the driving motor 6 is used to drive the meshing transmission between the driving gear 7 and the inner gear ring 8, so that the rotating chassis 2 drives the fixed frame 3 to rotate, and the picking direction of the picking robot is regulated. Then, the driving end of the regulating cylinder 5 is extended to control the top of the first movable frame 4 to lift upward, and the picking height of the picking robot is regulated. Then, the second movable frame 11 is controlled by the servo linear slide 10 to slide to one side along the first mounting frame 9, so that the picking arm 18 on one side of the second movable frame 11 approaches the picked fruit. At the same time, the output shaft of the first motor 12 is used to control the angle of the first mounting frame 9, and the output shaft of the second motor 13 is used to control the angle of the second mounting frame 9. The angle of the mounting frame 14 allows the three picking arms 18 on one side of the second mounting frame 14 to move to the side where the fruit is to be picked. Finally, the output shaft of the third motor 15 is used to control the driving screw 16 to rotate clockwise, and the driving screw 16 is used to drive the adjusting frame 19 to slide to one side along the limit rod 21, so that the three picking arms 18 are spread outward, and then the second movable frame 11 is driven by the servo linear slide 10 to move to the side where the fruit is to be picked, so that the picked fruit is between the three picking arms 18, and the output shaft of the third motor 15 is controlled to rotate counterclockwise to allow the three picking arms 18 to cover the picked fruit. Finally, the second movable frame 11 is controlled to reset along the first mounting frame 9 to complete the fruit picking operation.

[0046] Embodiment 2:

[0047] See also Figure 6 As shown, as a further specific explanation of the scheme in the above embodiment, the interior of the picking arm 18 is provided with a visual recognition module, a fruit state analysis module, a fruit picking evaluation module, and a database, the visual recognition module is respectively connected to the fruit state analysis module and the database, the fruit state analysis module is respectively connected to the fruit picking evaluation module and the database, and the fruit picking evaluation module is respectively connected to the execution terminal and the database.

[0048] The visual recognition module is used to collect data on the picking position, fruit image, and light intensity of each fruit in each picking area of ​​the picking robot. The specific collection process is as follows:

[0049] The camera arranged on the picking arm 18 is used to collect image data corresponding to each picking area when the picking robot is picking, and the position of the picked fruit is marked in the collected image data, and the coordinates and position data of the picking robot are recorded at the same time, and the distance data between the picking robot and the picked fruit are obtained in conjunction with the ranging sensor, and the picking position of each picking area corresponding to each picked fruit when the picking robot is picking is obtained by calculating the distance data between the picking robot and the picked fruit and the relative position coordinates between the picking robot and the picked fruit.

[0050] Among them, after obtaining the distance data between the picking robot and the picked fruit and the relative position coordinates between the picking robot and the picked fruit, the distance data is substituted into the relative position coordinates between the picking robot and the picked fruit, thereby obtaining the picking position of each picked fruit corresponding to each picking area of ​​the picking robot during picking.

[0051] By controlling the picking arm 18 to move to one side of the picked fruit, and then using the camera set on the picking arm 18 to obtain the image of the picked fruit, and by adjusting the position of the picking arm 18 relative to the picked fruit, a set of appearance images of each picked fruit corresponding to each picking area of ​​the picking robot during picking is obtained.

[0052] The light sensor arranged on the picking arm 18 acquires the light intensity data of each picking area corresponding to each picked fruit when the picking robot is picking, so as to obtain the light intensity of the picked fruit when picking.

[0053] In a specific embodiment, the present invention sets a visual recognition module on the picking robot to collect data on the picking position, fruit image and light intensity of the picked fruits in each picking area during the picking process. The picking robot is quickly positioned based on the picking position of the picked fruits in each picking area, thereby improving the picking efficiency of the picking robot for the fruits. Data is collected through the fruit image and light intensity of the picked fruits in each picking area, thereby providing strong data support for the subsequent control of the picking parameters of the picked fruits by the picking robot.

[0054] The fruit status analysis module is used to analyze the picking status of each fruit in each picking area of ​​the picking robot during picking. The specific analysis process is as follows:

[0055] Extracting the surface image of each picked fruit in the picking area from the shape image set of each picked fruit in each picking area when the picking robot is picking, obtaining the surface image of each picked fruit in the picking area, and obtaining the shape contour and skin texture of each picked fruit in the picking area based on the surface image of each picked fruit in the picking area;

[0056] Compare the outlines of each picked fruit corresponding to the picking area with the outlines of reference picked fruits stored in the database to obtain the overlapping area of ​​the outlines of each picked fruit corresponding to the picking area and the reference picked fruit, which is recorded as the overlapping area of ​​the outlines of each picked fruit corresponding to the picking area;

[0057] The reference skin texture of each picked fruit is obtained from the database, and a number of feature points are extracted on the reference skin texture. The skin texture of each picked fruit in the picking area corresponding to the fruit is compared with the skin texture of the reference picked fruit stored in the database for overlap of feature points, and the number of overlaps of feature points in the skin texture of each picked fruit in the picking area corresponding to the fruit and the skin texture of the reference picked fruit is obtained, which is recorded as the number of overlaps of feature points in the skin texture of each picked fruit in the picking area corresponding to the fruit.

[0058] It should be noted that the reference picked fruit's outline includes the size and shape of the ripened fruit, and the reference picked fruit's skin texture's feature points include color, bumps, lines, and other aspects.

[0059] The fruit picking evaluation module is used to evaluate the picking status of each picking area corresponding to each picked fruit during picking by the picking manipulator. The specific evaluation method is as follows:

[0060] The fruit picking evaluation parameters of each picking area corresponding to each picked fruit are composed of the overlapping area of ​​the picking area corresponding to the outer contour of each picked fruit, the overlapping number of feature points in the skin texture of each picked fruit and the light intensity of each picking area corresponding to each picked fruit.

[0061] From the fruit picking evaluation parameters of each picking area corresponding to each picked fruit, the values ​​of the overlap area, overlap number and light intensity of each picking area corresponding to each picked fruit are extracted and recorded as S j i , Q j i and L j i , i represents the number of each picked fruit, i=1,2,...,n, n represents the sum of the numbers of each picked fruit, j represents the number of each picking area, j=1,2,...,m, m represents the sum of the numbers of each picking area.

[0062] According to the formula The picking state evaluation index of each picked fruit in the picking area is calculated. ΔS is the difference in the allowable overlap area of ​​each picked fruit in the picking area. ΔQ is the difference in the allowable overlap quantity of each picked fruit in the picking area. ΔL is the difference in the allowable light intensity of each picked fruit in the picking area. S j i-1 It is expressed as the overlapping area of ​​the outline of the i-1th picked fruit in the picking area, Q j i-1 It is expressed as the number of overlapping feature points in the skin texture of the i-1th picked fruit in the picking area, L j i-1 It is represented as the light intensity of the picking area corresponding to the i-1th picked fruit, and a1, a2, and a3 represent the weight factors corresponding to the set overlapping area, overlapping number, and light intensity, respectively.

[0063] The picking state evaluation index thresholds of the set picking areas corresponding to each picked fruit are obtained through the database, and the picking state evaluation indexes of the picking areas corresponding to each picked fruit are compared with the set picking state evaluation index thresholds. The specific comparison results are as follows:

[0064] If γ i ≥q1, it means that the picking status of each fruit in the picking area is normal, and a normal signal is generated and sent to the execution terminal;

[0065] If q1>γ i >q2, it means that the picking status of each fruit in the picking area is poor quality, and a poor quality signal is generated and sent to the execution terminal;

[0066] If γ i ≤q2, it means that the picking status of each fruit in the picking area is abnormal, and an abnormal signal is generated and sent to the execution terminal.

[0067] Among them, q1 and q2 both represent the picking state evaluation index thresholds of each picked fruit corresponding to the set picking area, and q1>q2>0.

[0068] In a specific embodiment, the present invention calculates the picking state evaluation index of each picked fruit in the picking area by comprehensively analyzing the overlapping area of ​​the outer contours of each picked fruit, the overlapping number of feature points in the skin texture of each picked fruit, and the light intensity of each picked fruit in the picking area. The picking state evaluation index of each picked fruit in the picking area is then compared with the set picking state evaluation index threshold to obtain a corresponding picking signal, so that the picking robot adjusts the control parameters according to the corresponding picking signal to decide whether picking operation is needed for the fruit, as well as the recovery position of the fruit after picking, thereby significantly improving the picking efficiency of the fruit by the picking robot, and being able to realize automatic classification processing of the picked fruits, thereby improving the picking effect of the fruit.

[0069] The execution terminal performs corresponding operations based on the signal generated by the fruit picking evaluation module. The specific operation methods are as follows:

[0070] After receiving a normal signal, the execution terminal controls the picking robot to pick the fruit and put it into a normal collection frame. After receiving a poor quality signal, the execution terminal controls the picking robot to pick the fruit and put it into a diseased fruit collection frame. After receiving an abnormal signal, the execution terminal controls the picking robot not to pick the fruit.

[0071] The database is used to store the outline of reference picked fruits, skin texture and picking state evaluation index thresholds of each picked fruit corresponding to the picking area.

[0072] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0073] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0074] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A harvesting manipulator based on artificial intelligence visual recognition, comprising a mounting base (1), characterized in that: A rotating chassis (2) is rotatably provided on the top of the mounting base (1), a fixed frame (3) is fixedly provided on the top of the rotating chassis (2), and a first movable frame (4) is rotatably provided on the top of the fixed frame (3), an adjusting cylinder (5) is rotatably provided on one side of the fixed frame (3), and a driving end of the adjusting cylinder (5) is movably connected to the bottom of the first movable frame (4); A first mounting frame (9) is rotatably provided on one side of the top of the first movable frame (4), a servo linear slide (10) is fixedly provided on one side of the first mounting frame (9), and a second movable frame (11) is slidably provided on the surface of the servo linear slide (10), a second motor (13) is fixedly provided on one side of the top of the second movable frame (11), and a second mounting frame (14) is rotatably provided on the other side of the top of the second movable frame (11), one end of the output shaft of the second motor (13) is fixedly connected to one side of the second mounting frame (14), a third motor (15) is fixedly provided on one side of the second mounting frame (14), and a driving screw rod (16) is fixedly provided on one end of the output shaft of the third motor (15), a picking frame (17) is fixedly provided on one side of the second mounting frame (14), and three picking arms (18) are movably provided on one side of the picking frame (17); The picking arm (18) is internally provided with a visual recognition module, a fruit state analysis module, a fruit picking evaluation module, an execution terminal and a database; the visual recognition module is respectively connected to the fruit state analysis module and the database; the fruit state analysis module is respectively connected to the fruit picking evaluation module and the database; and the fruit picking evaluation module is respectively connected to the execution terminal and the database.

2. The picking manipulator based on artificial intelligence visual recognition according to claim 1 is characterized in that: A driving motor (6) is fixedly arranged on the top of the inner wall of the mounting base (1), and a driving gear (7) is fixedly arranged on the output shaft of the driving motor (6); an inner gear ring (8) is fixedly arranged on the inner wall of the rotating chassis (2), and the surface of the driving gear (7) meshes with the inner wall of the inner gear ring (8) for transmission.

3. The picking manipulator based on artificial intelligence visual recognition according to claim 1 is characterized in that: A first motor (12) is also fixedly arranged on the other side of the top of the first movable frame (4), and the output shaft of the first motor (12) is fixedly connected to one side of the first mounting frame (9).

4. The picking manipulator based on artificial intelligence visual recognition according to claim 1 is characterized in that: The surface of the driving screw rod (16) is threadedly provided with an adjustment frame (19), and the outer peripheral surface of the adjustment frame (19) is movably provided with three adjustment rods (20), one end of the three adjustment rods (20) is movably connected to one end of three picking arms (18).

5. The picking manipulator based on artificial intelligence visual recognition according to claim 1 is characterized in that: Three limiting rods (21) are also fixedly arranged on one side of the second mounting frame (14), and one end of the three limiting rods (21) is slidably connected to the inside of the adjustment frame (19).

6. The picking manipulator based on artificial intelligence visual recognition according to claim 1 is characterized in that: The visual recognition module is used to collect data on the picking position, shape image set, and light intensity of each picked fruit in each picking area corresponding to the picked fruit by the picking robot during picking.

7. The picking manipulator based on artificial intelligence visual recognition according to claim 6 is characterized in that: The fruit state analysis module is used to analyze the picking state of each picking area corresponding to each picked fruit when the picking robot is picking. The specific analysis process is as follows: Extracting the surface image of each picked fruit in the picking area from the shape image set of each picked fruit in each picking area when the picking robot is picking, obtaining the surface image of each picked fruit in the picking area, and obtaining the shape contour and skin texture of each picked fruit in the picking area based on the surface image of each picked fruit in the picking area; Compare the outlines of each picked fruit corresponding to the picking area with the outlines of reference picked fruits stored in the database to obtain the overlapping area of ​​the outlines of each picked fruit corresponding to the picking area and the reference picked fruit, which is recorded as the overlapping area of ​​the outlines of each picked fruit corresponding to the picking area; The reference skin texture of each picked fruit is obtained from the database, and a number of feature points are extracted on the reference skin texture. The skin texture of each picked fruit in the picking area corresponding to the fruit is compared with the skin texture of the reference picked fruit stored in the database for overlap of feature points, and the number of overlaps of feature points in the skin texture of each picked fruit in the picking area corresponding to the fruit and the skin texture of the reference picked fruit is obtained, which is recorded as the number of overlaps of feature points in the skin texture of each picked fruit in the picking area corresponding to the fruit.

8. The picking manipulator based on artificial intelligence visual recognition according to claim 7 is characterized in that: The fruit picking evaluation module is used to evaluate the picking status of each picking area corresponding to each picked fruit when the picking manipulator is picking. The specific evaluation method is as follows: The fruit picking evaluation parameters of each picking area corresponding to each picked fruit are formed by the overlapping area of ​​the picking area corresponding to the outer contour of each picked fruit, the overlapping number of feature points in the skin texture of each picked fruit, and the light intensity of each picking area corresponding to each picked fruit; From the fruit picking evaluation parameters of each picking area corresponding to each picked fruit, the values ​​of the overlap area, overlap number and light intensity of each picking area corresponding to each picked fruit are extracted and recorded as S j i , Q j i and L j i , i represents the number of each picked fruit, i=1,2,...,n, n represents the sum of the numbers of each picked fruit, j represents the number of each picking area, j=1,2,...,m, m represents the sum of the numbers of each picking area; According to the formula The picking state evaluation index of each picked fruit in the picking area is calculated. ΔS is the difference in the allowable overlap area of ​​each picked fruit in the picking area. ΔQ is the difference in the allowable overlap quantity of each picked fruit in the picking area. ΔL is the difference in the allowable light intensity of each picked fruit in the picking area. S j i-1 It is expressed as the overlapping area of ​​the outline of the i-1th picked fruit in the picking area, Q j i-1 It is expressed as the number of overlapping feature points in the skin texture of the i-1th picked fruit in the picking area, L j i-1 It is represented as the light intensity of the picking area corresponding to the i-1th picked fruit, and a1, a2, and a3 represent the weight factors corresponding to the set overlapping area, overlapping number, and light intensity, respectively.

9. The picking manipulator based on artificial intelligence visual recognition according to claim 8 is characterized in that: The picking state evaluation index of each picked fruit in the picking area is compared with the set picking state evaluation index threshold, and the specific comparison results are as follows: If γ i ≥q1, it means that the picking status of each fruit in the picking area is normal, and a normal signal is generated and sent to the execution terminal; If q1>γ i >q2, it means that the picking status of each fruit in the picking area is poor quality, and a poor quality signal is generated and sent to the execution terminal; If γ i ≤q2, it means that the picking status of each fruit in the picking area is abnormal, and an abnormal signal is generated and sent to the execution terminal; Among them, q1 and q2 both represent the picking state evaluation index thresholds of each picked fruit corresponding to the set picking area, and q1>q2>0.

10. The harvesting manipulator based on artificial intelligence visual recognition according to claim 9, characterized in that: The execution terminal performs corresponding operations based on the signal generated by the fruit picking evaluation module, and the specific operation method is as follows: After receiving a normal signal, the execution terminal controls the picking robot to pick the fruit and put it into a normal collection frame. After receiving a poor quality signal, the execution terminal controls the picking robot to pick the fruit and put it into a diseased fruit collection frame. After receiving an abnormal signal, the execution terminal controls the picking robot not to pick the fruit.

Images