Efficient eggplant picking and sorting device and picking and sorting method thereof

Abstract

The invention discloses an efficient eggplant picking and sorting device and a picking and sorting method thereof. The efficient eggplant picking and sorting device comprises a walking mechanism, and a picking mechanism is arranged on the side, close to plants, of the walking mechanism; the picking mechanism comprises a positioning module which can drive a picking manipulator to correspond to the transverse position of a picking target, and the picking manipulator can transversely move back and forth between the walking mechanism and the plants; the positioning module comprises a visual module, and the visual module is located on the picking manipulator and can identify and select a picking target when approaching the walking mechanism; in the process of approaching the plant, guiding the positioning module to correspond to a picking target; when the picking manipulator corresponds to a picking target, a pattern is collected for disease identification, and the sorting mechanism is used for adjusting the longitudinal corresponding relation between the inlets of the storage bins and the outlets of the receiving devices. According to the method, the time of the image recognition period can be reasonably utilized, other actions of picking operation are synchronously completed, the single fruit picking duration is effectively shortened, and the picking efficiency is improved.

Description

Technical Field

[0001] The invention relates to the technical field of eggplant picking. Background Art

[0002] Fruit and vegetable picking robots can replace manual picking, thereby reducing labor intensity and labor costs. Existing picking robots often use multi-degree-of-freedom robotic arms and end-point picking actuators to complete the picking action, and use additional visual systems to assist in target recognition, positioning, and fruit disease identification. However, since the entire identification and picking action needs to be completed in sequence, and the visual system's shooting and recognition process takes a long time, it takes a long time to pick a single fruit, and the picking efficiency is not ideal. Summary of the invention

[0003] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides an efficient eggplant picking and sorting device and a picking and sorting method thereof, which can reasonably utilize the time during image recognition, synchronously complete other actions of the picking operation, effectively shorten the single fruit picking time, and improve the picking efficiency.

[0004] Technical solution: To achieve the above-mentioned purpose, the present invention provides an eggplant efficient picking and sorting device and a picking and sorting method thereof, comprising a walking mechanism, a picking mechanism being arranged on a side close to the plant; the picking mechanism comprising a positioning module, which can drive a picking manipulator to correspond to a lateral position of a picking target, and the picking manipulator can move back and forth between the walking mechanism and the plant;

[0005] The positioning module includes a visual module, which is located on the picking robot and can identify and select picking targets when approaching the walking mechanism; can guide the positioning module to correspond with the picking target in the process of approaching the plant; can collect patterns for disease identification when the picking robot corresponds to the picking target, and can be used for the sorting mechanism to adjust the longitudinal correspondence between its multiple storage bin imports and its receiving device export.

[0006] Furthermore, the positioning module includes a transverse cantilever, which is driven by a dual-degree-of-freedom motion mechanism to move in a vertical plane and extends laterally to point to the target plant; the picking robot is slidably arranged on the upper part of the transverse cantilever, and a limiting structure is arranged at the end of the transverse cantilever away from the walking mechanism, and a telescopic drive device is provided to drive the limiting structure away from or close to the end of the transverse cantilever, and the picking robot corresponds to the picking target when it slides to abut against the limiting surface of the limiting structure; the limiting structure can extend to the bottom side of the picking target.

[0007] Further, the limiting structure includes a limiting plate, and the side of the limiting plate close to the transverse cantilever is the limiting surface; the limiting plate is fixedly connected to the telescopic end of the telescopic driving device, and a photoelectric switch for controlling the telescopic driving device to stop telescoping is arranged on the upper part of the limiting plate.

[0008] Further, the picking manipulator includes a sliding seat, and an attracting device is arranged on the end face of the sliding seat close to the picking target, which can attract the lower part of the picking target and attach to its side surface; a clamping jaw is arranged above the attracting device in a lifting manner, a pneumatic shear is arranged on the upper part of the clamping jaw, and a contour scanning sensor is arranged between the clamping jaw and the pneumatic shear.

[0009] Further, the vision module is located above the clamping jaw, and the vision module includes two camera units. During the ascending process of the clamping jaw, the two camera units can be symmetrically located on both sides of the picking target and respectively face the picking target for shooting.

[0010] Further, color sensors and concave-convex sensors are arranged on the upper parts of the two camera units respectively, which are used to trigger the shooting actions of the two camera units respectively.

[0011] Further, the clamping jaw includes a support body, which can laterally support the upper part of the picking target. Transverse extension members are arranged on both sides of the support body, and swing members are arranged at the ends of the two transverse extension members far from the support body. The two swing members swing relatively close to each other, and can press the upper part of the picking target to fit with the support surface of the support body.

[0012] Further, the two camera units are respectively rotatably arranged on the two transverse extension members, and during the process of the clamping jaw approaching the picking target, the two camera units can always face the picking target.

[0013] Further, an elastic cushion layer is arranged on the support surface of the support body, an elastic wrapping layer is arranged on the outer side of the swing member, and a pressure sensor is embedded in the elastic cushion layer.

[0014] Further, when picking the fruits of a certain plant, the following steps are specifically included:

[0015] S1. The vision module identifies and selects the picking target at a position close to the traveling mechanism;

[0016] S2. Both camera units rotate to face the picking target, and lock the picking target at the center of their respective images; then, through the movement adjustment of the transverse cantilever, the postures of the two camera units are symmetric with each other, and the limiting plate is longitudinally located below the picking target;

[0017] S3. The limiting plate extends to be located below the bottom of the picking target;

[0018] S4. The picking manipulator moves horizontally closer to the picking target. When it contacts the limit plate, the suction device is triggered to adsorb the lower part of the picking target, making the picking target in a vertical posture.

[0019] S5. The clamping jaws and the pneumatic shears move upward synchronously from the lower limit position until the contour scanning sensor detects a rapid narrowing of the contour and then stops. The clamping jaws wrap and hold the upper part of the picking target, and the pneumatic shears cut the pedicel. During the upward movement, the color sensor and the concave-convex sensor scan the entire fruit part of the picking target from bottom to top. When uneven color distribution, a large difference from the preset color, and obvious surface concavities and convexities are detected, the upward movement is paused and the two camera units are triggered to capture and collect image samples for maturity and disease identification.

[0020] S6. The picking manipulator carries the picking target and moves horizontally back to the traveling mechanism until it corresponds vertically with the receiving device of the sorting mechanism. Then, the picking target is dropped into the receiving device and naturally slides into the corresponding storage bin. During this fruit receiving operation, the analysis and identification of the collected image samples are completed. According to the analysis results, the sorting mechanism is controlled to adjust the longitudinal correspondence between the inlet of the corresponding storage bin and the outlet of its receiving device. And the two camera units rotate synchronously to face the direction of the plant and synchronously complete steps S1 - S3. After this fruit receiving operation, steps S4 - S6 are continued and repeated multiple times until all the fruits of the plant are picked.

[0021] Beneficial effects: An efficient eggplant picking and sorting device and its picking and sorting method of the present invention connect the two parts that need to involve visual processing in the whole picking and sorting operation and complete them synchronously with the remaining other actions, avoiding the mutual waiting between the picking action and the recognition action, and thus there will be no waiting time other than the image processing reaction time, so as to greatly shorten the picking time of a single fruit and ensure the continuity of multi-fruit picking of a single plant, effectively improving the picking efficiency. Description of the Drawings

[0022] Figure 1 It is a schematic diagram of the overall structure of an efficient eggplant picking and sorting device of the present invention;

[0023] Figure 2 It is a schematic diagram of the structural relationship between the clamping jaws and the camera module in an embodiment of the present invention;

[0024] Figure 3 It is a schematic diagram of the structure of the sorting mechanism in an embodiment of the present invention and the relative movement relationship with the transverse cantilever;

[0025] Figure 4 It is a schematic diagram of the continuous picking process of multiple fruits of a single plant of the present invention. Detailed Embodiments

[0026] The present invention will be further described in detail below with reference to the drawings.

[0027] As attached Figures 1-4 An efficient eggplant picking and sorting device and its picking and sorting method, including a traveling mechanism 1 that can walk linearly along the arrangement direction of the plants between the plant rows; a picking mechanism is arranged on one side of the traveling mechanism 1 close to the plants, and the picking mechanism includes a positioning module that can drive the picking manipulator 2 to correspond to the picking target 3 in the horizontal position;

[0028] The picking manipulator 2 can reciprocate horizontally between the traveling mechanism 1 and the plants, and when approaching the plants, it can wrap, clamp and cut off the picking target 3 on the plants; during the process of horizontally moving closer to the traveling mechanism 1, it carries the cut picking target 3 to correspond vertically to the receiving device 4 of the sorting mechanism;

[0029] The positioning module includes a vision module. The vision module is located on the picking manipulator 2 and can, when approaching the traveling mechanism 1, identify and select the picking target 3; during the process of approaching the plants, it can guide the positioning module to correspond to the picking target 3; when the picking manipulator 2 corresponds to the picking target 3, it can collect patterns for disease identification, and is used for the sorting mechanism to adjust the longitudinal corresponding relationship between the inlets 5 of its multiple storage bins and the outlet of the receiving device 4.

[0030] Based on the above structure, the basic picking sequence of this solution is that after the traveling mechanism corresponds to a certain plant position, first identify and select the optimal picking target, then position the picking manipulator to correspond to the target fruit horizontally, then extend the picking manipulator to complete the picking action, and finally bring back the fruit for sorting and storage, so as to complete a complete picking and sorting action. During this process, the vision module plays two roles. One is to realize the identification and selection of the optimal picking target in the state of being far from the plants; the other is to realize the collection of abnormal patterns of the picking target in the state of being close to the plants for disease identification, so as to carry out sorting and storage. Before picking the target fruit, the picking manipulator will surely approach the plants, and after picking, it will surely move away from the plants to recycle the picked fruit. Then, by making the vision module follow the picking manipulator, it can immediately enter the identification and selection of the next picking target after the previous fruit is picked. This makes full use of the vision module in the whole operation, allowing it to alternate between two modes, so that it will not be idle during continuous picking operations. That is, the two parts that need to involve vision processing in the whole picking and sorting operation are connected together and synchronized with the remaining other actions, avoiding the mutual waiting between the picking action and the identification action, and thus there will be no waiting time other than the image processing reaction time, so as to greatly shorten the picking time of a single fruit and ensure the continuity of multi-fruit picking of a single plant, effectively improving the picking efficiency.

[0031] Moreover, the vision module collects image samples at close range for disease identification. The features in the patterns will be larger and clearer. When compared with the control patterns, the response speed will be faster, thus significantly shortening the disease identification process. During the disease identification process, the picking and recycling of fruits are also synchronously completed, effectively shortening the duration of this part of the operation.

[0032] Moreover, although the time required for the picking manipulator to traverse back to the traveling mechanism is short, when it returns to be vertically aligned with the receiving device 4 of the sorting mechanism, the disease identification result may not have been generated yet or may just have been generated. Therefore, when the picking manipulator returns to its position, it may need to wait for a period of time for the sorting mechanism to respond according to the identification result before it can drop the fruit into the receiving device 4. However, this also enables the vision module to quickly enter the identification and selection of the next target. The reason is that as the vision module moves away from the plant, its field of view gradually expands. The faster it moves away, the faster the field of view expands. When the field of view is large enough to cover all the fruits on the plant, the selection of the next picking target can be carried out. Moreover, as the picking continues, the number of fruits on the plant decreases, and the required field of view range may correspondingly decrease, enabling the vision module to enter the identification and selection of the next picking target shortly after moving away from the plant. During the picking process of individual plants, the phenomenon of an increasingly faster single-fruit picking speed may occur.

[0033] In the specific identification and selection process, during the first identification, all recognizable objects within the azimuth view are counted and the count is decremented after each picking. Starting from the second identification and selection, as the field of view of the vision module increases, when there are two recognizable objects in the field of view, a picking priority comparison is made to obtain a preferred picking target. Every time a complete contour object enters the field of view, a picking priority comparison is made again with the previous preferred picking target until the number of fruits in the field of view reaches the count value. The positioning module can then execute the positioning action based on the current optimal picking target. In this way, the image recognition process of the disease and the visual selection and positioning adjustment process of the next target also have time overlap and can be almost synchronously completed, with only a short time gap before and after. This further shortens the duration of continuous operation and significantly improves the continuous picking efficiency.

[0034] Embodiment: The positioning module includes a lateral cantilever 6, which is driven by a two-degree-of-freedom motion mechanism to move in a vertical plane, including a transverse movement along the traveling direction of the traveling mechanism and a lifting movement in the vertical direction, and its longitudinal direction extends horizontally towards the target plant; the picking manipulator 2 is slidably arranged on the upper part of the lateral cantilever 6, a limiting structure is arranged at the end of the lateral cantilever 6 far from the traveling mechanism 1, and a telescopic driving device for driving the limiting structure to move away from or close to the end of the lateral cantilever 6, when the picking manipulator 2 slides to abut against the limiting surface 71 of the limiting structure, it corresponds to the picking target 3; the limiting structure can extend to the bottom side of the picking target 3.

[0035] The limiting structure includes a limiting plate 7, and the side of the limiting plate 7 close to the lateral cantilever 6 is the limiting surface 71; the limiting plate 7 is fixedly connected to the telescopic end of the telescopic driving device, and a photoelectric switch for controlling the telescopic driving device to stop telescoping is arranged on the upper part of the limiting plate 7.

[0036] Based on this, when the disease recognition result is not generated, the limiting plate can be first positioned at the bottom side of the picking target by the cooperation of the actions of the lateral cantilever 6 and the telescopic driving device through the control of the photoelectric switch and the vision module. Once the disease recognition result is generated and the picking manipulator completes the fruit dropping action, it can immediately move horizontally close to the picking target and, when contacting the limiting plate, just correspond to the picking target, so that a new round of pattern collection and picking actions can be quickly executed, meeting the high coincidence possibility of the disease image recognition process and the visual selection and positioning adjustment process of the next target.

[0037] Preferably, the picking manipulator 2 includes a sliding seat 21. An attracting device 22 is arranged on the end face of the sliding seat 21 close to the picking target 3, which can attract the lower part of the picking target 3 and adhere to its side surface. The attracting device 22 can adopt a negative pressure suction cup. A switch for controlling the start of the attracting device 22 can be arranged on the end face of the sliding seat. When it contacts the limiting surface of the limiting plate, the attracting action can be triggered by impact, and the obliquely growing fruit body can be adsorbed to the vertical posture as much as possible, so as to ensure more comprehensive image acquisition in the subsequent pattern acquisition and make the fruit more stable during picking. An upper clamping jaw 23 is arranged above the attracting device 22 in a lifting manner, and a pneumatic shear 24 is arranged above the clamping jaw 23. A contour scanning sensor is arranged between the clamping jaw 23 and the pneumatic shear 24. During the synchronous upward movement of the clamping jaw and the pneumatic shear, the contour of the eggplant is scanned by the contour scanning sensor. When the scanning passes through the fruit stalk, the contour will quickly become thinner. At this time, stop the upward movement, and the clamping jaw can just correspond to the upper connection part of the fruit stalk and the fruit body. This place is relatively harder than the fruit body, and a greater clamping force can be applied. Combining with the adsorption effect of the attracting device 22 on the lower part, the eggplant fruit can be firmly fixed without damaging the eggplant fruit body, and the pneumatic shear is just located at the stalk. After the clamping jaw is clamped in place, the stalk can be cut off to complete the picking action.

[0038] The vision module is located above the clamping jaw 23. The vision module includes two camera units 8. During the upward movement of the clamping jaw 23, the two camera units 8 can be symmetrically located on both sides of the picking target 3 and respectively face the picking target 3 for shooting. The visual fields of the two cameras just sweep across the whole eggplant fruit body during the upward movement, and then relatively comprehensive image samples can be collected.

[0039] Color sensors and concave-convex sensors are arranged above both of the camera units 8, which are used to trigger the shooting actions of the two camera units 8 respectively. Taking the eggplant mature color spectrum as a reference, through scanning and comparing the color approximation degree, as well as the uniformity and surface smoothness of the eggplant surface color, a preliminary judgment is made. For individuals with a smooth and bright surface color, there is no need to take photos for sampling. For individuals that may have diseases, when a certain abnormal feature is detected by scanning, pattern acquisition is carried out again, which can greatly shorten the duration of pattern acquisition and collect key patterns for disease identification, reduce the amount of image processing for disease identification, and lower the processing difficulty, effectively improving the pattern acquisition efficiency and disease identification efficiency.

[0040] Preferably, the clamping jaw 23 includes a support body 231, which can laterally support the upper part of the picking target 3. Transverse extension members 232 are arranged on both sides of the support body 231. Swing members 233 are arranged at the ends of the two transverse extension members 232 away from the support body 231. The two swing members 233 swing relatively close to each other, and can press the upper part of the picking target 3 to fit with the support surface of the support body 231.

[0041] The two camera units 8 are respectively rotatably arranged on the two laterally extending members 232, so as to realize free switching between the target recognition selection and pattern acquisition modes; and during the process of the gripper 23 approaching the picking target 3, the two camera units 8 can always face the picking target 3 directly. During the process of the picking manipulator sliding close to the picking target 3, the position of the limiting plate can be further finely adjusted, so as to determine the final docking position of the picking manipulator, and ensure good pattern acquisition effect and picking effect.

[0042] An elastic cushion layer is arranged on the supporting surface of the support body 231, an elastic wrapping layer is arranged on the outer side of the swinging member 233, and a pressure sensor is buried in the elastic cushion layer. The clamping force can be reasonably controlled through the feedback of the pressure sensor to ensure that the fruit is not damaged by clamping. In the picking operation with the requirement of screening the fruit maturity, it can also be combined with a color sensor. When a color difference is detected on the surface of the eggplant, the gripper gently clamps the fruit body appropriately, and then slowly and uniformly opens. The elasticity and softness of the eggplant are judged by detecting the numerical value of the pressure sensor, so as to judge the maturity.

[0043] The receiving device 4 can adopt a bucket-shaped structure or a ramp structure, and an outlet allowing a single eggplant to pass through is arranged at the lower end. The bucket opening or the slope surface is relatively wide, and the transverse cantilever 6 can be horizontally moved to any position along the traveling direction of the traveling mechanism and can be located above the receiving device 4. The inlets 5 of the multiple storage bins of the sorting mechanism are arranged in a row on a moving adjustment mechanism. Through moving adjustment, the multiple inlets 5 of the storage bins can be sequentially corresponding to the lower end outlet of the receiving device 4. Each inlet 5 of the storage bin is communicated with each storage bin through a flexible mesh cloth channel, which can make the adjustment of the inlet 5 of the storage bin more flexible and has a certain buffering and protecting effect on the falling eggplant.

[0044] When picking the fruits of a certain plant, the following steps are specifically included:

[0045] S1. The vision module recognizes and selects the picking target 3 at a position close to the traveling mechanism;

[0046] S2. The two camera units 8 are both rotated to face the picking target 3 directly, and the picking target 3 is locked at the image center of each of them; then through the movement adjustment of the transverse cantilever 6, the postures of the two camera units 8 are made symmetrical to each other, and the limiting plate 7 is longitudinally located below the picking target 3;

[0047] S3. The limiting plate 7 extends to be located below the bottom of the picking target 3;

[0048] S4. The picking manipulator 2 horizontally moves close to the picking target 3, and when it contacts the limiting plate 7, the attracting device 22 is triggered to adsorb the lower part of the picking target 3, so that the picking target 3 is in a vertical posture;

[0049] S5. The gripper 23 and the pneumatic shear 24 move up synchronously from the lower limit position until they stop when the contour scanning sensor detects that the contour rapidly tapers. The gripper 23 wraps and holds the upper part of the picking target 3, and the pneumatic shear 24 cuts the pedicel. During the upward movement, the color sensor and the concave-convex sensor scan the entire fruit part of the picking target 3 from bottom to top. When uneven color distribution, a large difference from the preset color, and obvious concavities and convexities on the surface are detected, the upward movement is paused and the two camera units 8 are triggered to capture image samples for maturity and disease identification.

[0050] S6. The picking manipulator 2 carries the picking target 3 and traverses back to the traveling mechanism 1 until it corresponds vertically to the receiving device 4 of the sorting mechanism. Then, the picking target 3 is dropped into the receiving device 4 and naturally slides down to the corresponding storage bin. During this fruit receiving operation, the analysis and identification of the captured image samples are completed, and according to the analysis results, the sorting mechanism is controlled to adjust the vertical correspondence between the inlet 5 of the corresponding storage bin and the outlet of its receiving device 4. Moreover, the two camera units 8 rotate synchronously to face the direction of the plant and synchronously complete steps S1 - S3. After this fruit receiving operation, steps S3 - S6 are continued, and the cycle is repeated multiple times until all the fruits of the plant are picked.

[0051] The above description is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art of this technology, without departing from the above principles of the present invention, several improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the present invention.

Claims

1. An efficient eggplant picking and sorting device, characterized in that: It includes a traveling mechanism (1), and a picking mechanism is arranged on one side close to the plant; the picking mechanism includes a positioning module, which can drive the picking manipulator (2) to correspond to the picking target (3) in the lateral position, and the picking manipulator (2) can reciprocate laterally between the traveling mechanism (1) and the plant; The positioning module includes a vision module, which is located on the picking manipulator (2), and can identify and select the picking target (3) when approaching the traveling mechanism (1); can guide the correspondence between the positioning module and the picking target (3) during the process of approaching the plant; can collect patterns for disease identification when the picking manipulator (2) corresponds to the picking target (3), and is used for the sorting mechanism to adjust the longitudinal correspondence between the inlets (5) of its multiple storage bins and the outlet of its receiving device (4).

2. The high-efficiency eggplant picking and sorting device according to claim 1, wherein: The positioning module includes a lateral cantilever (6), which is driven by a two-degree-of-freedom motion mechanism to move in a vertical plane and extends laterally towards the target plant; the picking manipulator (2) is slidably arranged on the upper part of the lateral cantilever (6), and a limiting structure is arranged at the end of the lateral cantilever (6) far from the traveling mechanism (1), and a telescopic driving device for driving the limiting structure to move away from or close to the end of the lateral cantilever (6) is provided. The picking manipulator (2) corresponds to the picking target (3) when it slides to abut against the limiting surface (71) of the limiting structure; the limiting structure can extend to the bottom side of the picking target (3).

3. The efficient eggplant picking and sorting device according to claim 2, characterized in that: The limiting structure includes a limiting plate (7), and the side of the limiting plate (7) close to the lateral cantilever (6) is the limiting surface (71); the limiting plate (7) is fixedly connected to the telescopic end of the telescopic driving device, and a photoelectric switch for controlling the telescopic driving device to stop telescoping is arranged on the upper part of the limiting plate (7).

4. An efficient eggplant picking and sorting device according to claim 3, characterized in that: The picking manipulator (2) includes a sliding seat (21), and an attracting device (22) is arranged on the end face of the sliding seat (21) close to the picking target (3), which can attract the lower part of the picking target (3) and adhere to its side surface; a clamping jaw (23) is arranged to lift above the attracting device (22), a pneumatic shear (24) is arranged on the upper part of the clamping jaw (23), and a contour scanning sensor is arranged between the clamping jaw (23) and the pneumatic shear (24).

5. An efficient eggplant picking and sorting device according to claim 4, characterized in that: The vision module is located above the clamping jaw (23), and the vision module includes two camera units (8). During the rising process of the clamping jaw (23), the two camera units (8) can be symmetrically located on both sides of the picking target (3) and respectively face the picking target (3) for shooting.

6. The efficient eggplant picking and sorting device according to claim 5, characterized in that: Color sensors and concave-convex sensors are arranged on the upper parts of the two camera units (8) respectively, which are used to trigger the shooting actions of the two camera units (8) respectively.

7. An efficient eggplant picking and sorting device according to claim 5, characterized in that: The clamping jaws (23) include a support body (231) capable of laterally supporting the upper part of the picking target (3). Transverse extensions (232) are arranged on both sides of the support body (231). Swing members (233) are arranged at the ends of the two transverse extensions (232) away from the support body (231). The two swing members (233) swing relatively close to each other, and can press the upper part of the picking target (3) against the support surface of the support body (231).

8. The efficient eggplant picking and sorting device according to claim 7, wherein: The two camera units (8) are respectively rotatably arranged on the two transverse extensions (232), and during the process of the clamping jaws (23) approaching the picking target (3), the two camera units (8) can always face the picking target (3) directly.

9. The efficient eggplant picking and sorting device according to claim 8, characterized in that: An elastic cushion layer is arranged on the support surface of the support body (231), an elastic wrapping layer is arranged on the outer side of the swing member (233), and a pressure sensor is embedded in the elastic cushion layer.

10. A picking and sorting method for an efficient eggplant picking and sorting device according to any one of claims 1-9, characterized in that: When picking the fruits of a certain plant, the following steps are specifically included: S1. The vision module identifies and selects the picking target (3) at a position close to the traveling mechanism. S2. The two camera units (8) are both rotated to face the picking target (3) directly, and the picking target (3) is locked at the image center of each of them; then, through the movement adjustment of the transverse cantilever (6), the postures of the two camera units (8) are made symmetrical to each other, and the limiting plate (7) is longitudinally located below the picking target (3). S3. The limiting plate (7) extends to be located below the bottom of the picking target (3). S4. The picking manipulator (2) moves horizontally close to the picking target (3), and when it contacts the limiting plate (7), the attracting device (22) is triggered to adsorb the lower part of the picking target (3), so that the picking target (3) is in a vertical posture. S5. The clamping jaws (23) and the pneumatic shears (24) synchronously move upward from the lower limit position until the contour scanning sensor detects that the contour rapidly becomes thinner and then stops. The clamping jaws (23) wrap and clamp the upper part of the picking target (3), and the pneumatic shears (24) cut off the pedicel; during the upward movement, the color sensor and the concave-convex sensor sweep across the entire fruit part of the picking target (3) from bottom to top. When uneven color distribution, a large difference from the preset color, and obvious surface concavities and convexities are detected, the upward movement is paused and the two camera units (8) are triggered to take and collect image samples for maturity and disease identification. S6. The picking manipulator (2) carries the picking target (3) and moves horizontally back to the traveling mechanism (1) until it corresponds vertically to the receiving device (4) of the sorting mechanism. Then, the picking target (3) is dropped into the receiving device (4) and naturally slides into the corresponding storage bin; during this fruit receiving operation process, the analysis and identification of the collected image samples are completed, and according to the analysis results, the sorting mechanism is controlled to adjust the longitudinal correspondence between the inlet (5) of the corresponding storage bin and the outlet of its receiving device (4); and the two camera units (8) synchronously rotate to face the plant direction and synchronously complete steps S1 - S3; after this fruit receiving operation, steps S3 - S6 are continued, and the cycle is repeated multiple times until all the fruits of this plant are picked.

Images