Small intelligent rapid sorting system and method for multi-varieties of citrus

Abstract

The present application relates to the technical field of citrus sorting, in particular to a small intelligent rapid sorting system and method for multi-variety citrus, which realizes orderly single-fruit conveying of the citrus to be sorted through multi-stage speed regulation control, collects weight data and appearance data of the citrus to be sorted in the same conveying process, and combines preset multi-variety citrus classification standards and a fuzzy comprehensive evaluation algorithm to complete quality judgment and sorting bin matching, so as to realize rapid, accurate, non-destructive and intelligent grading and sorting of multi-variety citrus, solve the core problem that existing sorting equipment cannot simultaneously consider miniaturization, high accuracy, multi-variety adaptation and low-cost operation, and fully meet the intelligent sorting operation needs of the majority of small and medium-sized planting subjects.

Description

Technical Field

[0001] This invention relates to the field of citrus sorting technology, specifically to a small-scale intelligent rapid sorting system and method for multiple varieties of citrus. Background Technology

[0002] Citrus is one of the most widely planted and economically valuable specialty fruits in my country. Post-harvest grading and sorting is the core process for realizing the commercialization of citrus and increasing the added value of products. The accuracy, efficiency and fruit damage rate of sorting directly determine the market circulation grade of citrus and the economic benefits of growers.

[0003] Currently, my country's citrus planting industry is characterized by significant fragmentation, with small and medium-sized farmers and small family farms being the main growers. Industry data shows that 85% of citrus growers in my country are small-scale farmers and family farms with planting areas of less than 50 mu (approximately 3.3 hectares), and over 65% of online citrus orders are directly shipped from these businesses. However, due to their small planting scale, diverse varieties, dispersed operations, and limited capital investment, these small and medium-sized businesses still generally rely on inefficient, subjective, and high-damage manual sorting methods.

[0004] With the upgrading of domestic consumption and the transformation of agricultural product distribution, new sales models such as fresh food e-commerce have rapidly emerged, placing higher demands on the standardization of grading, consistency of quality control, and timeliness of delivery of citrus fruits. The traditional manual sorting method used by small and medium-sized citrus growers can no longer meet current production and sales needs, making automated sorting an inevitable trend. However, existing automated citrus sorting technologies are all developed for centralized sorting scenarios in large cooperatives and orchards. Their equipment form, core architecture, and functional design are severely out of sync with the actual operational needs of the vast majority of small and medium-sized growers, resulting in many insurmountable technical defects, as follows:

[0005] (1) Existing citrus sorting technology solutions are all designed around the large-scale centralized sorting scenario. The equipment is mainly large-scale fixed or vehicle-mounted integrated. It has inherent problems such as large footprint, high purchase and use costs, and limited operating scenarios. It is completely unable to meet the small-scale decentralized operation needs of small and medium-sized farmers and family farms.

[0006] (2) Existing technologies can only detect basic appearance indicators such as size, color and shape of citrus fruits. They cannot complete the quantitative identification of fruit surface defects according to national industry standards. Furthermore, weight detection and appearance detection are separated from each other, making it impossible to achieve simultaneous comprehensive judgment of multi-dimensional quality. The grading accuracy and standardization are seriously insufficient.

[0007] (3) Existing technologies all adopt a step-by-step process architecture of "detection-sorting". The detection process and the sorting process are separated from each other, which not only leads to complex equipment structure and lengthy process, but also makes it easy to cause grading errors due to the timing misalignment of detection and sorting actions. It is difficult to balance operating efficiency and grading accuracy.

[0008] (4) Existing technologies all adopt fixed grading parameters for a single variety, without a multi-variety sorting parameter library and a one-click quick switching function, which cannot adapt to the current situation of multi-variety mixed planting and small-batch multi-batch sorting in my country's citrus producing areas, and the equipment has extremely poor versatility.

[0009] For example, the patent application CN109332200B, titled "A Mobile On-site Quality Testing and Grading Method for Citrus Fruits," describes a mobile sorting system built on a truck. It achieves on-site grading of citrus orchards through a method of baffle-type feeding, single-fruit conveying, industrial camera appearance inspection, and solenoid valve impact dropping. However, the equipment relies on a dedicated vehicle carrier, resulting in high investment costs. It can only complete basic appearance indicators such as size, color, and shape. The step-by-step execution of testing and sorting is prone to grading deviations. Furthermore, it is only suitable for fixed grading standards for a single variety and cannot meet the sorting needs of small and medium-sized farmers for multiple varieties and low costs.

[0010] For example, the "Automatic Arrangement and Sorting Device for Citrus" published by CN119500612B uses a large fixed frame structure. It uses an industrial camera to visually screen the quality of the fruit and uses a tilting turntable to sort the fruit by aperture size. However, it is a large fixed device with limited application scenarios. It can only achieve basic screening of good and bad fruit and fruit size grading. Visual inspection and fruit size grading are completed step by step, which is a lengthy process. Moreover, the grading structure is only suitable for the fruit size specifications of a single variety, and its adaptability to multiple varieties is extremely poor. It cannot meet the needs of small and medium-sized farmers for decentralized operations.

[0011] In summary, existing automated citrus sorting technologies are all developed around large-scale centralized sorting scenarios. They suffer from several common technical defects in terms of equipment form, core architecture, and functional design, which cannot be overcome simultaneously. They cannot meet the core needs of small and medium-sized growers for sorting equipment that is miniaturized, low-cost, adaptable to multiple varieties, precise grading, highly reliable operation, and low-damage sorting. This seriously restricts the improvement of the post-harvest commercialization rate of citrus in my country and the growth of planting income for small and medium-sized farmers. Summary of the Invention

[0012] The purpose of this invention is to address the shortcomings of existing technologies by providing a small-scale intelligent rapid sorting system and method for multiple varieties of citrus. Through multi-level speed control, the system achieves orderly conveying of individual citrus fruits to be sorted, and collects weight and appearance data of the citrus fruits during the same conveying process. Combining preset multi-variety citrus classification standards with a fuzzy comprehensive evaluation algorithm, the system completes quality judgment and sorting bin matching. This enables rapid, accurate, non-destructive, and intelligent grading and sorting of multiple citrus varieties, solving the core problems of existing sorting equipment's inability to simultaneously achieve miniaturization, high precision, multi-variety adaptability, and low-cost operation. It can fully meet the intelligent sorting needs of a wide range of small and medium-sized growers.

[0013] The objective of this invention is achieved through the following approach:

[0014] A small, intelligent, rapid sorting system for multiple citrus varieties includes a desktop or mobile frame, a drive assembly, a human-machine interface module, a feeding and cleaning module, an integrated sorting and collection module, and a control and monitoring module.

[0015] The human-machine interaction module is used to provide a selection interface for sorting standards of multiple varieties of citrus and a manual correction operation interface, and to display the overall machine operation, sorting output, and equipment fault information. At the same time, it transmits the selected citrus sorting standards and manual correction information to the control and monitoring module.

[0016] The feeding and impurity removal module is used to automatically convey citrus fruits. It uses multiple conveyor belts that increase speed in sequence to create speed differences to separate dense citrus fruits. Combined with the arc-shaped closing structure at the tail and multiple baffles, it enables the orderly dropping of individual citrus fruits. The fan speed is dynamically matched according to the diameter of the citrus fruits to be sorted, and impurities are blown away in a directional manner.

[0017] The integrated sorting and collection module is used to carry individual citrus fruits through multiple independent weighing and unloading units arranged circumferentially by the annular material conveying unit. These units carry the fruits in a closed loop and transfer them along the loop. During the loop transfer, the weighing and unloading units acquire the weight of the citrus fruits, and the image detection unit acquires the surface defect indicators of the citrus fruits. The weight, surface defect indicators, and the ID of the weighing and unloading unit are sent to the control and monitoring module to obtain sorting and unloading instructions. When the annular material conveying unit transports the citrus fruits to the sorting bin specified by the sorting and unloading instructions, the weighing and unloading units dump the citrus fruits to complete the unloading. The sorting bins collect the citrus fruits according to their quality grades. The surface defect indicators include the number of spots and the diameter of the spots.

[0018] The control and monitoring module is used to calculate the quality grade of citrus fruits based on their weight and surface defect indicators using a fuzzy comprehensive evaluation algorithm. It also generates sorting and unloading instructions by combining the citrus sorting standards issued by the human-machine interaction module. Simultaneously, it monitors the material jamming faults of the feeding and impurity removal module and the integrated sorting and collection module in real time. Upon detecting an abnormality, it triggers an audible and visual alarm and controls the entire machine to shut down for protection. It coordinates the entire process of the human-machine interaction module, the feeding and impurity removal module, and the integrated sorting and collection module to work together.

[0019] Preferably, the human-computer interaction module is configured with an independent manual correction operation interface, allowing operators to perform two types of correction operations:

[0020] (1) Correction of single fruit grading results: retrieve the weight, surface defect index and grading results of sorted citrus, manually modify the quality grade labels of the automatic grading results that have deviations, and upload the modified quality grade labels to the control and supervision module as input data for the self-learning optimization module. The quality grade labels include special grade, first grade, second grade and third grade.

[0021] (2) Correction of sorting standards for multiple varieties of citrus: Select the citrus variety to be modified, manually modify the fuzzy membership boundary parameters, weights of each quality indicator and thresholds of each quality grade for the citrus variety, and after the modification is completed, send the updated classification standards for multiple varieties of citrus to the control and supervision module as input data for the self-learning optimization module.

[0022] Preferably, the integrated sorting and collection module includes a ring-shaped material conveying unit, an image detection unit, and a weighing and unloading unit, wherein...

[0023] The circular material conveying unit is used to carry multiple weighing and unloading units to continuously and cyclically convey single citrus fruits along a closed circular path.

[0024] The image detection unit is used to acquire front and back images of citrus fruits, and obtain surface defect indicators of citrus fruits by combining them with the built-in image processor, and transmit the surface defect indicators to the control and monitoring module.

[0025] The weighing and unloading unit is used to collect the weight data of a single fruit and transmit the weight data and the ID of the current weighing and unloading unit to the control and monitoring module. At the same time, it receives the sorting and unloading instructions issued by the control and monitoring module. When the weighing and unloading unit rotates to the top of the designated sorting bin, it pours the citrus to complete the unloading and automatically clears the weight data of the citrus. The citrus then returns to the feeding end of the circular material conveyor unit for cyclic operation.

[0026] Preferably, the feeding and impurity removal module includes an automatic feeding unit, a multi-stage speed-regulating sorting unit, and an adjustable airflow impurity removal unit, wherein...

[0027] The automatic feeding unit is used to transport citrus fruits from a low position to a feeding station via a stepped conveyor belt and automatically dump the citrus fruits.

[0028] The multi-stage speed-regulating fruit sorting unit is used to generate a speed difference through multiple sections of conveyor belts that increase speed in sequence, thereby breaking up the stacked citrus fruits and increasing the spacing between the fruits. At the same time, a pair of limiting baffles set on both sides of the conveyor belt, with the spacing gradually narrowing along the conveying direction, gather and limit the citrus fruits. After being gathered and arranged by the arc-shaped closing structure at the end of the conveyor belt, the fruits are discharged in an orderly manner with the help of the liftable baffles. The liftable baffles are linked to the rotation position of the weighing and unloading unit.

[0029] The adjustable wind power cleaning unit is used to match the corresponding blowing wind power according to the current fruit diameter of the sorted citrus, and remove residual leaves, fruit debris and field debris attached to the surface of the citrus through directional airflow.

[0030] Preferably, the control and monitoring module includes a sorting instruction generation unit, a fault alarm protection unit, and a self-learning optimization unit, wherein...

[0031] The sorting instruction generation unit is used to receive the citrus sorting standards issued by the human-computer interaction module, calculate the comprehensive quality grade of the citrus based on the weight and surface defect index of the citrus using a fuzzy comprehensive evaluation algorithm, and generate sorting and unloading instructions in combination with the citrus sorting standards.

[0032] The fault alarm protection unit is used to monitor the material jamming fault of the feeding and impurity removal module and the integrated sorting and collection module in real time. After detecting the material jamming abnormality, it triggers an audible and visual alarm and controls the whole machine to stop for protection.

[0033] The self-learning optimization unit receives manually corrected information uploaded by the human-computer interaction module as input. It incorporates a small-sample gradient descent fine-tuning strategy suitable for embedded controllers, aiming to minimize the error between the system's automatic grading prediction results and the manually corrected information. It iteratively updates the weight parameters corresponding to each quality indicator in the multi-variety citrus classification standard. After the iterative optimization is completed, the updated set of indicator weights is synchronously stored in the multi-variety citrus classification standard database of the sorting instruction generation unit for subsequent citrus grading judgment, continuously reducing the system's grading misjudgment rate.

[0034] The method for rapid sorting of multiple varieties of citrus using the above-mentioned small intelligent rapid sorting system includes the following steps:

[0035] 1) Select the citrus sorting criteria in the human-computer interaction module and transmit the selected citrus sorting criteria to the control and monitoring module;

[0036] 2) The citrus fruits to be sorted are automatically fed by the feeding and impurity removal module. After being gradually accelerated and dispersed into individual fruits, the individual citrus fruits are output to the feeding end of the integrated sorting and collection module in sequence.

[0037] 3) The integrated sorting and collection module uses a ring-shaped material conveying unit to transport the citrus to be sorted along a ring path. During the transport process, the weighing and unloading unit collects the weight of the citrus to be sorted, and the image detection unit obtains the surface defect index of the citrus to be sorted. The weight of the citrus to be sorted, the surface defect index, and the ID of the weighing and unloading unit are transmitted to the control and monitoring module.

[0038] 4) The control and monitoring module matches and binds the weight and surface defect index of the citrus to be sorted with the ID of the weighing and unloading unit. It uses a fuzzy comprehensive evaluation algorithm to calculate the comprehensive quality grade of the citrus to be sorted, and generates a sorting and unloading instruction for the citrus to be sorted in combination with the citrus sorting standard. The sorting and unloading instruction is then transmitted to the integrated sorting and collection module.

[0039] 5) According to the sorting and unloading command, the ring-shaped material conveying unit of the integrated sorting and collection module continues to rotate. When the weighing and unloading unit carrying the citrus to be sorted rotates to the top of the sorting bin specified by the sorting and unloading command, the weighing and unloading unit performs the unloading operation to complete the grading and collection of citrus.

[0040] Preferably, in step 4), the sorting instruction is generated in the following manner:

[0041] 4-1) Assign a unique number to each sorting bin and a unique identification ID to each weighing and unloading unit in advance;

[0042] 4-2) The citrus fruits to be sorted fall above the weighing and unloading unit, which automatically collects the weight information of the citrus fruits to be sorted and transmits it to the control and monitoring module along with the ID of the weighing and unloading unit.

[0043] 4-3) When the citrus fruits to be sorted enter the image acquisition range of the image detection unit, the image detection unit acquires the front and back images of the citrus fruits to be sorted, and then uses image processing or deep learning algorithms to obtain the surface defect indicators of the citrus fruits to be sorted and sends them to the control and monitoring module.

[0044] 4-4) In the control and monitoring module, the received weight and surface defect information are matched, and the weight information and surface defect information of the same citrus to be sorted are bound and stored with the ID of the weighing and unloading unit carrying the citrus to be sorted;

[0045] 4-5) In the control and supervision module, the fuzzy comprehensive evaluation algorithm is used to calculate the comprehensive grading judgment value of the citrus to be sorted based on the weight and surface defect index of the citrus to be sorted.

[0046] 4-6) In the control and supervision module, the calculated comprehensive grading judgment value is compared with the preset threshold range of each quality grade in the citrus sorting standard to determine the final quality grade of the citrus to be sorted.

[0047] 4-7) In the control and supervision module, based on the pre-established mapping relationship between quality grade and sorting bin number, the final quality grade is matched to the corresponding target sorting bin number, and sorting instructions for the citrus to be sorted are generated.

[0048] Preferably, in steps 4-5), the comprehensive grading value of the citrus fruits to be sorted is calculated in the following manner:

[0049] 4-5-1) The weight and surface defect index of the citrus to be sorted are used as quality evaluation indicators. The fuzzy membership function is used to map each quality evaluation indicator to a continuous membership value in the range of 0 to 1, which respectively characterizes the degree of conformity of the citrus to be sorted as special grade, first grade, second grade and third grade. Among them, the weight adopts the trapezoidal fuzzy membership function, and the number of spots and the diameter of spots adopt the Z-type fuzzy membership function.

[0050] 4-5-2) The values ​​of each quality indicator of the citrus to be sorted are fuzzified, and the membership degree of each quality indicator under the special grade, first grade, second grade and third grade is calculated respectively.

[0051] 4-5-3) Retrieve the weights of each quality evaluation indicator in the citrus sorting standard. The sum of all indicator weights equals 1. Calculate the membership degree of each dimension in the membership degree matrix by weighted summation to obtain the comprehensive membership degree corresponding to each grade: special grade, first grade, second grade, and third grade.

[0052]

[0053] In the formula, w represents the overall membership degree of a certain level k. i Let i be the weight of the i-th quality evaluation index. Let i be the membership degree of the i-th indicator at the k-th level. The total number of quality evaluation indicators;

[0054] 4-5-4) Compare the comprehensive membership degrees of the special grade, first grade, second grade and third grade, and take the comprehensive membership degree with the largest value as the comprehensive grading judgment value of the citrus to be sorted.

[0055] Preferably, in step 5), after the citrus is graded and collected, the control and monitoring module will clear all the detection data and sorting instructions of the corresponding weighing and unloading unit. At the same time, the weighing and unloading unit returns to the feeding and receiving position with the circular material conveying unit to receive the next citrus to be sorted, thereby realizing the continuous grading and collection of citrus.

[0056] The beneficial effects of this invention include the following:

[0057] ① The existing mainstream citrus sorting equipment is mainly divided into two mature solutions: one is large fixed frame equipment, with a single set purchase cost of up to 80,000 yuan; the other is truck-mounted integrated design equipment, with a single set cost of up to 150,000 yuan. The high investment greatly raises the threshold for using the equipment. The overall size (length × width × height) of the equipment in these two solutions is close to 78m × 6m × 3.3m, and the width of the main line is close to 1.5m.

[0058] This invention uses a small desktop or mobile small frame as the overall load-bearing structure, combined with a streamlined conveying layout of stepped feeding and multi-level speed-adjustable sorting. By compressing the equipment depth through a ring-shaped integrated sorting and collection architecture, it abandons the large fixed frame and truck-mounted integrated design of the existing technology, optimizing the overall size (length × width × height) to 10m × 3m × 1.5m, with a single set cost of only 8,000 yuan, less than one-tenth of the existing fixed frame equipment and one-twentieth of the truck-mounted integrated equipment.

[0059] Compared to existing mainstream large-scale sorting equipment, the overall size of the machine of this invention is greatly reduced, which not only significantly reduces the manufacturing cost of the equipment, but also has a strong spatial adaptability. It can be flexibly placed in small spaces such as fields and small warehouses, perfectly meeting the small-scale and decentralized sorting needs of small and medium-sized farmers, and effectively reducing the threshold for using citrus sorting equipment.

[0060] ② This invention employs an image detection mechanism with dual cameras installed facing each other to simultaneously acquire images of the citrus surface. It can strictly follow the NY / T1190-2006 industry standard to quantitatively detect defect indicators such as the number and diameter of spots on the fruit surface. At the same time, the weighing module is integrated into the bearing tray of each weighing and unloading unit arranged in a ring. As the citrus flows along the ring path, the weight acquisition and appearance defect detection are completed simultaneously. Combined with the fuzzy comprehensive evaluation algorithm, the multi-dimensional quality of weight and surface defects is judged simultaneously, which significantly improves the grading accuracy and standardization of citrus sorting.

[0061] ③ This invention adopts an integrated sorting and collection module, which integrates the three core processes of weight detection, image detection, and fixed-point unloading on the same circular material conveying path. As the fruit rotates and flows with the circular material conveying unit, weight information and fruit surface images are collected simultaneously. The control and monitoring module uses the unique ID of the weighing and unloading unit as an identifier, binds the weight and defect data of a single fruit, and completes the quality grading judgment. Then, it directly controls the corresponding weighing and unloading unit to complete the unloading above the designated sorting bin by magnetic attraction de-energization and self-weight deflection. This realizes the integrated and continuous execution of detection and sorting, which simplifies the overall structure of the equipment, shortens the operation process, and completely avoids the grading error problem caused by the misalignment of the detection and sorting action sequence, thus improving the operating efficiency and grading accuracy of the equipment.

[0062] ④ This invention incorporates a multi-variety citrus sorting parameter library, covering grading standards for various mainstream varieties such as loose-skinned mandarins, pomelos, and oranges. Each variety's grading standard is accompanied by corresponding fuzzy comprehensive evaluation algorithm parameters (including membership function parameters, indicator evaluation weight vectors, and output evaluation sets). These fuzzy comprehensive evaluation algorithm parameters are based on manually corrected grading labels and undergo adaptive parameter iterative fine-tuning through a self-learning module, continuously reducing the deviation between the system's grading results and manual judgment standards.

[0063] Meanwhile, this invention features a one-click switching function for grading standards of corresponding varieties on the host computer. The grower can directly and quickly retrieve the corresponding grading standards and algorithm parameters on the host computer according to the actual variety of citrus to be sorted. There is no need to replace the equipment hardware components or manually modify complex parameters, which can quickly complete the sorting of different varieties of citrus. This fully adapts to the needs of small-batch, multi-batch sorting operations in multi-variety mixed-crop production areas, greatly improving the versatility and applicability of the equipment.

[0064] ⑤ This invention optimizes the low-damage design of the entire process from conveying, carrying, and unloading. In the conveying stage, a conveyor belt with a food-grade silicone flexible buffer layer and anti-slip texture is used, combined with multi-level speed control and an arc-shaped tail end guide structure to achieve orderly and flexible conveying of individual citrus fruits, buffering the impact of fruit collisions and avoiding skin damage caused by rigid contact and mutual squeezing. In the carrying stage, an arc-shaped concave carrying plate is used to adapt to the shape of the fruit, and in the unloading stage, a self-weight slow tilting unloading method is used to reduce rigid impact throughout the process and fully ensure the commercial integrity of the sorted fruits.

[0065] On the other hand, a jamming alarm module linked to the controller is set up to monitor the entire process of equipment operation in real time. Once a jamming abnormality is detected, a signal is immediately fed back to the controller, which simultaneously triggers an audible and visual alarm and controls the whole machine to perform a shutdown protection action. This not only fully ensures the integrity of the sorted fruit, but also greatly improves the reliability and stability of the equipment in long-term operation, and reduces equipment maintenance costs and downtime due to failure. Attached Figure Description

[0066] Figure 1 This is a schematic diagram of the structure of a small intelligent rapid sorting system in an embodiment of the present invention;

[0067] Figure 2 This is a partial structural diagram of the integrated sorting and collection module in an embodiment of the present invention;

[0068] Figure 3 This is a top view of the integrated sorting and collection module in an embodiment of the present invention;

[0069] Figure 4 This is a schematic diagram of the weighing and unloading unit in an embodiment of the present invention. Figure 4 (a) shows the loading state of the weighing and unloading unit. Figure 4 (b) shows the unloading status of the weighing and unloading unit;

[0070] Figure 5 This is a flowchart of the operation of a small intelligent rapid sorting system in an embodiment of the present invention. Detailed Implementation

[0071] like Figures 1 to 5 As shown, a small-scale intelligent rapid sorting system for multiple varieties of citrus includes a machine frame, a drive assembly, a human-machine interaction module, a feeding and impurity removal module, an integrated sorting and collection module, and a control and monitoring module.

[0072] The machine frame is a small desktop or movable small frame, and the overall dimensions (length × width × height) are 10m × 3m × 1.5m;

[0073] The human-machine interaction module, as the host computer 1, is used to provide a selection interface for sorting standards of multiple varieties of citrus and a manual correction operation interface, and to display the overall machine operation, sorting output, and equipment fault information. At the same time, it transmits the selected citrus sorting standards and manual correction information to the control and monitoring module.

[0074] The specific content of the multi-variety citrus sorting standards is stored in the control and supervision module. It mainly includes parameters such as the membership function parameters, indicator evaluation weight vectors, and output judgment sets of different citrus varieties such as Wogan and Sugar Mandarin. The human-computer interaction module only provides a selection interface. After the operator selects the citrus varieties to be sorted on the multi-variety citrus sorting standard selection interface, the selection result will be transmitted to the control and supervision module. The control and supervision module will then retrieve the corresponding citrus sorting standards and use them as a reference for citrus grading in the subsequent citrus sorting process.

[0075] In this embodiment, the human-computer interaction module is configured with an independent manual correction operation interface, and the operator can perform two types of correction operations:

[0076] (1) Correction of single fruit grading results: retrieve the weight, surface defect index and grading results of sorted citrus, manually modify the quality grade labels of the automatic grading results that have deviations, and upload the modified quality grade labels to the control and supervision module as input data for the self-learning optimization module. The quality grade labels include special grade, first grade, second grade and third grade.

[0077] (2) Correction of sorting standards for multiple varieties of citrus: Select the citrus varieties to be modified, such as Wogan and Sugar Mandarin, and manually modify the fuzzy membership boundary parameters, weights of each quality indicator and thresholds of each quality grade for the citrus variety. After the modification is completed, the updated classification standards for multiple varieties of citrus will be sent to the control and supervision module as input data for the self-learning optimization module.

[0078] Both types of manually corrected information are transmitted in real time to the self-learning optimization unit of the control and supervision module as true value samples for gradient descent iterative optimization, used to iteratively update indicator weights, and reduce the error between the system's automatic classification judgment and the manual judgment results.

[0079] The feeding and impurity removal module is used to automatically transport citrus fruits. It uses multiple conveyor belts that increase speed in sequence to create speed differences to separate dense citrus fruits. Combined with the arc-shaped closing structure at the tail and multiple baffles, it enables the orderly dropping of individual citrus fruits. The fan speed is dynamically matched according to the diameter of the citrus fruits to be sorted, and the fan blows in a directional manner to remove impurities such as fruit peel residue, fruit scraps, and field debris.

[0080] In this embodiment, the feeding and impurity removal module includes an automatic feeding unit, a multi-level speed-regulating sorting unit, and an adjustable airflow impurity removal unit, wherein...

[0081] The automatic feeding unit includes a hopper and a stepped conveyor belt 9. After the citrus fruits are fed into the hopper, the stepped conveyor belt 9 smoothly transports the fruits from a low position to a high feeding station 2 (also called a fruit feeding device) and automatically tilts the citrus fruits. The surface of the stepped conveyor belt is equipped with stepped retaining structures to support the citrus fruits and prevent them from rolling off the conveyor. Figure 1 As shown.

[0082] The multi-stage speed-regulating fruit sorting unit is used to generate a speed difference through multiple sections of citrus conveyor belt 3 that increase speed in sequence, thereby breaking up the stacked citrus and increasing the distance between the fruits. The citrus conveyor belt 3 includes an infeed conveyor belt 301, a first-stage speed-regulating belt 302, and a second-stage speed-regulating belt 303. At the same time, a pair of limiting baffles set on both sides of the conveyor belt and facing each other with the spacing gradually narrowing along the conveying direction are used to gather and limit the citrus. After being gathered and arranged by the arc-shaped closing structure 4 at the end of the conveyor belt, the citrus is discharged in an orderly manner with the help of the liftable baffles.

[0083] In this embodiment, the liftable baffle is a vertically liftable opening and closing baffle, set at the outlet of the arc-shaped converging structure 4. The baffle is synchronously linked with the rotation position of the weighing and unloading unit. When the first weighing and unloading unit rotates to the fruit receiving position directly below the arc-shaped converging structure, the control mechanism drives the baffle to descend and open, and the first citrus fruit falls naturally into the loading unit by gravity. After the fruit has completely fallen into the loading unit, the liftable baffle immediately rises and closes to prevent subsequent fruits from falling. When the annular loading and conveying unit drives the next weighing and unloading unit to the designated fruit receiving position, the liftable baffle descends and opens again to release the next citrus fruit. This cycle repeats to achieve a precise one-to-one correspondence between the fruit and the loading unit.

[0084] The lifting baffle is controlled by the sorting instruction generation unit in the control and monitoring module. The lifting of the baffle is driven by a micro servo motor. The control and monitoring module determines the receiving position based on the rotary encoder signal of the weighing and unloading unit and controls the lifting of the baffle at the corresponding time.

[0085] In this embodiment, the arc-shaped closing structure 4 is located at the discharge end of the citrus conveyor belt 3, and is used to orderly limit and gather the citrus fruits conveyed by the citrus conveyor belt 3, ensuring that the citrus fruits fall steadily and individually into the weighing pan below. The surface of the citrus conveyor belt 3 is also covered with a flexible buffer pad layer, which is made of food-grade silicone and has an anti-slip texture to prevent fruit collision.

[0086] In other words, this embodiment uses a three-stage progressively faster citrus conveyor belt with an arc-shaped tail end structure. By utilizing the speed difference between the conveyor belts, dense citrus fruits are automatically separated and the fruit spacing is increased. The arc-shaped tail end then completes the limiting and gathering. Without the need for an additional independent fruit sorting mechanism, the orderly and stable transport of individual citrus fruits can be achieved. This simplifies the overall structure of the equipment, reduces manufacturing costs, and adapts to the space layout requirements of miniaturized equipment.

[0087] The adjustable wind power removal unit (i.e. Figure 1 The blower 10 is installed on one side of the multi-stage speed-adjustable fruit sorting unit. It is used to match the corresponding blowing force and wind speed parameters according to the current fruit diameter of the sorted citrus. Small-sized fruits adopt the low-pressure micro-wind blowing mode, while large-sized fruits adopt the standard wind blowing mode. It removes residual leaves, fruit chips and field debris attached to the surface of citrus fruits through directional airflow, taking into account the impurity removal effect and conveying stability of multiple varieties of fruits.

[0088] In this embodiment, after the operator selects a citrus variety, the control and monitoring module can automatically retrieve the standard fruit diameter range and preset wind speed for that variety. It can also identify the fruit diameter in real time through the image detection unit and dynamically fine-tune the fan speed.

[0089] The integrated sorting and collection module is used to carry individual citrus fruits through multiple independent weighing and unloading units 5 arranged circumferentially by the annular material conveying unit 6, which circulate and transfer them along a closed annular path. During the annular transfer, the weighing and unloading units 5 obtain the weight of the citrus fruits, and the image detection unit 7 obtains the surface defect indicators of the citrus fruits. The weight of the citrus fruits, the surface defect indicators, and the ID of the weighing and unloading unit 5 are sent to the control and monitoring module to obtain sorting and unloading instructions. When the annular material conveying unit 6 transports the citrus fruits to the sorting bin 8 specified by the sorting and unloading instructions, the weighing and unloading units 5 dump the citrus fruits to complete the unloading. The sorting bin 8 collects the citrus fruits according to their quality grade. The surface defect indicators include the number of spots and the diameter of the spots.

[0090] In this embodiment, the integrated sorting and collection module includes a weighing and unloading unit 5, a circular material conveying unit 6, and an image detection unit 7, wherein...

[0091] The weighing and unloading unit 5 is evenly arranged around the circumference of the annular material conveying unit. It integrates the dual functions of weight acquisition and fixed-point unloading. It is used to collect the weight data of a single fruit and transmit the weight data and the ID of the current weighing and unloading unit to the control and monitoring module. At the same time, it receives the sorting and unloading instructions issued by the control and monitoring module. When the weighing and unloading unit rotates to the top of the designated sorting bin 8, it pours the citrus to complete the unloading and automatically clears the weight data of the citrus. It then returns to the feeding end of the annular material conveying unit for cyclic operation.

[0092] The weighing and unloading unit 5 includes a concave bearing plate 501, a hinged support 502, a swing bracket 503, and a support column 504 fixedly mounted on the annular material conveying unit 6. The hinged support 502 is located at an eccentric position at the bottom of the concave bearing plate 501. The hinged support 502 is hinged to one end of the swing bracket 503. The rod of the swing bracket 503 near the other end is hinged to the support column 504 via a pin.

[0093] A first magnetic attraction structure 505 is provided at the bottom of the concave bearing plate 501 on the side away from the hinge support 502, and a second magnetic attraction structure 506 is provided at the top of the support column 504. The first magnetic attraction structure 505 and the second magnetic attraction structure 506 cooperate to lock the attitude of the concave bearing plate 501. A third magnetic attraction structure 507 is provided at the other end of the swing bracket 503. The third magnetic attraction structure 507 cooperates with a fourth magnetic attraction structure 508 fixedly installed on the annular material conveying unit 6 to lock the attitude of the swing bracket 503.

[0094] A return spring 509 is provided on the swing bracket 503 in the area between the two hinges. One end of the return spring 509 is fixed on the swing bracket 503 and the other end is fixed on the annular material conveying unit 6. It is used to drive the swing bracket 503 to return to its original position after the oranges are unloaded from the concave bearing plate 501.

[0095] The weighing module 510 is located at the bottom of the inner cavity of the concave bearing plate 501. It is used to collect the weight information of a single citrus fruit carried in the concave bearing plate 501 when the concave bearing plate 501 is in a horizontal bearing posture and the weighing and unloading unit 5 moves to the preset weighing position with the annular material conveying unit 6.

[0096] When the citrus fruits fall from the arc-shaped concave structure 4 onto the concave support plate 501, the first magnetic attraction structure 505, the second magnetic attraction structure 506, the third magnetic attraction structure 507, and the fourth magnetic attraction structure 508 are respectively energized and engaged. At this time, the concave support plate 501 is locked to the support column 504, and the swing bracket 503 is locked to the fourth magnetic attraction structure 508, ensuring that the citrus fruits will not tip over during transport. In this state, the return spring 509 is in a relaxed state (e.g., Figure 4 (as shown in (a))

[0097] When the citrus fruits arrive at the target sorting bin 8 along with the annular conveyor unit 6, the first magnetic attraction structure 505 and the second magnetic attraction structure 506, the third magnetic attraction structure 507 and the fourth magnetic attraction structure 508 are simultaneously de-energized, releasing the attraction lock. Because the hinge fulcrum between the swing bracket 503 and the support column 504 is offset and not located at the center of the swing bracket 503, under the weight torque of the concave bearing plate 501 and the citrus fruits inside, the swing bracket 503 deflects downwards around the pin at the support column 504. Simultaneously, during the unloading and falling of the citrus fruits, the hinge support 502 can adaptively rotate relative to the pin at the end of the swing bracket 503 to match the unloading angle, facilitating the smooth unloading of the citrus fruits from the concave bearing plate 501. In this state, the return spring 509 is compressed and stored energy by the swing bracket 503 (e.g., Figure 4 (b) shown).

[0098] After the citrus fruits are completely poured out, the return spring 509 releases its elastic potential energy, driving the swing bracket 503 to rotate in the opposite direction around the pin. The hinge support 502 of the concave bearing plate 501 is eccentrically located at its bottom, so that the center of gravity of the concave bearing plate 501 when unloaded is always located on the side of the hinge point closer to the second magnetic structure 506. Therefore, after the citrus fruits are unloaded, the concave bearing plate 501 rotates around the hinge axis of the hinge support 502 on the side of the second magnetic structure 506 under the action of the restoring torque generated by its own gravity. At the same time, the rotation angle is limited within a preset range under the constraint of the limiting structure, ensuring that the alignment end faces of the first magnetic structure 505 and the second magnetic structure 506 are accurately attached. After the alignment end faces of the two sets of magnetic structures are attached, the control and monitoring module controls the first magnetic structure 505 and the second magnetic structure 506, and the third magnetic structure 507 and the fourth magnetic structure 508 to be simultaneously energized and locked, so that the concave bearing plate 501 returns to the horizontal bearing posture.

[0099] During the rotation and reset of the swing bracket 503, because the hinge support 502 is located at the eccentric position at the bottom of the concave bearing plate 501, the center of gravity of the concave bearing plate 501 always falls on the side of the hinge point closer to the second magnetic attraction structure 506. This causes the hinge support 502 at the bottom of the concave bearing plate 501 to rotate in the opposite direction relative to the end pin of the swing bracket 503, so that the first magnetic attraction structure 505 and the second magnetic attraction structure 506 are aligned and attracted, and the two remain in an electrically locked state, thus completing the posture reset of the concave bearing plate 501.

[0100] A limiting structure is provided at the hinge position between the hinge support 502 and the swing bracket 503 to restrict the hinge support 502 from rotating at a small angle relative to the swing bracket 503. This ensures that the center of gravity of the concave bearing plate 501 remains on the side of the hinge point close to the second magnetic structure 506, and that the concave bearing plate 501 can return to its original position smoothly by its own weight after unloading, so as to achieve reliable alignment and attraction between the first magnetic structure 505 and the second magnetic structure 506.

[0101] The first magnetic attraction structure 505, the second magnetic attraction structure 506, the third magnetic attraction structure 507 and the fourth magnetic attraction structure 508 are all energized and attracting type DC electromagnets with a rated voltage of DC 24V. The power supply control lines of each magnetic attraction structure are collected through the wiring groove inside the annular material conveying unit 6 and then electrically connected to the control and monitoring module through the slip ring structure. The control and monitoring module uniformly controls the energization and de-energization of each magnetic attraction structure.

[0102] In this embodiment, the concave support plate 501 is covered with a material with a high coefficient of friction, and the curved surface conforms to the contour of the fruit. This concave support plate 501 effectively secures the fruit through high friction and its curved shape, preventing the fruit from sliding, rolling, or falling during rotation or tilting. The concave support plate 501 can be made of a transparent material to avoid obscuring the bottom surface of the citrus fruit during image recognition.

[0103] The sorting bins 8 include a first sorting bin 801, a second sorting bin 802, a third sorting bin 803, and a fourth sorting bin 804. The four sorting bins are arranged circumferentially along the outer periphery of the annular material conveying unit, and correspond to the special grade, first grade, second grade, and third grade quality grades, respectively. The weighing and unloading unit pours the citrus fruits outward from the annular ring, and they fall into the corresponding sorting bins.

[0104] The fourth magnetic structure 508 is mounted on a magnetic support 601 on the annular material conveying unit 6.

[0105] The annular material conveying unit 6 is a circulating transfer carrier of the integrated sorting and collection module. It rotates at a constant speed according to a circular trajectory and is used to carry multiple weighing and unloading units to continuously circulate and transport single citrus fruits along a closed annular path.

[0106] In this embodiment, the ring-shaped material conveying unit adopts an integrated ring layout to replace the traditional equipment weighing station and image station separate arrangement scheme, which can eliminate the data misalignment and cycle imbalance caused by fruit transfer across stations. At the same time, the supporting drive components realize stepless speed regulation, adapt to the spatial layout of miniaturized equipment, and ensure that the citrus fruits are smoothly and continuously transferred to the image detection area and each graded unloading point.

[0107] The image detection unit 7 includes a light box, a light source, dual cameras, and an image processor. The light source is installed on the inner wall of the light box to provide a uniform, shadow-free supplementary lighting environment for the citrus fruits on the weighing and unloading unit, eliminating surface reflections and blind spots in the shooting. The dual cameras are installed facing each other inside the light box to capture the front and back of the citrus fruits respectively, ensuring that the entire citrus fruit is captured. The image processor is electrically connected to the dual cameras. After receiving the front and back images of the citrus fruits captured by the dual cameras, it performs image processing on the front and back images to obtain surface defect information of the citrus fruits and outputs it to the controller.

[0108] In this embodiment, the dual cameras are positioned inside the lightbox to capture the full view of the citrus fruit. For example, the dual cameras can be arranged vertically facing each other, with the top camera shooting vertically downwards at the upper surface of the citrus fruit and the bottom camera shooting vertically upwards through the transparent support plate at the lower surface of the citrus fruit, thus covering the main fruit surface area.

[0109] The image detection unit 7 is located above the middle section of the annular material conveying unit. This module has a through-hole for the annular material conveying unit 6 to drive the weighing and unloading unit 5 through one side of the through-hole and out the other side of the through-hole. During this process, it acquires images of the front and back of the citrus fruit (using a uniform light source from a lightbox to eliminate reflections and shadows on the peel, and using two cameras facing each other to simultaneously acquire complete images of the front and back of the citrus fruit). Combined with the built-in image processor, it obtains surface defect indicators such as the number and diameter of spots on the citrus fruit and transmits the surface defect indicators to the control and monitoring module.

[0110] In this embodiment, the image detection unit 7 is positioned above the middle section of the circular conveyor unit 6. Utilizing the circular co-station design, the citrus fruits can be simultaneously imaged and weighed without secondary repositioning, thus avoiding the defects of mismatched multi-source detection data from the source. Furthermore, the multiple weighing and unloading units 5 mounted on the circular conveyor unit 6 can collect the weight of the citrus fruits during the conveying process. This embodiment, by integrating weight acquisition and panoramic appearance detection into the same rotating conveyor station, can reduce the overall machine size and achieve miniaturization while avoiding the problems of mismatched single-fruit data and inconsistent conveying rhythm caused by multi-station transfer.

[0111] The control and monitoring module, which is the controller, is the core control unit of the small intelligent rapid sorting system. It is used to calculate the quality grade of citrus fruits based on their weight and surface defect indicators using a fuzzy comprehensive evaluation algorithm. Combined with the citrus sorting standards issued by the human-machine interaction module, it generates sorting and unloading instructions. At the same time, it monitors the material jamming faults of the feeding and impurity removal module and the integrated sorting and collection module in real time. After detecting an abnormality, it triggers an audible and visual alarm and controls the entire machine to stop for protection. It coordinates the entire process of the human-machine interaction module, the feeding and impurity removal module, and the integrated sorting and collection module to work together.

[0112] In this embodiment, the control and monitoring module integrates weighing detection, image detection, sorting control, jamming alarm functions, and self-learning functions. It can receive multi-variety classification standards set by the host computer, process the weight and surface defect indicators transmitted by the weighing module and the image detection mechanism in real time, complete the comprehensive judgment of citrus grade, and accurately control the integrated rotary conveyor mechanism to perform corresponding sorting and discharging actions based on the judgment results. At the same time, it can also receive abnormal signals from the jamming alarm module, trigger alarm prompts, and control the equipment to stop for protection. It coordinates the collaborative operation of various mechanisms to ensure the automated, stable, and orderly execution of the sorting process. Furthermore, the controller relies on the self-learning module to iteratively optimize the core parameters of the multi-variety classification standards, continuously reduce the grading judgment error, and improve the long-term sorting accuracy of the system.

[0113] In this embodiment, the control and monitoring module includes a sorting instruction generation unit, a fault alarm protection unit, and a self-learning optimization unit, wherein...

[0114] The sorting instruction generation unit is used to receive the citrus sorting standards issued by the human-computer interaction module, calculate the comprehensive quality grade of the citrus based on the weight and surface defect index of the citrus using a fuzzy comprehensive evaluation algorithm, and generate sorting and unloading instructions in combination with the citrus sorting standards.

[0115] The fault alarm protection unit is used to monitor the material jamming fault of the feeding and impurity removal module and the integrated sorting and collection module in real time. After detecting the material jamming abnormality, it triggers an audible and visual alarm and controls the whole machine to stop for protection.

[0116] In this embodiment, the detection end of the fault alarm protection unit is located at the connection station between the citrus conveyor belt 3 and the weighing and unloading unit 5. The signal output end of the jamming alarm module is electrically connected to the controller. The controller is bidirectionally connected to the host computer 1. The self-learning module receives modification information from the host computer 1 and modifies the multi-variety classification standard built into the controller.

[0117] The self-learning optimization unit receives manually corrected information uploaded by the human-computer interaction module as input. It incorporates a small-sample gradient descent fine-tuning strategy suitable for embedded controllers (low computational cost, can be directly deployed in the device controller). With the optimization objective of minimizing the error between the system's automatic grading prediction results and the manually corrected information, it iteratively updates the weight parameters corresponding to each quality indicator in the multi-variety citrus classification standard. After the iterative optimization is completed, the updated set of indicator weights is synchronously stored in the multi-variety citrus classification standard database of the sorting instruction generation unit for subsequent citrus grading judgment, continuously reducing the system's grading misjudgment rate.

[0118] The drive assembly includes drive motors that are respectively connected to the fruit feeding device 2, the citrus conveyor belt 3, and the weighing and unloading unit 5. Each drive motor is electrically connected to the control and monitoring module to provide power for the conveying and sorting of citrus fruits.

[0119] The discharge end of the feeding and impurity removal module is connected to the feed end of the integrated sorting and collection module;

[0120] The data output terminal of the human-computer interaction module is connected to the first data input terminal of the control and supervision module, the data input terminal of the human-computer interaction module is connected to the first data output terminal of the control and supervision module, the data output terminal of the integrated sorting and collection module is connected to the second data input terminal of the control and supervision module, and the data input terminal of the integrated sorting and collection module is connected to the second data output terminal of the control and supervision module.

[0121] The human-machine interface module is fixedly mounted on the side frame of the feeding and impurity removal module. Both the feeding and impurity removal module and the integrated sorting and collection module are fixedly installed on the overall machine frame. The control and monitoring module's hardware is primarily a controller, integrated within the overall machine frame at the connection area between the feeding and impurity removal module and the integrated sorting and collection module. The discharge end of the feeding and impurity removal module connects to the feed end of the integrated sorting and collection module. Both the feeding and impurity removal module and the integrated sorting and collection module are fixedly installed on the overall machine frame. The control and monitoring module's hardware is primarily a controller, integrated within the overall machine frame.

[0122] The method for rapid sorting of multiple varieties of citrus using the above-mentioned small intelligent rapid sorting system includes the following steps:

[0123] 1) Initialize the small intelligent rapid sorting system, select the citrus sorting criteria in the human-computer interaction module, and transmit the selected citrus sorting criteria to the control and monitoring module;

[0124] 2) The citrus fruits to be sorted are automatically fed by the feeding and impurity removal module. After being gradually accelerated and dispersed into individual fruits, the individual citrus fruits are output to the feeding end of the integrated sorting and collection module in sequence.

[0125] 3) The ring-shaped material conveying unit 6 of the integrated sorting and collection module continuously conveys along the closed ring path. When the weighing and unloading unit 5 carrying citrus fruits moves to the preset weighing position and the concave bearing plate 501 is in a horizontal and stable bearing posture, the weighing module 510 triggers weight acquisition to obtain the weight information of a single citrus fruit and sends the weight information and the weighing and unloading unit ID to the control and monitoring module.

[0126] 4) The control and monitoring module matches, binds, and stores the weight and surface defect indicators of the citrus fruits to be sorted with the ID of the weighing and unloading unit. It then uses a fuzzy comprehensive evaluation algorithm to calculate the overall quality grade of the citrus fruits to be sorted. Combined with the citrus sorting standards, it generates sorting and unloading instructions for the citrus fruits to be sorted and transmits these instructions to the integrated sorting and collection module. The specific process is as follows:

[0127] 4-1) Each sorting bin is pre-assigned a unique number. Each weighing and unloading unit is assigned a unique identification ID. ;

[0128] 4-2) The citrus fruits to be sorted fall above the weighing and unloading unit 5, and the weighing and unloading unit 5 automatically collects the weight information of the citrus fruits to be sorted, and transmits it together with the ID of the weighing and unloading unit 5 to the control and monitoring module.

[0129] 4-3) When the citrus fruits to be sorted enter the image acquisition range of the image detection unit 7, the image detection unit 7 acquires front and back images of the citrus fruits to be sorted, and then uses image processing or deep learning algorithms to obtain surface defect indicators of the citrus fruits to be sorted, and generates category labels based on the surface defect indicators. Send the category label to the control and supervision module;

[0130] Considering the limited bus transmission bandwidth and insufficient computing power of the embedded controller under high-speed continuous flow conditions of the annular material loading unit, in this embodiment, after the image detection unit 7 acquires surface defect indicators such as the number of spots and the diameter of spots, it generates category labels based on the surface defect indicators. Sending category labels to the control and monitoring module serves two purposes: firstly, to reduce the amount of data transmitted and avoid data transmission delays and time sequence errors between single fruit IDs and defect information during high-speed sorting; secondly, to delegate the pre-processing of defect quantification and classification to the image processor, thereby reducing the real-time computational load on the control and monitoring module and ensuring rapid response in fuzzy classification judgment. At the same time, standardized category labels facilitate the system in completing single fruit data binding, manual classification correction, and self-learning weight iteration.

[0131] 4-4) In the sorting instruction generation unit of the control and supervision module, first start from the category label The surface defect information is obtained by parsing, and then the received weight and surface defect information are matched. The weight information and surface defect information of the same citrus to be sorted are bound and stored with the ID of the weighing and unloading unit carrying the citrus to be sorted.

[0132] 4-5) In the sorting instruction generation unit of the control and supervision module, the fuzzy comprehensive evaluation algorithm is used to calculate the comprehensive grading judgment value of the citrus to be sorted based on the weight and surface defect index of the citrus to be sorted. The specific process is as follows:

[0133] It should be noted that conventional grading methods often use fixed-weight linear weighting. For example, the controller can retrieve the pre-selected citrus sorting criteria from the host computer 1 and, based on the comprehensive grading judgment formula shown below, quantify the single-fruit detection data after binding and matching to obtain the comprehensive grading judgment value of the citrus to be sorted:

[0134]

[0135] In the formula, This is a comprehensive grading determination value. The weighting coefficient is the weight index. This refers to the weighting coefficient for appearance defect indicators. This is the normalized value of the weight data collected by the corresponding weighing module. This is the normalized value corresponding to the defect category label;

[0136] However, this method relies solely on fixed weight coefficients to perform linear weighting calculations, which cannot adapt to the coupling relationship characteristics of weight and appearance defect indicators of different citrus varieties. The grading criteria are too rigid, which can easily lead to grading misjudgments and poor adaptability.

[0137] Therefore, this embodiment does not adopt the conventional method, but instead combines a fuzzy comprehensive evaluation algorithm. It uses citrus weight and surface defects as input factors for the fuzzy comprehensive evaluation, dynamically allocating indicator weights and constructing multi-dimensional evaluation intervals through fuzzification, fuzzy rule reasoning, and defuzzification processes. This approach better meets the actual sorting needs of different citrus varieties, significantly improving the accuracy and versatility of the grading results, as detailed below:

[0138] 4-5-1) The weight and surface defect index of the citrus to be sorted are used as quality evaluation indicators. The fuzzy membership function is used to map each quality evaluation indicator to a continuous membership value in the interval of 0 to 1, which respectively characterizes the degree of conformity of the citrus to be sorted as special grade, first grade, second grade and third grade. Among them, the weight adopts the trapezoidal fuzzy membership function, and the surface defect index specifically includes the number of spots and the diameter of spots. The number of spots and the diameter of spots both adopt the Z-type fuzzy membership function.

[0139] The weight index is modeled using a trapezoidal fuzzy function, and its expression is:

[0140]

[0141] In the formula, a, b, c, and d are the fuzzy boundary parameters of the index at the corresponding level.

[0142] The diameter and number of spots are modeled using a Z-type fuzzy function. The larger the defect, the lower the membership degree. The expression is:

[0143]

[0144] In the formula, a and b are the fuzzy boundary parameters of the index at the corresponding level.

[0145] 4-5-2) The values ​​of each quality indicator of the citrus to be sorted are fuzzified, and the membership degree of each quality indicator under the special grade, first grade, second grade and third grade is calculated respectively.

[0146] In this embodiment, the values ​​of various indicators of the citrus fruits to be graded are fuzzified, and their individual membership degrees at each grade are calculated. Specifically, let u be the membership degree of the i-th indicator at the k-th grade. ik This yields a membership matrix with multiple indicators and multiple levels:

[0147]

[0148] In the formula, n is the number of indicators, and the grade set can be represented as {Special Grade, Grade 1, Grade 2, Grade 3}.

[0149] 4-5-3) Retrieve the weights of each quality evaluation indicator in the citrus sorting criteria. Sub-indicators such as the number of spots and spot diameter each have their own weights, and the sum of all indicator weights equals 1. Calculate the individual membership degrees of each dimension in the membership matrix using a weighted summation method to obtain the comprehensive membership degrees corresponding to the special grade, first grade, second grade, and third grade, respectively. Specifically, let the weight corresponding to the i-th indicator be... And satisfy =1, then the comprehensive membership degree of a certain level k is... It can be represented as:

[0150]

[0151] In the formula, w represents the overall membership degree of a certain level k. i Let i be the weight of the i-th quality evaluation index. Let i be the membership degree of the i-th indicator at the k-th level. The total number of quality evaluation indicators;

[0152] 4-5-4) Compare the comprehensive membership degrees of the premium, first-grade, second-grade, and third-grade citrus fruits, and take the comprehensive membership degree with the highest value as the comprehensive grading judgment value for the citrus fruits to be sorted:

[0153]

[0154] In the formula, This is the comprehensive grading value for the citrus fruits to be sorted. The overall membership degree of a certain level k. It is a function that maximizes the independent variable.

[0155] 4-6) In the sorting instruction generation unit of the control and supervision module, the calculated comprehensive grading judgment value is compared with the preset threshold range of each quality grade in the citrus sorting standard to determine the final quality grade of the citrus to be sorted.

[0156] 4-7) In the sorting instruction generation unit of the control and supervision module, based on the pre-established mapping relationship between quality grade and sorting bin number, the final quality grade is matched to the corresponding target sorting bin number to generate the sorting instruction for the citrus to be sorted.

[0157] Furthermore, during the comprehensive judgment process in step 4), when the highest comprehensive membership degree of the citrus to be sorted is lower than the preset confidence threshold (e.g., 0.5), or the difference between the highest membership degree and the second highest membership degree is less than the preset difference threshold (e.g., 0.1), the citrus is determined to be a problematic fruit. The system does not generate an automatic sorting instruction, but instead transports the citrus to a separate problematic fruit collection bin for manual re-inspection and grading by operators. The manual grading results of the problematic fruit are also used as input samples for the self-learning optimization unit.

[0158] 5) According to the sorting and unloading command, the annular conveying unit 6 of the integrated sorting and collection module continues to rotate. When the weighing and unloading unit 5 carrying the citrus to be sorted rotates above the sorting bin 8 specified by the sorting and unloading command, the weighing and unloading unit performs the unloading operation to complete the grading and collection of the citrus. The specific process includes:

[0159] 5-1) The ring-shaped material conveying unit 6 continues to rotate circumferentially. The control and monitoring module tracks the circumferential position of each weighing and unloading unit 5 in real time. When the weighing and unloading unit 5 carrying the citrus to be sorted rotates to the preset tilting trigger angle directly above the sorting bin 8 specified by the sorting and unloading command, the control and monitoring module sends a power-off unloading command to the two sets of electromagnetic attraction parts corresponding to the weighing and unloading unit 5 simultaneously.

[0160] 5-2) When the two sets of electromagnetic attraction components are de-energized simultaneously, the attraction lock is released. Under the action of the self-weight torque of the eccentrically arranged concave bearing plate 501 and the swing bracket 503, they deflect downward around the pin axis of the support column 504. During the deflection process, the concave bearing plate 501 adaptively adjusts the tilt angle through the bottom hinge support 502, and tilts the oranges carried inside into the corresponding sorting bin 8 below. At this time, the return spring 509 is compressed as the swing bracket 503 deflects, storing elastic potential energy.

[0161] 5-3) After the citrus fruits are completely dumped, the return spring 509 releases its elastic potential energy, driving the swing bracket 503 to rotate in the opposite direction around the pin shaft; the concave bearing plate 501 returns to the center synchronously under its own weight and the constraint of the limiting structure. After the alignment end faces of the two sets of electromagnetic attraction components are attached, the electromagnetic attraction components are simultaneously energized and locked, so that the concave bearing plate 501 returns to the horizontal bearing posture; at the same time, the control monitoring module clears the citrus weight, surface defect index and unloading command bound to the ID of the weighing and unloading unit 5, so that the weighing and unloading unit 5 returns to the waiting-to-load state;

[0162] 5-4) The weighing and unloading unit 5, after being reset, continues to circulate with the annular material conveying unit 6, returning to the receiving station at the feeding end to wait for the next citrus to be sorted, thereby realizing continuous grading and sorting operations for multiple varieties of citrus.

[0163] In this embodiment, in step 5), after the grading and collection of citrus fruits is completed, the control and monitoring module will clear all the detection data and sorting instructions of the corresponding weighing and unloading unit 5. At the same time, the weighing and unloading unit returns to the feeding and receiving position with the circular material conveying unit to receive the next citrus fruit to be sorted, thereby realizing the continuous grading and collection of citrus fruits.

[0164] In summary, this invention addresses the shortcomings of existing citrus sorting equipment, such as its large size, high cost, poor adaptability, and insufficient grading accuracy.

[0165] At the structural level, this invention reduces equipment footprint and hardware costs by miniaturizing and making the whole machine frame mobile, and adopts an integrated sorting and collection module to realize the integrated layout of detection and unloading stations. It is combined with a multi-level speed-adjustable citrus conveyor belt to ensure orderly feeding of individual fruits, and is supplemented by a jamming alarm module to improve the stability of equipment operation. The overall structure is compact and flexible in deployment, and can be adapted to various scenarios such as small production warehouses and orchard sites.

[0166] At the algorithm and control level, this invention applies the fuzzy comprehensive evaluation algorithm to the fusion and grading of multi-dimensional quality data such as weight and surface defects. It realizes the binding, matching and verification of multi-source detection data of single fruit through the unique ID of the loading weighing module, supports one-click switching of classification standards for multiple varieties to adapt to the sorting requirements of different citrus varieties, and iteratively optimizes the grading parameters through the small sample gradient descent strategy of the self-learning module suitable for embedded controllers, continuously improving the grading accuracy and batch adaptability.

[0167] Based on the combined improvements at the two levels mentioned above, this invention ultimately achieves low-cost, refined, multi-variety continuous and rapid sorting of citrus fruits. It effectively solves the practical problems of high investment threshold for sorting equipment, low efficiency of manual sorting, and inconsistent standards for small and medium-sized growers, effectively alleviating the pressure of labor costs in production. It also provides a feasible technical path for the standardization of quality grading, digitalization of production management and control, and high-quality development of the citrus industry, and has important application value in helping to revitalize rural characteristic industries.

[0168] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications made to the present invention by those skilled in the art without departing from the spirit of the present invention shall fall within the protection scope of the present invention.

Claims

1. A small, intelligent, rapid sorting system for multiple varieties of citrus, characterized in that, It includes the overall machine frame, human-machine interface module, feeding and impurity removal module, integrated sorting and collection module, and control and monitoring module. The human-machine interaction module is used to provide a selection interface for sorting standards of multiple varieties of citrus and a manual correction operation interface, and to display the overall machine operation, sorting output, and equipment fault information. At the same time, it transmits the selected citrus sorting standards and manual correction information to the control and monitoring module. The feeding and impurity removal module is used to automatically convey citrus fruits. It uses multiple conveyor belts that increase speed in sequence to create speed differences to separate dense citrus fruits. Combined with the arc-shaped closing structure at the tail and multiple baffles, it enables the orderly dropping of individual citrus fruits. The fan speed is dynamically matched according to the diameter of the citrus fruits to be sorted, and impurities are blown away in a directional manner. The integrated sorting and collection module is used to carry individual citrus fruits through multiple independent weighing and unloading units arranged circumferentially by the annular material conveying unit. These units carry the fruits in a closed loop and transfer them along the loop. During the loop transfer, the weighing and unloading units acquire the weight of the citrus fruits, and the image detection unit acquires the surface defect indicators of the citrus fruits. The weight, surface defect indicators, and the ID of the weighing and unloading unit are sent to the control and monitoring module to obtain sorting and unloading instructions. When the annular material conveying unit transports the citrus fruits to the sorting bin specified by the sorting and unloading instructions, the weighing and unloading units dump the citrus fruits to complete the unloading. The sorting bins collect the citrus fruits according to their quality grades. The surface defect indicators include the number of spots and the diameter of the spots. The control and monitoring module is used to calculate the quality grade of citrus fruits based on their weight and surface defect indicators using a fuzzy comprehensive evaluation algorithm. It also generates sorting and unloading instructions by combining the citrus sorting standards issued by the human-machine interaction module. Simultaneously, it monitors the material jamming faults of the feeding and impurity removal module and the integrated sorting and collection module in real time. Upon detecting an abnormality, it triggers an audible and visual alarm and controls the entire machine to shut down for protection. It coordinates the entire process of the human-machine interaction module, the feeding and impurity removal module, and the integrated sorting and collection module to work together.

2. The small intelligent rapid sorting system according to claim 1, characterized in that, The human-computer interaction module is equipped with an independent manual correction interface, allowing operators to perform two types of correction operations: (1) Correction of single fruit grading results: retrieve the weight, surface defect index and grading results of sorted citrus, manually modify the quality grade labels that have deviations in the automatic grading results, and upload the modified quality grade labels to the control and supervision module. The quality grade labels include special grade, first grade, second grade and third grade. (2) Correction of sorting standards for multiple varieties of citrus: Select the citrus variety to be modified, manually modify the fuzzy membership boundary parameters, weights of each quality indicator and thresholds of each quality grade for the citrus variety, and send the updated classification standards for multiple varieties of citrus to the control and supervision module after the modification is completed.

3. The small intelligent rapid sorting system according to claim 1, characterized in that, The integrated sorting and collection module includes a ring-shaped material conveying unit, an image detection unit, and a weighing and unloading unit, wherein... The circular material conveying unit is used to carry multiple weighing and unloading units to continuously and cyclically convey single citrus fruits along a closed circular path. The image detection unit is used to acquire front and back images of citrus fruits, and obtain surface defect indicators of citrus fruits by combining them with the built-in image processor, and transmit the surface defect indicators to the control and monitoring module. The weighing and unloading unit is used to collect the weight data of a single fruit and transmit the weight data and the ID of the current weighing and unloading unit to the control and monitoring module. At the same time, it receives the sorting and unloading instructions issued by the control and monitoring module. When the weighing and unloading unit rotates to the top of the designated sorting bin, it pours the citrus to complete the unloading and automatically clears the weight data of the citrus. The citrus then returns to the feeding end of the circular material conveyor unit for cyclic operation.

4. The small intelligent rapid sorting system according to claim 1, characterized in that, The feeding and impurity removal module includes an automatic feeding unit, a multi-stage speed-regulating processing unit, and an adjustable airflow impurity removal unit, wherein... The automatic feeding unit is used to transport citrus fruits from a low position to a feeding station via a stepped conveyor belt and automatically dump the citrus fruits. The multi-stage speed-regulating fruit sorting unit is used to generate a speed difference through multiple sections of conveyor belts that increase speed in sequence, thereby breaking up the stacked citrus fruits and increasing the spacing between the fruits. At the same time, a pair of limiting baffles set on both sides of the conveyor belt and facing each other with the spacing gradually narrowing along the conveying direction are used to gather and limit the citrus fruits. After being gathered and arranged by the arc-shaped closing structure at the end of the conveyor belt, the fruits are discharged in an orderly manner with the help of the liftable baffles. The adjustable wind power cleaning unit is used to match the corresponding blowing wind power according to the current fruit diameter of the sorted citrus, and remove residual leaves, fruit debris and field debris attached to the surface of the citrus through directional airflow.

5. The small intelligent rapid sorting system according to claim 1, characterized in that, The control and monitoring module includes a sorting instruction generation unit, a fault alarm protection unit, and a self-learning optimization unit, wherein... The sorting instruction generation unit is used to receive the citrus sorting standards issued by the human-computer interaction module, calculate the comprehensive quality grade of the citrus based on the weight and surface defect index of the citrus using a fuzzy comprehensive evaluation algorithm, and generate sorting and unloading instructions in combination with the citrus sorting standards. The fault alarm protection unit is used to monitor the material jamming fault of the feeding and impurity removal module and the integrated sorting and collection module in real time. After detecting the material jamming abnormality, it triggers an audible and visual alarm and controls the whole machine to stop for protection. The self-learning optimization unit receives manually corrected information uploaded by the human-computer interaction module as input. It incorporates a small-sample gradient descent fine-tuning strategy suitable for embedded controllers, aiming to minimize the error between the system's automatic grading prediction results and the manually corrected information. It iteratively updates the weight parameters corresponding to each quality indicator in the multi-variety citrus classification standard. After the iterative optimization is completed, the updated set of indicator weights is synchronously stored in the multi-variety citrus classification standard database of the sorting instruction generation unit for subsequent citrus grading judgment, continuously reducing the system's grading misjudgment rate.

6. A method for rapid sorting of multiple varieties of citrus using the small intelligent rapid sorting system described in claim 1, characterized in that, Includes the following steps: 1) Select the citrus sorting criteria in the human-computer interaction module and transmit the selected citrus sorting criteria to the control and monitoring module; 2) The citrus fruits to be sorted are automatically fed by the feeding and impurity removal module. After being gradually accelerated and dispersed into individual fruits, the individual citrus fruits are output to the feeding end of the integrated sorting and collection module in sequence. 3) The integrated sorting and collection module uses a ring-shaped material conveying unit to transport the citrus to be sorted along a ring path. During the transport process, the weighing and unloading unit collects the weight of the citrus to be sorted, and the image detection unit obtains the surface defect index of the citrus to be sorted. The weight of the citrus to be sorted, the surface defect index, and the ID of the weighing and unloading unit are transmitted to the control and monitoring module. 4) The control and monitoring module matches and binds the weight and surface defect index of the citrus to be sorted with the ID of the weighing and unloading unit. It uses a fuzzy comprehensive evaluation algorithm to calculate the comprehensive quality grade of the citrus to be sorted, and generates a sorting and unloading instruction for the citrus to be sorted in combination with the citrus sorting standard. The sorting and unloading instruction is then transmitted to the integrated sorting and collection module. 5) According to the sorting and unloading command, the ring-shaped material conveying unit of the integrated sorting and collection module continues to rotate. When the weighing and unloading unit carrying the citrus to be sorted rotates to the top of the sorting bin specified by the sorting and unloading command, the weighing and unloading unit performs the unloading operation to complete the grading and collection of citrus.

7. The method according to claim 6, characterized in that, In step 4), the sorting instruction is generated in the following manner: 4-1) Assign a unique number to each sorting bin and a unique identification ID to each weighing and unloading unit in advance; 4-2) The citrus fruits to be sorted fall above the weighing and unloading unit, which automatically collects the weight information of the citrus fruits to be sorted and transmits it to the control and monitoring module along with the ID of the weighing and unloading unit. 4-3) When the citrus fruits to be sorted enter the image acquisition range of the image detection unit, the image detection unit acquires the front and back images of the citrus fruits to be sorted, and then uses image processing or deep learning algorithms to obtain the surface defect indicators of the citrus fruits to be sorted and sends them to the control and monitoring module. 4-4) In the control and monitoring module, the received weight and surface defect information are matched, and the weight information and surface defect information of the same citrus to be sorted are bound and stored with the ID of the weighing and unloading unit carrying the citrus to be sorted; 4-5) Using the fuzzy comprehensive evaluation algorithm, calculate the comprehensive grading judgment value of the citrus to be sorted based on the weight and surface defect index of the citrus to be sorted; 4-6) Compare the calculated comprehensive grading judgment value with the preset threshold range of each quality grade in the citrus sorting standard to determine the final quality grade of the citrus to be sorted. 4-7) Based on the pre-established mapping relationship between quality grades and sorting bin numbers, the final quality grade is matched to the corresponding target sorting bin number, and sorting instructions for the citrus to be sorted are generated.

8. The method according to claim 7, characterized in that, In steps 4-5), the comprehensive grading value of the citrus fruits to be sorted is calculated in the following manner: 4-5-1) The weight and surface defect index of the citrus to be sorted are used as quality evaluation indicators. The fuzzy membership function is used to map each quality evaluation indicator to a continuous membership value in the range of 0 to 1, which respectively characterizes the degree of conformity of the citrus to be sorted as special grade, first grade, second grade and third grade. Among them, the weight adopts the trapezoidal fuzzy membership function, and the number of spots and the diameter of spots adopt the Z-type fuzzy membership function. 4-5-2) The values ​​of each quality indicator of the citrus to be sorted are fuzzified, and the membership degree of each quality indicator under the special grade, first grade, second grade and third grade is calculated respectively. 4-5-3) Retrieve the weights of each quality evaluation indicator in the citrus sorting standard. The sum of all indicator weights equals 1. Calculate the membership degree of each dimension in the membership degree matrix by weighted summation to obtain the comprehensive membership degree corresponding to each grade: special grade, first grade, second grade, and third grade. ； In the formula, w represents the overall membership degree of a certain level k. i Let i be the weight of the i-th quality evaluation index. Let i be the membership degree of the i-th indicator at the k-th level. The total number of quality evaluation indicators; 4-5-4) Compare the comprehensive membership degrees of the special grade, first grade, second grade and third grade, and take the comprehensive membership degree with the largest value as the comprehensive grading judgment value of the citrus to be sorted.

9. The method according to claim 6, characterized in that, In step 5), after the citrus is graded and collected, the control and monitoring module will clear all the detection data and sorting instructions of the corresponding weighing and unloading unit. At the same time, the weighing and unloading unit returns to the feeding and receiving position with the circular material conveying unit to receive the next citrus to be sorted, thus realizing the continuous grading and collection of citrus.

Images