Intelligent ginseng quality identification method and system

Abstract

The present application relates to ginseng detection technology field, specifically to a kind of ginseng quality intelligent identification method and system, the method includes: whole ginseng is placed in the profiled box with root support area, main root support area and fibrous root overhanging avoidance slot, by conveying device sequentially conveyed through three-dimensional form collection station, spectrum collection station, weighing station and sorting station;Collect the three-dimensional point cloud of ginseng and determine the form parameters;Collect epidermis image and near infrared spectrum and determine the spectral characteristics;Collect quality and determine flesh firmness in combination with point cloud volume;Morphological parameters, spectral characteristics, flesh firmness and quality are fused and processed, to determine the final quality grade of ginseng;The present application can complete automatic grading and sorting under the condition that ginseng fixed posture is maintained and fibrous root is avoided extrusion, improve the accuracy and consistency of ginseng quality identification.

Description

Technical Field

[0001] This invention relates to the field of ginseng testing technology, specifically to an intelligent identification method and system for ginseng quality. Background Technology

[0002] As a high-value traditional Chinese medicine, the quality grade of ginseng directly depends on its appearance, such as the number of rhizome bowls, the size of the main body, the density of ring patterns, and the integrity of the fibrous roots, as well as its internal components, such as saponin content. During the harvesting, processing, and distribution of ginseng, each ginseng root needs to be individually identified and graded to determine its commercial value and intended use.

[0003] In existing technologies, ginseng quality identification mostly relies on manual visual inspection combined with experience. Some processing enterprises have introduced image or spectral detection equipment: taking pictures of the ginseng appearance with a camera to obtain size information, or using a near-infrared spectroscopy device to detect the ginseng's component content, and then giving a grade by manual or simple rules. After grading, the ginseng is sorted into the corresponding collection bins by manual or air-blowing sorting devices.

[0004] However, the existing technologies mentioned above have the following shortcomings: First, ginseng has an irregular shape and fine fibrous roots, which are prone to rolling and breaking under pressure during transportation, resulting in inconsistent sampling standards and large detection errors. Second, the spectral sampling location is arbitrary and is affected by the fibrous roots, epidermal damage, and differences in the thickness of ginseng, making it difficult to stably characterize the saponin content. Third, the detection data such as appearance and composition are independent of each other, lacking a unified fusion and grading rule, and the grading results are highly subjective.

[0005] Therefore, the present invention provides a method and system for intelligent identification of ginseng quality. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] The technical solution adopted by this invention to solve its technical problem is: a method for intelligent identification of ginseng quality, comprising the following steps:

[0008] S1. Place the whole ginseng in a contoured tray. The contoured tray is provided with a rhizome support area, a main root support area, and a fibrous root suspension avoidance groove, so that the ginseng maintains a fixed posture and the fibrous roots hang naturally during the transportation process. The contoured tray is transported by the conveying device through the three-dimensional morphology acquisition station, the spectrum acquisition station, the weighing station, and the sorting station in sequence. The sorting station is provided with at least two sorting bins.

[0009] S2. Collect three-dimensional point cloud of ginseng under fixed posture, extract morphological parameters from the three-dimensional point cloud, the morphological parameters include at least the number of reed bowls, the main body size, the ring density and the integrity of the fibrous roots, and determine the spectral detection area of ​​the middle section of the main root of ginseng based on the three-dimensional point cloud.

[0010] S3. Collect near-infrared spectra of the corresponding region of ginseng according to the location of the spectral detection area to obtain spectral characteristics characterizing saponin content;

[0011] S4. Obtain the mass of the ginseng and calculate the volume of the ginseng using three-dimensional point computing. Obtain the density of the ginseng from the mass and volume, and determine the firmness of the ginseng flesh based on the density and ring density.

[0012] S5. The morphological parameters, spectral characteristics, flesh firmness, and quality are fused to determine the final quality grade of the ginseng. Based on the final quality grade, the air nozzles in the pneumatic sorting module corresponding to the target sorting bin are controlled to blow the ginseng into the target sorting bin. The triggering time of the air nozzles is determined based on the conveying relationship between the weighing station and the target sorting bin, the air nozzle response delay, and the drift time corresponding to the ginseng quality. The blowing pressure is determined based on the ginseng quality.

[0013] Preferably, the conformal support box further includes a tail support area, and the support depths of the rhizome support area, the main root support area, and the tail support area are respectively adapted to the radial dimensions of the corresponding parts of the ginseng; the fibrous root hanging avoidance groove is provided on at least one side of the conformal support box to allow the ginseng fibrous roots to hang down naturally; the conformal support box is provided with an elastic liner, which is used to restrict the rolling and turning of the ginseng during the transportation process.

[0014] Preferably, step S2 specifically comprises:

[0015] S231. At the three-dimensional morphology acquisition station, multi-view point cloud data of ginseng are simultaneously acquired by at least three depth cameras located above the contour tray and on both sides of the conveying path.

[0016] S232. The point clouds collected by each depth camera are converted to the same coordinate reference and stitched together. The main axis of the ginseng is fitted according to the main point cloud. For the missing area of ​​the point cloud on the side of the ginseng that is attached to the contour box, the radial distance of the collected point cloud relative to the main axis of the ginseng is mapped to the missing area in the same axial section to obtain the completed three-dimensional point cloud.

[0017] S233. Segment the completed 3D point cloud to obtain point cloud partitions corresponding to the rhizome, main body and fibrous roots respectively;

[0018] S234. Count the number of concave areas in the reed head point cloud partition to obtain the number of reed bowls; obtain the axial length and maximum radial dimension of the main body point cloud partition to obtain the main body dimension; divide the main body point cloud partition into three equal segments along the axial direction, count the number of annular protrusions around the main body in each segment, and take the ratio of the number of annular protrusions in each segment to the axial length of the corresponding segment as the annular density of that segment.

[0019] S235. Based on the end morphology of the fibrous roots in the fibrous root point cloud partition, the ratio of the number of complete fibrous roots at the end to the total number of fibrous roots is calculated to obtain the fibrous root integrity. The total number of fibrous roots is the sum of the number of identified complete fibrous roots at the end, truncated fibrous roots at the end, and fibrous root remnants. The fibrous root integrity is used as a morphological parameter in the fusion and classification of S5.

[0020] Preferably, step S3 specifically comprises:

[0021] At the spectral acquisition station, the shaped tray is positioned by a positioning mechanism to keep the ginseng stationary during the spectral acquisition process; the middle third section of the main root of the ginseng is determined based on the completed three-dimensional point cloud; the position of the fibrous root protrusion is identified based on the three-dimensional point cloud; and the position of epidermal damage is identified based on the epidermal image registered with the three-dimensional point cloud acquired by the image acquisition device. The surface area of ​​the main root that does not contain the fibrous root protrusion position and the epidermal damage position is determined as the spectral detection area.

[0022] Multiple detection points are determined within the spectral detection area. The spectral probe is moved sequentially to each detection point and near-infrared spectra are collected respectively. The absorption intensity of each detection point in the pre-calibrated saponin characteristic band is normalized according to the equivalent optical path at the corresponding detection point. The normalized absorption intensities are then averaged to obtain the spectral characteristics characterizing the relative content level of saponins.

[0023] Preferably, step S4 specifically comprises:

[0024] When the ginseng passes through the dynamic weighing sensor set on the conveying path along the contour tray, multiple mass values ​​are continuously sampled and the median value is taken. Then, the tare weight of the corresponding contour tray that is pre-stored is deducted to obtain the mass of the ginseng.

[0025] The completed 3D point cloud is reconstructed using a closed surface, and the volume of the ginseng is calculated. The density of the ginseng is determined based on its mass and volume. The average value of the density of the three rings is calculated, and the ginseng density and the average ring density are converted into density score and ring score respectively according to the same normalization rule. The average value of the density score and the ring score is used as the firmness score of the ginseng flesh.

[0026] Preferably, in step S5, the morphological parameters, spectral characteristics, flesh firmness, and quality are fused, specifically as follows:

[0027] S561. Compare the maximum radial dimension of the ginseng body with the preset specification limit. If the maximum radial dimension is greater than the specification limit, the ginseng is classified into the large specification category; if the maximum radial dimension is less than or equal to the specification limit, the ginseng is classified into the small specification category.

[0028] S562. Convert the integrity of the rootlets, spectral characteristics, and quality into dimensionless scores according to the pre-stored normalization upper and lower limits, and use the determined flesh firmness score as the dimensionless score corresponding to the flesh firmness; weight and fuse each dimensionless score according to the fusion weight corresponding to the ginseng's specification category to obtain the fusion score; among them, in the fusion weight corresponding to the large specification category, the weight of the rootlet integrity is greater than the weight of the spectral characteristics; in the fusion weight corresponding to the small specification category, the weight of the spectral characteristics is greater than the weight of the rootlet integrity.

[0029] S563. Determine the initial quality grade of ginseng based on the fusion score and the pre-stored fusion score grade threshold.

[0030] S564. Verify the initial quality level based on the integrity of the hair roots to determine the final quality level; wherein, when the integrity of the hair roots is lower than a preset integrity threshold, the initial quality level is downgraded by one level to obtain the final quality level; when the integrity of the hair roots is not lower than the preset integrity threshold, the initial quality level is determined as the final quality level.

[0031] Preferably, step S5, which controls the air nozzles in the pneumatic sorting module corresponding to the target sorting bin according to the final quality level, includes:

[0032] S571. After obtaining the final quality grade of ginseng, call the pre-stored correspondence between the final quality grade and the sorting warehouse, determine the target sorting warehouse corresponding to the final quality grade, and assign the high quality grade to the first sorting warehouse, the medium quality grade to the second sorting warehouse, and the low quality grade to the third sorting warehouse.

[0033] S572. Determine the target sorting bin for ginseng based on the final quality grade of ginseng and the corresponding relationship, and generate a sorting instruction carrying the target sorting bin information;

[0034] S573. Based on the conveying relationship between the ginseng-containing contour tray and the target sorting bin along the conveying path, determine the execution time of the sorting instruction, and when the execution time is reached, send the sorting instruction to the pneumatic sorting module, which will then blow the ginseng into the target sorting bin.

[0035] Preferably, after determining the final quality grade of ginseng in step S5, the process also includes establishing and tracing quality records, specifically:

[0036] The final quality grade, morphological parameters, spectral characteristics, flesh firmness and quality of ginseng are linked and stored with the number of the molded tray and the batch to which it belongs, forming a quality file for each ginseng.

[0037] The system calculates the percentage of high, medium, and low quality grades among ginseng that has completed grading within the same batch. When the number of ginseng that has completed grading within the statistical period reaches the minimum number of samples stored in advance, and the percentage of low quality grades increases sequentially for at least two consecutive statistical periods, a batch quality warning is generated. The system also retrieves the quality records of ginseng that has completed grading within the same batch based on the batch information, enabling full traceability of the entire process from self-testing and grading of a single ginseng to warehousing.

[0038] Preferably, the pneumatic sorting module's process of blowing ginseng into the target sorting bin specifically includes the following steps:

[0039] S901. Using the weighing station outlet as the trigger reference position, determine the remaining conveying time for the ginseng to move from the trigger reference position to the target sorting bin based on the distance along the conveying path between the trigger reference position and the target sorting bin and the conveying speed of the conveying device; use the sum of the response delay of the air nozzle corresponding to the target sorting bin, the drift time corresponding to the ginseng mass, and the time correction amount corresponding to the target sorting bin as the trigger advance amount; when the remaining conveying time reaches the trigger advance amount, trigger the air nozzle corresponding to the target sorting bin.

[0040] S902. Compare the quality of the ginseng with the average quality of the same batch of ginseng. If the quality of the ginseng is less than the average quality, the ginseng is determined to be in the first quality grade; otherwise, the ginseng is determined to be in the second quality grade. Read the corresponding drift time and blowing pressure from the pre-stored pneumatic parameter calibration table according to the quality grade.

[0041] S903. A photoelectric sensor is installed at the entrance of the sorting bin to detect the actual landing point of the ginseng relative to the center of the target sorting bin entrance, and the correction direction is determined based on the landing point deviation between the actual landing point and the center of the target sorting bin entrance.

[0042] S904. When the ginseng landing point deviation corresponding to the same target nozzle is too far back in at least two consecutive sorting processes, the time correction amount corresponding to the target nozzle is increased by a preset correction step size; when the ginseng landing point deviation corresponding to the same target nozzle is too far forward in at least two consecutive sorting processes, the time correction amount corresponding to the target nozzle is decreased by a preset correction step size.

[0043] S905. The time correction amount corresponding to each target nozzle is accumulated independently; when the actual landing point after correction returns to the allowable deviation range of the target sorting warehouse entrance center, the current time correction amount is maintained; when the subsequent landing point deviation direction is opposite to the accumulated correction direction, the time correction amount is backed back to the initial value by one correction step.

[0044] A ginseng quality intelligent identification system includes a conveying device, at least two contoured trays spaced apart on the conveying device, a three-dimensional morphology acquisition device, an image acquisition device, a spectral acquisition device and a weighing device arranged sequentially along the conveying path of the conveying device, a positioning mechanism set at the spectral acquisition station, a grading processing module, a pneumatic sorting module, at least two pairs of photoelectric sensors set at the entrance of each sorting bin, and a storage module.

[0045] The conformal tray is used to support the entire ginseng and maintain its fixed posture; the three-dimensional morphology acquisition device is used to acquire the three-dimensional point cloud of the ginseng and extract morphological parameters; the image acquisition device is used to acquire the ginseng epidermal image registered with the three-dimensional point cloud; the positioning mechanism is used to limit the conformal tray during the spectral acquisition process; the spectral acquisition device is used to acquire the near-infrared spectrum of the ginseng and obtain spectral characteristics; the weighing device is used to obtain the mass of the ginseng.

[0046] The grading module is used to determine the ginseng density based on its mass and volume, and to determine the flesh firmness based on the ginseng density and ring density. It also integrates the morphological parameters, spectral characteristics, flesh firmness, and mass to determine the final quality grade of the ginseng. The pneumatic sorting module includes air nozzles corresponding to the sorting bins, used to blow the ginseng into the corresponding sorting bins according to the final quality grade. The photoelectric sensor is used to detect the actual landing point of the ginseng. The storage module stores the tartrate weight of the conformal tray, saponin characteristic bands, normalized upper and lower limits, pneumatic parameter calibration tables, and quality records.

[0047] This invention provides a method and system for intelligent identification of ginseng quality. It has the following beneficial effects:

[0048] 1. By using a contoured tray, ginseng is kept in a fixed posture during transportation and its fibrous roots hang naturally. The tray is used as a spatial reference to complete multi-view point cloud splicing, occlusion completion and part segmentation. This achieves automatic quantification of appearance indicators such as the number of reed bowls, ring density and fibrous root integrity, avoiding detection errors caused by ginseng rolling and turning, and reducing fibrous root damage during transportation and sorting.

[0049] 2. Near-infrared spectroscopy is used to sample at fixed points in the middle section of the main root by point cloud guidance, eliminating root points, damaged epidermis and contaminated areas, and normalizing the spectral absorption according to the local equivalent root diameter, thus eliminating the interference of ginseng morphology, posture and thickness differences on spectral detection; the quality is graded by integrating morphological, spectral, quality and density characteristics, taking into account both the appearance and internal components of ginseng, and improving the accuracy and stability of grading results.

[0050] 3. By determining the nozzle trigger advance and blowing pressure based on the ginseng quality, and using the actual landing point deviation to independently correct the trigger timing of each sorting bin, detection, grading and sorting are coordinated and controlled. This can automatically compensate for landing point drift caused by air pressure fluctuations, nozzle aging and conveying speed changes, improve the accuracy of multi-level sorting and the stability of long-term continuous operation of the equipment, and reduce manual calibration and intervention. Attached Figure Description

[0051] Figure 1 This is a flowchart of the method of the present invention;

[0052] Figure 2 This is a schematic diagram of the system structure of the present invention;

[0053] Figure 3 This is a flowchart of the fusion hierarchy process of the present invention. Detailed Implementation

[0054] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the invention. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the invention for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0055] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0056] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0057] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0058] like Figures 1 to 3 As shown, this invention proposes an intelligent identification method for ginseng quality, comprising the following steps:

[0059] S1. Place the whole ginseng in a contoured tray. The contoured tray is equipped with a rhizome support area, a main root support area, and a fibrous root suspension avoidance groove, so that the ginseng maintains a fixed posture and the fibrous roots hang naturally during the transportation process. The contoured tray is transported by the conveying device through the three-dimensional morphology acquisition station, the spectrum acquisition station, the weighing station, and the sorting station in sequence. The sorting station is equipped with at least two sorting bins.

[0060] S2. Collect three-dimensional point clouds of ginseng in a fixed posture, extract morphological parameters from the three-dimensional point cloud, including at least the number of reed bowls, main body size, ring density and fibrous root integrity, and determine the spectral detection area of ​​the middle section of the main root of ginseng based on the three-dimensional point cloud.

[0061] S3. Collect near-infrared spectra of the corresponding region of ginseng based on the location of the spectral detection area to obtain spectral characteristics characterizing saponin content;

[0062] S4. Obtain the mass of the ginseng and calculate the volume of the ginseng using three-dimensional point computing. Calculate the density of the ginseng from the mass and volume, and determine the firmness of the ginseng flesh based on the density and ring density.

[0063] S5. The morphological parameters, spectral characteristics, flesh firmness and quality are fused to determine the final quality grade of ginseng. Based on the final quality grade, the air nozzles in the pneumatic sorting module corresponding to the target sorting bin are controlled to blow the ginseng into the target sorting bin. The triggering time of the air nozzles is determined based on the conveying relationship between the weighing station and the target sorting bin, the air nozzle response delay and the drift time corresponding to the ginseng quality, and the blowing pressure is determined based on the ginseng quality.

[0064] It should be noted that the thickness of a ginseng root varies significantly along its axis: the rhizome is short and thick, the main body is swollen, and the tail gradually tapers, with multiple fibrous roots extending from the lower part of the main body. If the tray's accommodating cavity is a groove of uniform depth, the ginseng will only partially contact the tray after being placed inside, making it prone to rolling and shifting during transport, unable to maintain a fixed posture; at the same time, the fibrous roots, lying flat on the tray's surface, are easily squeezed, pulled, and broken during transport, affecting the ginseng's appearance and value.

[0065] Based on this, in this embodiment, the accommodating cavity of the contour tray is divided into a rhizome support section, a main body support section, and a tail support section along the ginseng's axial direction. The depth of the three sections varies according to the thickness of the corresponding parts of the ginseng, with the main body support section being the deepest and the tail support section being the shallowest. Specifically, the depth of each support section can be preset according to the specifications and dimensions of the ginseng to be sorted, so that after the ginseng is placed in the accommodating cavity, all parts of its axial direction are supported, thereby being transported in a fixed posture, providing a consistent acquisition benchmark for subsequent three-dimensional point cloud acquisition and spectral acquisition.

[0066] Furthermore, the accommodating cavity of the conformal tray is provided with root hanging clearance grooves on both sides for ginseng rootlets to pass through. The root hanging clearance grooves are slots that penetrate through the side wall of the conformal tray. The ginseng rootlets hang down naturally through the root hanging clearance grooves and do not come into contact with the conveying surface, thereby reducing the risk of the rootlets being squeezed, pulled and broken during the conveying process.

[0067] In one embodiment, the inner wall of the accommodating cavity is also provided with an elastic liner. The elastic liner can be made of materials with elastic deformation capabilities such as silicone or sponge. After the ginseng is placed in, the elastic liner is pressed tightly against the surface of the ginseng body. In this way, on the one hand, it can adapt to the differences in thickness and branching degree between different individuals, improving the versatility of the tray; on the other hand, it can limit the rolling and movement of the ginseng during the transportation process, further ensuring the stability of the fixed posture.

[0068] Therefore, this implementation method, through the combination of segmented support, root suspension, and elastic close contact, ensures that ginseng maintains a stable and undamaged posture throughout the transportation process, thereby improving the consistency of subsequent quality inspection and the accuracy of grading results.

[0069] In one embodiment, the process of acquiring a three-dimensional point cloud and extracting morphological parameters using a three-dimensional morphological acquisition device is further described as follows:

[0070] It should be noted that the quality grade of ginseng is closely related to its appearance: the number of rhizome bowls reflects the growth years of ginseng, the main body size reflects its specifications, the density of ring patterns reflects the growth years and texture, and the integrity of the fibrous roots directly affects the commercial value. However, ginseng has an irregular shape and fine fibrous roots, and the point cloud collected from a single perspective has occlusion and missing parts, making it difficult to fully represent the above-mentioned morphology. Based on this, this implementation method uses the following steps to extract morphological parameters.

[0071] In S231, depth cameras are set above and to the side of the transport path. A depth camera is an imaging device that can acquire depth information of each point on the surface of the object being photographed, such as a structured light camera or a time-of-flight camera. The upper depth camera collects the point cloud of the top surface and upper side of the ginseng, and the side depth camera collects the point cloud of the side of the ginseng, thereby covering most of the outer surface of the ginseng.

[0072] In S232, since the ginseng is supported in a fixed posture within the contoured tray, the poses of each depth camera relative to the ginseng are determined. Therefore, the point clouds from each viewpoint can be stitched together according to the coordinate reference corresponding to this fixed posture. Furthermore, the point cloud on the side of the ginseng that is against the contoured tray is missing. Considering that the ginseng body is approximately a shape rotating around its axis, this embodiment uses the symmetry of the main body point cloud relative to the ginseng axis to deduce the missing point cloud on the side against the tray from the already collected point cloud on one side. The multi-view point cloud data is then stitched together, and the stitched 3D point cloud is symmetrically completed to obtain the completed 3D point cloud.

[0073] It should be noted that this completion method is feasible precisely because it relies on the premise of a fixed ginseng posture: after fitting the main body point cloud along its axis, the main body point cloud is converted into cylindrical coordinates based on the axis; for the missing circumferential region on the cascade side, the radial distance of the already acquired side relative to the axis within the same axial section is used to generate corresponding completion points; when no usable symmetrical points exist on the same axial section, interpolation is performed using the radial distance between adjacent axial sections. The completion points are used for calculating the main body dimensions, equivalent optical path, and volume; the integrity of the fibrous roots is determined based on the actual acquired fibrous root point cloud.

[0074] In S233, the completed 3D point cloud is segmented to obtain point cloud partitions corresponding to the rhizome, the main body, and the fibrous roots. Specifically, the partitions can be made according to the radial dimension changes of the completed 3D point cloud along the ginseng axis: the rhizome is a short segment with a sudden change in radial dimension at one end of the axis, the main body is the middle segment with the largest radial dimension, and the fibrous roots are slender branches extending from the surface of the ginseng main body. Thus, each point cloud partition corresponds one-to-one with the actual part of the ginseng, providing a regional basis for subsequent parameter extraction.

[0075] In S234, the reed bowls are the depressions formed by the gradual shedding of stem scars from the reed head over the years. Therefore, by counting the number of depression areas in the reed head dot cloud region, the number of reed bowls can be obtained. The unit is "pieces," and the main dimensions include the axial length L and the maximum radial dimension D of the main point cloud partition, both in cm. The ring density is determined by the following formula:

[0076] Where i=1,2,3, is the segment number after the main point cloud partition is divided into three segments along the axis; Let be the ring density of the i-th segment, in units of rings / cm; This represents the number of annular ridge and ridge stripes on the surface of the i-th segment, in units of stripes. The axial length of the i-th segment is in cm. It should be noted that the annular density is used here instead of the total number of annular rings in order to eliminate the influence of the difference in the length of the ginseng body and to make the annular characteristics of ginseng of different specifications comparable. The reason for further segmented statistics instead of overall statistics is that the annular rings of ginseng are not evenly distributed along the axial direction and are usually denser near the root head. The segmented density can reflect the texture differences that are masked by the overall density.

[0077] In S235, the total number of fibrous roots is the sum of the number of identified intact fibrous roots, truncated fibrous roots, and fibrous root remnants. The integrity of the fibrous roots is determined by the following formula:

[0078] , For the integrity of the fibrous roots; This refers to the number of intact terminal fibrous roots; The number of fibrous roots cut off at the end. To identify the number of root remnants, This represents the total number of fibrous roots;

[0079] It should be noted that complete fibrous roots, truncated fibrous roots, and fibrous root remnants are identified as follows: branches are extracted from the fibrous root point cloud partitions, each independent branch extending from the surface of the main point cloud partition is identified as a fibrous root, and the skeleton line of the branch is extracted; the skeleton line refers to the line that represents the direction of the branch center after the branch point cloud is gradually contracted along its extension direction, and the starting point of the skeleton line is the connection point between the branch and the main surface.

[0080] Multiple cross-sections are cut along the skeleton line at preset intervals, and the radial dimension of each cross-section is calculated: From the connection point to the end along the skeleton line, the radial dimension of each cross-section decreases sequentially, and the radial dimension of the end cross-section is less than the product of the radial dimension of the cross-section at the connection point and a preset ratio; these are identified as complete end roots and included in the calculation. If the radial dimension of the end cross-section is greater than or equal to the product of the radial dimension of the cross-section at the connection point and a preset ratio, and the length of the skeleton line is greater than the radial dimension of the cross-section at the connection point, it is determined as an end-cut fibrous root and included in the calculation. If the length of the skeleton line is less than or equal to the radial dimension of the cross-section at the connection point, its extension length is insufficient to determine the end morphology and is identified as a root remnant, and is included in the calculation. The preset ratio can be determined and stored in advance during the equipment calibration stage based on the statistical analysis of the end morphology of complete ginseng rootlets of the same variety.

[0081] The ratio of intact rootlets to the total number is used as the integrity value to quantify the degree of broken rootlets, so that the differences in appearance between different ginsengs have a unified comparative scale. Rootlet integrity value, as one of the morphological parameters, is input into the grading module along with the number of rootlets, the main body size, and the density of ring patterns to participate in the fusion grading. Rootlet integrity value plays a dominant role in the fusion weight of large-scale categories and serves as the basis for subsequent verification of the initial quality grade.

[0082] Therefore, this implementation eliminates acquisition occlusion through multi-view acquisition and symmetrical completion, and provides clear regional basis for the extraction of each morphological parameter through point cloud partitioning. Each parameter is determined in a quantifiable way, avoiding the subjectivity of manual visual inspection and grading. The extracted root integrity, together with the number of reed bowls, the main body size, and the ring density, constitute the appearance quality dimension of the fusion grading, providing input for subsequent specification-based weighted fusion and root integrity verification.

[0083] In one embodiment, the process of acquiring the near-infrared spectrum of ginseng using a spectral acquisition device is further described.

[0084] It should be noted that the distribution of ginsenosides in the main root, lateral roots, and fibrous roots is not uniform, with the saponin content in the middle section of the main root being the most representative. Furthermore, the main root itself is uneven in thickness, and the epidermis may have damage and rust spots. Therefore, if only single-point spectra are collected, it is easily affected by the collection location and the condition of the epidermis, making it difficult to stably characterize the saponin level of the entire ginseng root. Based on this, the following collection process is adopted in this embodiment:

[0085] The characteristic bands of saponins are determined through standard sample calibration. Specifically, standard samples of the same variety as the ginseng to be tested are selected, and the total saponin content of each standard sample is determined by physicochemical detection methods. Near-infrared spectra of the middle section of the main root of each standard sample are also collected. The correlation between the absorbance of each band and the total saponin content is calculated. One or more bands that meet the preset requirements for correlation with the total saponin content are selected as saponin characteristic bands, and the saponin characteristic bands are pre-stored in the grading processing module. During detection, the average absorbance of the spectrum of the detection point within the saponin characteristic bands is taken as the absorption intensity of the detection point.

[0086] The image acquisition device and the depth camera are pre-registered to ensure that the pixel positions in the epidermal images acquired by the image acquisition device can be mapped to the corresponding surface positions in the 3D point cloud. The hierarchical processing module determines the location of epidermal damage based on color and texture anomaly areas in the epidermal images and maps the epidermal damage locations to the 3D point cloud for removal from the spectral detection area.

[0087] Determination of the spectral detection zone: Specifically, based on the three-dimensional point cloud obtained by the three-dimensional morphology acquisition device, the position and direction of the ginseng taproot are first determined. Then, the surface area of ​​the middle third section of the taproot is taken as the spectral detection zone. When determining the detection zone, the location of epidermal damage is identified based on the epidermal image registered with the ginseng three-dimensional point cloud, and the location of epidermal damage is mapped to the completed three-dimensional point cloud to determine the corresponding three-dimensional spatial position. That is, the detection zone only covers the intact surface area of ​​the middle section of the taproot epidermis. It should also be noted that the saponin distribution at the fibrous roots differs significantly from that at the taproot. The exposure of tissue at the epidermal damage site will change the reflectance characteristics of the spectrum. Excluding these two factors can avoid interference and ensure that the acquired spectrum truly reflects the saponin content of the taproot.

[0088] Before the spectral probe collects data point by point, the positioning mechanism limits the shape-following tray, keeping it stationary at the spectral acquisition station, thus keeping the ginseng still during the spectral acquisition process; after all detection points are collected, the positioning mechanism releases the shape-following tray, and the conveying device continues to transport the shape-following tray to the weighing station.

[0089] Arrangement and acquisition of detection points: Multiple detection points are determined within the spectral detection area, for example, three detection points are arranged at intervals along the main root axis. The spectral acquisition device includes a position-adjustable spectral probe, which is an optical component that emits near-infrared light and receives reflected light from ginseng. It can be mounted on a robotic arm or a moving slide to change its position. The spectral probe moves sequentially to each detection point and acquires a near-infrared spectrum at each detection point. Since the position of each detection point is determined by the three-dimensional point cloud, the movement of the spectral probe has a clear target position and does not require manual alignment.

[0090] Normalization and fusion of absorption intensity: For the near-infrared spectra of each detection point, the absorption intensity at the pre-calibrated characteristic band of saponins is taken. Where j is the detection point number, the transmission path length of near-infrared light into the ginseng is related to the thickness of the taproot: the thicker the taproot, the longer the optical path, and the greater the natural absorption intensity. This difference is caused by the thickness rather than the saponin content. To eliminate the interference of thickness differences, the absorption intensity of each detection point is normalized according to the radial dimension of the taproot at its corresponding detection point:

[0091] ,in The normalized absorption intensity at the j-th detection point; denoted as the average absorbance of the j-th detection point within the characteristic band of saponins; Let be the equivalent optical path length at the j-th detection point, in cm. After normalization, the absorption intensity only reflects the absorption level per unit optical path length. The saponin characteristics of ginseng of different coarse and fine grades can thus be compared. The equivalent optical path length can be determined by the local radial dimension obtained from the three-dimensional point cloud and the incident angle of the spectral probe. For perpendicular incidence, the approximate equivalent optical path length can be taken as the local radial dimension.

[0092] The normalized absorption intensities are then combined to obtain the spectral characteristics representing the saponin content:

[0093] ,in Spectral characteristics; n is the number of detection points, in units of points. It should be noted that multi-point fusion is used here instead of single-point sampling because the distribution of saponins within the main root fluctuates. Multi-point averaging can suppress random deviations at single points, making the spectral characteristics more stable in representing the saponin levels of the entire ginseng root.

[0094] Therefore, this implementation method uses three-dimensional point clouds to determine the detection area, arrange detection points, and provide the radial dimensions required for normalization, so that the location, quantity, and data processing of spectral acquisition are all adapted to the actual morphology of individual ginseng. At the same time, through triple processing of damage avoidance, coarse and fine normalization, and multi-point fusion, the accuracy and comparability of spectral features are improved, providing reliable input for subsequent fusion and grading.

[0095] In one embodiment, the method by which the ginseng mass is obtained by the weighing device is further described as follows:

[0096] Since the ginseng passes through the weighing position along with the conformal tray during transportation, the weighing result includes the weight of the tray itself. Furthermore, weighing during movement is subject to vibration interference, resulting in significant fluctuations in the values ​​of a single sample. Additionally, grading solely by quality is unfair: for ginseng of similar size, those with firmer flesh are of higher quality, and firmness cannot be solely reflected by quality. Therefore, this implementation method adopts the following approach:

[0097] Dynamic weighing with tare weight: Specifically, the weighing device is a dynamic weighing sensor set on the conveyor path. The dynamic weighing sensor is a sensor that continuously measures the mass of an object as it moves through the conveyor, such as a belt-type weighing sensor. The tare weight of each contoured tray is pre-stored. Tare weight refers to the mass of the contoured tray itself. It can be obtained by weighing each contoured tray one by one before sorting and stored in relation to its number.

[0098] When the ginseng passes through the weighing device with the conformal tray, the dynamic weighing sensor continuously samples multiple mass values ​​and takes the median value. Then, the tare weight of the corresponding conformal tray is subtracted to obtain the mass of the ginseng. It should be noted that the median value is taken instead of the average value because abnormal sampling values ​​caused by vibration usually deviate from the normal range. The median value is not affected by individual abnormal values ​​and has stronger anti-interference ability. Subtracting the tare weight eliminates the influence of the difference in the self-weight of different trays on the weighing result.

[0099] The density and firmness of the ginseng are obtained by computing the volume of the ginseng using three-dimensional point clouds. Since the three-dimensional point cloud fully represents the shape of the ginseng, a closed surface model of the ginseng body is reconstructed based on the completed three-dimensional point cloud. The closed surface model is then divided and integrated to obtain the volume V of the ginseng body. The density of the ginseng is determined by the following formula:

[0100] , The density of ginseng is expressed in g / cm³. The ginseng mass after basal removal is expressed in g; V is the ginseng volume calculated from the closed surface reconstructed from the completed 3D point cloud, expressed in cm³. Density, rather than mass, is used in the flesh quality evaluation because density eliminates the influence of differences in ginseng size and reflects the mass per unit volume, which is directly related to the fullness of the flesh.

[0101] Furthermore, the average value of the three ring density segments is calculated, and the ginseng density and average ring density are converted into density score and ring score respectively according to the same normalization rule. The average value of density score and ring score is used as the ginseng flesh firmness score.

[0102] Finally, the fusion process is carried out: meat firmness, morphological parameters, spectral characteristics, and quality are all included in the fusion process as the basis for determining the quality grade. Thus, in the conventional weighing process, this implementation method improves the accuracy of quality data by removing the tare weight and taking the median value. Furthermore, by combining quality with point cloud volume and ring density, meat firmness with a clear quality orientation is derived. This allows the weighing data and three-dimensional point cloud data to be reused in the grading process, enriching the dimensions of the fusion grading.

[0103] In some implementations, the focus of quality evaluation differs for ginseng of different sizes: larger ginseng is often sold as a gift or whole, and the integrity of its rootlets directly affects its appearance and value; smaller ginseng is often used as raw material for processing, and its internal saponin content is more critical. If a uniform fusion weight is applied to ginseng of all sizes, it will lead to a misalignment of evaluation focus, and the grading results will not match the actual value. In addition, the quality grades corresponding to each parameter may be inconsistent. For example, a high saponin content may be accompanied by many broken rootlets. In this case, clear rules are needed for grading. Based on this, this implementation adopts the fusion processing steps S561-S564:

[0104] In S561, the main body size of the ginseng is compared with a preset specification limit. Ginseng with a main body size larger than the specification limit is classified into the large specification category, and ginseng with a main body size smaller than or equal to the specification limit is classified into the small specification category. Specifically, the main body size can be the maximum radial dimension D of the main body, and the specification limit can be preset to 2.0 cm. That is, ginseng with D greater than 2.0 cm is classified into the large specification category, and ginseng with D less than or equal to 2.0 cm is classified into the small specification category. The specific value of the specification limit can be adjusted in advance according to the ginseng commodity grading standard, and this embodiment does not limit it.

[0105] S562. Convert the integrity of the rootlets, spectral characteristics, flesh firmness, and quality into normalized scores according to the pre-stored lower and upper normalization limits of each parameter. Then, perform weighted fusion of each score according to the fusion weight corresponding to the ginseng's specification category. The fusion process is performed according to the following calculation formula:

[0106] Where S is the fusion score; , , , These are normalized scores for root integrity, spectral characteristics, flesh firmness, and quality, respectively. , , , For the corresponding fusion weights, and =1. It should be noted that the four parameters mentioned above have different dimensions and cannot be directly weighted. Therefore, they are first converted into dimensionless scores and then fused so that the fused scores can uniformly represent the comprehensive quality of ginseng.

[0107] For the k-th parameter whose quality evaluation increases with increasing value, the lower limit of the parameter is determined based on the calibrated samples of the same variety. and parameter upper limit and the measured value of the k-th parameter Convert to dimensionless score according to the following relationship. and These are the lower and upper limits of the k-th parameter, respectively, determined based on calibrated samples of the same variety:

[0108]

[0109] in Let k be the measured value of the k-th parameter. The dimensionless score for the k-th parameter; the meat firmness score is obtained by averaging the ginseng density score and the average ring density score;

[0110] Among the fusion weights corresponding to large-scale categories, the weight of root integrity is greater than the weight of spectral features, for example... , , , This makes the integrity of the fibrous roots the dominant factor in the quality grade, consistent with the evaluation focus of large-sized ginseng based on its appearance; in the fusion weighting corresponding to smaller-sized categories, the weight of spectral characteristics is greater than the weight of the integrity of the fibrous roots, for example... , , , This makes spectral characteristics dominant, which is consistent with the evaluation focus of small-sized ginseng on its medicinal efficacy;

[0111] In S563, the fusion score is divided into three quality levels—high, medium, and low—according to multiple preset score intervals. Each score interval corresponds to a quality level, and the quality level corresponding to the interval into which the fusion score falls is the initial quality level of the ginseng. The boundary values ​​of the score intervals can be preset and adjusted according to the ginseng's commercial grading standards; this implementation method does not limit this. It should be noted that by first weighting the scores according to specifications and then uniformly grading them according to score intervals, the initial quality levels of ginseng of different specifications are comparable on the same scale, facilitating subsequent unified verification of root integrity.

[0112] In S564, the initial quality grade is checked based on the integrity of the ginseng rootlets to determine the final quality grade. Specifically, the integrity of the ginseng rootlets is compared with a pre-stored integrity threshold. When the integrity of the rootlets is lower than the integrity threshold, the initial quality grade is downgraded by one grade to obtain the final quality grade. When the integrity of the rootlets is not lower than the integrity threshold, the initial quality grade is determined as the final quality grade. When the initial quality grade is already the lowest grade and the integrity of the rootlets is lower than the integrity threshold, the lowest grade is maintained. This setting is because the integrity of the rootlets is a hard indicator affecting the value of ginseng as a commodity. Even if the internal quality of ginseng is up to standard, the actual commodity value of ginseng with damaged appearance is still determined at a lower grade.

[0113] In some implementations, after determining the final quality grade of ginseng, this grade needs to be converted into a sorting action that can be performed by the pneumatic sorting module. This implementation uses processes S571-S573 to achieve the connection between quality grade and sorting execution:

[0114] In S571, the pre-stored correspondence between the final quality grade and the sorting bin is invoked. Specifically, multiple sorting bins are set at intervals along the conveying path of the conveying device, with the high quality grade corresponding to the first sorting bin, the medium quality grade corresponding to the second sorting bin, and the low quality grade corresponding to the third sorting bin. The correspondence is pre-stored in the grading processing module and can be adjusted according to the actual number and layout of the sorting bins. In this embodiment, the specific location of each sorting bin is not limited.

[0115] It should be noted that by establishing and storing the correspondence between grades and sorting bins in advance, each ginseng can have a unique destination once its final quality grade is determined, avoiding delays caused by ad-hoc decisions during sorting.

[0116] In S572, the target sorting bin for the ginseng is determined based on its final quality grade and the aforementioned correspondence, and a sorting instruction carrying the target sorting bin information is generated. Specifically, the sorting instruction is bound to the number of the shaped tray as the shaped tray is conveyed, so that the ginseng, the tray, and the target sorting bin correspond one-to-one. Even if multiple ginseng roots are in a state of waiting to be sorted on the conveying path at the same time, the sorting instructions will not be confused with each other.

[0117] In S573, the execution time of the sorting instruction is determined based on the conveying relationship between the ginseng-containing contour tray and the target sorting bin along the conveying path. When the execution time is reached, the sorting instruction is sent to the pneumatic sorting module, which then blows the ginseng into the target sorting bin.

[0118] It should be noted that the ginseng passes through each sorting bin sequentially during the transport process. The sorting instruction needs to be issued at an appropriate time before the ginseng arrives at the target sorting bin, allowing time for the air nozzle to respond and for the ginseng to drift. The execution time is determined based on the real-time position of the tray and the transport relationship between the tray and the target sorting bin, rather than being issued at a fixed cycle, thus adapting to working conditions with fluctuating transport speeds.

[0119] Therefore, this implementation method, through a three-level connection of grade mapping, instruction generation, and timed issuance, enables the quality grade identification result to be directly and uniquely converted into a sorting action. No manual intervention is required between identification and sorting, and it provides an instruction basis for the precise control of the trigger lead time in the following text.

[0120] In one implementation, after determining the final quality grade of ginseng, the process also includes establishing and tracing quality records, as detailed below:

[0121] It should be noted that after ginseng grading, if the ginseng is simply blown into the corresponding sorting bins according to its grade, the abundant data obtained during the testing process—including morphological parameters extracted from 3D point clouds, spectral characteristics extracted from near-infrared spectroscopy, and flesh firmness and quality derived from weighing—will be discarded after sorting. This data cannot support the verification of the ginseng's origin, the handling of quality disputes, and the improvement of processes in the planting and processing stages. Furthermore, the overall quality level of the same batch of ginseng also needs to be monitored quantitatively to promptly detect batch-specific quality declines. Therefore, this implementation method establishes quality files and enables traceability after grading, as detailed below:

[0122] Establishment of quality files: The final quality grade, morphological parameters, spectral characteristics, flesh firmness, and quality of ginseng are linked and stored with the serial number of the ginseng's packaging box and the batch number it belongs to, forming a quality file for each ginseng root; among which, the morphological parameters include at least the number of rhizome bowls. Main body dimensions (axial length L and maximum radial dimension D), and ring density of each segment. and the integrity of the fibrous roots Spectral characteristics are represented by the fusion absorption intensity, which characterizes the saponin content. The firmness of the meat is determined by its density. And the ring density is determined;

[0123] Each parameter is stored in a corresponding manner with its quality grade, tray number, and batch information, ensuring that each ginseng has a complete and independently searchable data record.

[0124] Batch quality monitoring: This involves statistically analyzing the percentage of high, medium, and low quality grades of ginseng within the same batch. Specifically, let N be the total number of graded ginseng in the same batch, measured in pieces, and let the quantity of low-quality ginseng be [missing information]. If the unit is "piece", the proportion of low-quality grades is determined by the following formula:

[0125] Where μ represents the proportion of low-quality grades, and the statistical period is the preset detection duration or the preset number of trays; when the number of ginseng that has been graded within the statistical period reaches the preset minimum number of samples, the proportion of low-quality grades in adjacent statistical periods is compared to avoid false alarms caused by insufficient sample quantity.

[0126] It should be noted that the low quality grade of a single ginseng is mostly due to individual differences, while the continuous increase in the proportion of low quality grade usually reflects batch factors, such as insufficient sap and gas due to early harvesting, or increased breakage of ginseng roots due to improper processing.

[0127] Therefore, "continuous increase" is used as the early warning condition. When the proportion of low-quality grade μ increases sequentially within at least two consecutive statistical periods, a batch quality early warning is generated to remind operators to check the source and processing status of the ginseng in that batch, so as to avoid batch quality problems being masked by the averaged grading results. When the number of ginseng that has been graded within a statistical period is less than the minimum number of samples stored in advance, that statistical period will not participate in the continuity judgment and will not generate a batch quality early warning, so as to avoid false alarms caused by insufficient sample quantity.

[0128] Full-process traceability: Based on batch information, the quality files of ginseng that have completed grading within the same batch are retrieved, enabling full traceability of a single ginseng from self-testing and grading to warehousing. Specifically, traceability can be conducted in two directions: First, by forward retrieval of batch information, the quality files of all ginseng in that batch are obtained to obtain the grade distribution and parameter distribution of that batch, which is used for batch quality evaluation and settlement basis; Second, by reverse retrieval of the quality file of a single ginseng from its tray number or warehousing information, the various testing parameters and grading process of that ginseng are restored, which is used for verification of quality disputes.

[0129] Therefore, this implementation method will precipitate the identification data into quality archives after classification, and realize batch quality monitoring through proportion statistics and full-process traceability through two-way retrieval, so that the identification results are transformed from a one-time sorting basis into a reusable quality data asset.

[0130] As an optional embodiment, based on the sorting instruction issued by S573, there is a response delay from the air nozzle receiving the sorting instruction to the actual formation of airflow. After being subjected to airflow, ginseng also needs to undergo a drifting process before falling into the sorting bin. If a fixed triggering time is used, when the conveying speed fluctuates or the ginseng quality is different, there may be a situation where the airflow has not yet formed or the drift has not been completed when the ginseng arrives at the sorting bin, resulting in missorting. If a uniform blowing pressure is used, light ginseng is prone to tumbling and damage due to excessive impact, while heavy ginseng may not be blown into place due to insufficient pressure. In addition, there are individual differences in the installation angle and air output efficiency of each air nozzle, and the air pressure will decrease with the use time, resulting in a recurring and fixed systematic deviation between the actual landing point of the ginseng and the target sorting bin.

[0131] Based on this, the pneumatic sorting module of this embodiment performs blowing according to the process S901-S905, as detailed below:

[0132] Specifically, in S901, the weighing station exit is used as the trigger reference position. Based on the distance along the conveying path between the trigger reference position and the target sorting bin, as well as the real-time conveying speed of the conveying device, the remaining conveying time required for the ginseng to move from the trigger reference position to the target sorting bin is determined.

[0133] ;in, The remaining conveying time is in seconds; X is the distance along the conveying path between the trigger reference position and the target sorting bin, in meters; v is the conveying speed of the conveying device, in meters per second; the conveying speed is obtained in real time by the conveying device, so the remaining conveying time can be updated in real time according to the fluctuation of the conveying speed.

[0134] The nozzle response delay is measured during the equipment commissioning phase: a trigger signal is sent to the nozzle corresponding to the target sorting bin, and the actual time of airflow formation is detected. The time difference between the trigger signal issuance time and the actual airflow formation time is taken as the nozzle response delay. After measurement, the data is pre-stored in the grading module; drift time Based on the quality grade of the ginseng, the parameters are read from the pre-stored pneumatic parameter calibration table.

[0135] The trigger lead time is determined according to the following relationship:

[0136] ,in The trigger lead time is measured in seconds (s). The drift time is expressed in seconds. This represents the time correction amount corresponding to the target sorting warehouse, in seconds. When no correction occurs... ;

[0137] The remaining transport time before the ginseng reaches the target sorting warehouse. Reaching the trigger lead time At that time, the grading module triggers the air nozzle corresponding to the target sorting bin. The instruction waiting time from when the ginseng passes the trigger reference position to when the trigger air nozzle is triggered is: ;

[0138] in The instruction wait time is expressed in seconds (s). At that time, after the ginseng passes the trigger reference position, the air nozzle is triggered after the specified instruction waiting time; when The air nozzle is triggered when the ginseng passes the trigger reference position;

[0139] The reason for using the above relationship is that the air nozzle needs to go through an air nozzle response delay after receiving the trigger signal. Only then can an airflow be formed, and ginseng needs time to drift after being affected by the airflow. Only then can it enter the target sorting warehouse, time correction amount Used to compensate for the landing point deviation caused by the corresponding air nozzle in the target sorting bin; the air nozzle is triggered when the remaining conveying time is consistent with the sum of the air nozzle response delay, drift time and time correction amount, so that the timing of airflow formation, ginseng drift and ginseng arrival in the target sorting bin can be matched.

[0140] In S902, the quality of ginseng is compared with the average quality of ginseng in the same batch, and the drift time and blowing pressure are determined according to the comparison results.

[0141] At the start of a batch, the pre-stored average quality of ginseng of the same variety is used as the initial average batch quality. After each ginseng is weighed, the average batch quality is updated on a rolling basis based on the quality of the ginseng that has been weighed.

[0142] Meanwhile, the drifting process of ginseng under the influence of airflow is related to its mass: lighter ginseng has less inertia, is significantly accelerated by airflow, and drifts for a large distance and takes a long time; heavier ginseng has greater inertia, accelerates slowly under the same airflow, and drifts for a small distance and takes a short time. If the trigger lead is calculated based on a uniform drift time, lighter ginseng will fall in front of the sorting bin earlier due to its longer actual drift time, while heavier ginseng will fall behind the sorting bin due to its shorter actual drift time. Correspondingly, heavier ginseng requires a larger blowing pressure to be blown away from the conveying path, while lighter ginseng can be blown away with a smaller blowing pressure. Excessive blowing pressure may cause it to be impacted, rolled over, or even have its roots damaged.

[0143] Specifically, when the mass m of ginseng is less than the average mass of ginseng in the same batch, the drift time is taken as the first drift time. The blowing pressure adopts the first blowing pressure. When the ginseng's mass is greater than or equal to the average batch mass, the drift time is taken as the second drift time. The blowing pressure adopts the second blowing pressure. .

[0144] Because the quality distribution of ginseng varies significantly across different producing areas and harvesting periods, using fixed boundaries could result in entire batches of ginseng falling into the same category, rendering grading meaningless. However, the batch average is automatically determined by the measured quality of the ginseng in that batch, adapting to different batches and ensuring that each batch of ginseng is naturally divided into two groups based on weight. , , , The specific value can be obtained through calibration tests based on the air pressure conditions of the nozzle and the mass range of ginseng.

[0145] During the equipment commissioning phase, the nozzle response delay and drift time of ginseng samples at the first and second mass levels were measured under different blowing pressures. The drift time and blowing pressure corresponding to each mass level were stored as a pneumatic parameter calibration table.

[0146] In S903, when ginseng falls into the sorting bin, the actual landing point of the ginseng is detected by photoelectric sensors installed on both sides of the sorting bin entrance. The distance between the actual landing point and the center of the target sorting bin along the conveying direction is determined as the deviation amount, and the direction in which the actual landing point is forward or backward relative to the center of the target sorting bin is determined as the deviation direction.

[0147] Specifically, the photoelectric sensor includes a light emitting end and a light receiving end arranged opposite each other. The light emitted by the light emitting end is received by the light receiving end. When an object passes between them, the light path is blocked, and the output signal of the light receiving end changes accordingly. Multiple pairs of photoelectric sensors are arranged at intervals on both sides of the sorting bin entrance along the conveying direction. When ginseng passes through the sorting bin entrance, it blocks the light path of one pair of photoelectric sensors. The actual landing point of the ginseng is determined based on the location of the photoelectric sensor whose light path is blocked. The center of the target sorting bin can be predetermined according to the structural dimensions of the sorting bin. The distance between the actual landing point and the center of the target sorting bin is the deviation P, in meters.

[0148] A landing point slightly ahead indicates an earlier triggering time, while a landing point slightly behind indicates a later triggering time. Therefore, the deviation reflects the magnitude of the error in the triggering timing, and the direction of the deviation reflects the timing of the triggering.

[0149] In S904, when the deviation direction of the same sorting bin occurs continuously in the same direction, the adjustment amount is determined according to the ratio of the deviation amount to the conveying speed. The correction amount of the sorting bin is cumulatively adjusted in the direction opposite to the deviation direction, and the trigger advance amount corresponding to the sorting bin is corrected with the adjusted correction amount. The trigger advance amount corresponding to the other sorting bins remains unchanged.

[0150] It should be noted that systematic deviations are caused by individual differences in each air nozzle, and they occur repeatedly in the same sorting bin with a fixed direction; while deviations caused by individual differences in ginseng posture occur randomly with uncertain direction and are considered sporadic deviations. If they are not corrected indiscriminately, sporadic deviations will cause frequent changes in the trigger advance, which will reduce the sorting stability.

[0151] Using "continuous same direction" as the adjustment condition can distinguish between systematic deviations and occasional deviations: occasional deviations are random in direction and will not occur continuously in the same direction, so they will not trigger adjustments, thereby avoiding interference from individual differences in ginseng in the correction process.

[0152] During adjustment, the adjustment amount is determined based on the ratio of the deviation to the conveying speed:

[0153] Where is the single adjustment amount in seconds; P is the deviation in meters; and v is the conveying speed of the conveying device in meters per second. It should be noted that the landing point deviation essentially reflects the timing of the triggering: the time required for the ginseng to travel the distance corresponding to the deviation at the conveying speed is the time when the triggering timing is too early or too late. Therefore, dividing the deviation by the conveying speed can convert the spatial deviation into the time error of the triggering timing, making the dimensions of the adjustment amount and the trigger advance amount consistent.

[0154] The correction amount is the cumulative value of all previous adjustments, determined as follows: if the actual landing point is later, it indicates the trigger time is too late, requiring an increase in the trigger advance, and the adjustment amount is positive; if the actual landing point is earlier, it indicates the trigger time is too early, requiring a decrease in the trigger advance, and the adjustment amount is negative; the correction amount ΔM is the cumulative value of all previous adjustments, both positive and negative, and is zero when no adjustment has occurred; the corrected trigger advance is the sum of the original trigger advance and the correction amount. The sum of the results. Since the systematic deviation is caused by the individual differences of each air nozzle, the deviation of different sorting bins is different. Therefore, the correction amount of each sorting bin is adjusted and takes effect independently. The correction of one sorting bin does not affect other sorting bins.

[0155] In S905, the time correction amount corresponding to each target nozzle is accumulated and takes effect independently. The correction of one sorting bin does not affect other sorting bins. After correction, the actual landing point of the subsequent ginseng in the sorting bin is processed according to the detection results of the photoelectric sensor: when the corrected actual landing point returns to the allowable deviation range of the center of the target sorting bin entrance, it indicates that the correction amount is consistent with the systematic deviation, and the current time correction amount remains unchanged; when the deviation direction of the subsequent landing point is opposite to the cumulative correction direction, it indicates that the cumulative correction amount has exceeded the actual requirement, and the time correction amount is backed back by one correction step to avoid the correction amount changing repeatedly between the two directions, so that the correction process converges.

[0156] In a preferred configuration, the air nozzle is positioned on one side of the conveying device, and the sorting bin is positioned on the opposite side of the conveying device. The air nozzle is angled downwards toward the corresponding sorting bin, allowing the ginseng to acquire an initial downward velocity. After being acted upon by the airflow, the ginseng falls into the sorting bin along a parabolic trajectory, resulting in a low and flat flight path and a concentrated landing point. Specifically, the angle between the airflow direction and the horizontal plane can be set to 30° to 60°. This angle range can be determined experimentally based on the height of the conveying device and the distance between the air nozzle and the sorting bin.

[0157] The sorting bin entrance is equipped with a guide surface that slopes inwards. The surface of the guide surface is covered with a buffer layer, which can be made of materials with elastic deformation capabilities, such as silicone or sponge. After the ginseng falls into the sorting bin, it slides along the guide surface and is caught by the buffer layer. In this way, the guide surface prevents the ginseng from being stuck at the entrance of the sorting bin, and the buffer layer absorbs the impact when the ginseng falls in, preventing the ginseng, especially the rootlets, from rigidly colliding with the sorting bin and breaking, thus ensuring that the sorting process is damage-free.

[0158] A ginseng quality intelligent identification system, which is used to implement the above-mentioned ginseng quality intelligent identification method, includes a conveying device, multiple contoured trays spaced apart on the conveying device, a three-dimensional morphology acquisition device, a spectral acquisition device and a weighing device arranged sequentially along the conveying path of the conveying device, a grading processing module and a pneumatic sorting module.

[0159] The conformal tray is used to support the whole ginseng and keep it in a fixed position; the three-dimensional morphology acquisition device is used to acquire the three-dimensional point cloud of the ginseng and extract morphological parameters; the spectral acquisition device is used to acquire the near-infrared spectrum of the ginseng and obtain spectral characteristics; the weighing device is used to obtain the mass of the ginseng; the grading processing module is used to fuse the morphological parameters, spectral characteristics and mass to determine the final quality grade of the ginseng; the pneumatic sorting module is used to blow the ginseng into the corresponding sorting bin according to the quality grade, and the pneumatic sorting module includes multiple air nozzles that correspond one-to-one with multiple sorting bins.

[0160] In summary: By using a contoured tray to maintain a fixed posture for ginseng during transport and allowing the fibrous roots to hang naturally, a unified acquisition benchmark is provided for multi-view point cloud stitching, occlusion completion, and part segmentation. This enables the automatic quantification of appearance indicators such as the number of reed bowls, ring density, and fibrous root integrity, and reduces fibrous root damage during transport and sorting. Furthermore, by using point cloud-guided near-infrared spectroscopy to sample at fixed points in the middle section of the main root and normalizing the equivalent optical path, the interference of differences in ginseng posture, thickness, and epidermal condition on saponin detection is eliminated.

[0161] The system integrates morphology, spectrum, meat firmness, and quality to differentiate and grade ginseng by specifications, taking into account both appearance and internal composition. The grading rules are clear and unambiguous. The grading results are directly mapped to the sorting bin and generate instructions. The triggering timing and blowing pressure are adaptively adjusted according to the quality of ginseng. The landing point deviation is corrected independently in a closed loop by the sorting bin. Identification, grading, and sorting form a data closed loop. The accuracy of grading and the long-term operational stability of the equipment are improved in tandem. The identification data can also be stored as a traceable quality file.

[0162] Those skilled in the art will understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention can be combined and / or combined in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or combinations fall within the scope of the present invention.

[0163] The embodiments of the present invention have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of the invention. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of the invention, and all such substitutions and modifications should fall within the scope of the invention.

Claims

1. A method for intelligent identification of ginseng quality, characterized in that: Includes the following steps: S1. Place the whole ginseng in a contoured tray. The contoured tray is provided with a rhizome support area, a main root support area, and a fibrous root suspension avoidance groove, so that the ginseng maintains a fixed posture and the fibrous roots hang naturally during the transportation process. The contoured tray is transported by the conveying device through the three-dimensional morphology acquisition station, the spectrum acquisition station, the weighing station, and the sorting station in sequence. The sorting station is provided with at least two sorting bins. S2. Collect three-dimensional point cloud of ginseng under fixed posture, extract morphological parameters from the three-dimensional point cloud, the morphological parameters include at least the number of reed bowls, the main body size, the ring density and the integrity of the fibrous roots, and determine the spectral detection area of ​​the middle section of the main root of ginseng based on the three-dimensional point cloud. S3. Collect near-infrared spectra of the corresponding region of ginseng according to the location of the spectral detection area to obtain spectral characteristics characterizing saponin content; S4. Obtain the mass of the ginseng and calculate the volume of the ginseng using three-dimensional point computing. Obtain the density of the ginseng from the mass and volume, and determine the firmness of the ginseng flesh based on the density and ring density. S5. The morphological parameters, spectral characteristics, flesh firmness, and quality are fused to determine the final quality grade of the ginseng. Based on the final quality grade, the air nozzles in the pneumatic sorting module corresponding to the target sorting bin are controlled to blow the ginseng into the target sorting bin. The triggering time of the air nozzles is determined based on the conveying relationship between the weighing station and the target sorting bin, the air nozzle response delay, and the drift time corresponding to the ginseng quality. The blowing pressure is determined based on the ginseng quality.

2. The intelligent identification method for ginseng quality according to claim 1, characterized in that: The conformal support box also includes a tail support area, and the support depths of the rhizome support area, the main root support area, and the tail support area are respectively adapted to the radial dimensions of the corresponding parts of the ginseng; the fibrous root hanging avoidance groove is provided on at least one side of the conformal support box to allow the ginseng fibrous roots to hang down naturally; the conformal support box is provided with an elastic liner, which is used to restrict the rolling and turning of the ginseng during the transportation process.

3. The intelligent identification method for ginseng quality according to claim 1, characterized in that: Step S2 specifically involves: S231. At the three-dimensional morphology acquisition station, multi-view point cloud data of ginseng are simultaneously acquired by at least three depth cameras located above the contour tray and on both sides of the conveying path. S232. The point clouds collected by each depth camera are converted to the same coordinate reference and stitched together. The main axis of the ginseng is fitted according to the main point cloud. For the missing area of ​​the point cloud on the side of the ginseng that is attached to the contour box, the radial distance of the collected point cloud relative to the main axis of the ginseng is mapped to the missing area in the same axial section to obtain the completed three-dimensional point cloud. S233. Segment the completed 3D point cloud to obtain point cloud partitions corresponding to the rhizome, main body and fibrous roots respectively; S234. Count the number of concave areas in the reed head point cloud partition to obtain the number of reed bowls; obtain the axial length and maximum radial dimension of the main body point cloud partition to obtain the main body dimension; divide the main body point cloud partition into three equal segments along the axial direction, count the number of annular protrusions around the main body in each segment, and take the ratio of the number of annular protrusions in each segment to the axial length of the corresponding segment as the annular density of that segment. S235. Based on the end morphology of the hair roots in the hair root point cloud partition, the ratio of the number of complete hair roots at the end to the total number of hair roots is calculated to obtain the hair root integrity. The total number of hair roots is the sum of the number of identified complete hair roots at the end, truncated hair roots at the end, and hair root remnants. The hair root integrity is used as a morphological parameter in the fusion and grading of S5. The truncated hair roots at the end and the hair root remnants are identified based on the skeleton line length and radial dimension of the end cross-section of each hair root branch in the hair root point cloud partition.

4. The intelligent identification method for ginseng quality according to claim 3, characterized in that: Step S3 specifically involves: At the spectral acquisition station, the shaped tray is positioned by a positioning mechanism to keep the ginseng stationary during the spectral acquisition process; the middle third section of the main root of the ginseng is determined based on the completed three-dimensional point cloud; the position of the fibrous root protrusion is identified based on the three-dimensional point cloud; and the position of epidermal damage is identified based on the epidermal image registered with the three-dimensional point cloud acquired by the image acquisition device. The surface area of ​​the main root that does not contain the fibrous root protrusion position and the epidermal damage position is determined as the spectral detection area. Multiple detection points are determined within the spectral detection area. The spectral probe is moved sequentially to each detection point and near-infrared spectra are collected respectively. The absorption intensity of each detection point in the pre-calibrated saponin characteristic band is normalized according to the equivalent optical path at the corresponding detection point. The normalized absorption intensities are then averaged to obtain the spectral characteristics characterizing the relative content level of saponins.

5. The intelligent identification method for ginseng quality according to claim 1, characterized in that: Step S4 specifically involves: When the ginseng passes through the dynamic weighing sensor set on the conveying path along the contour tray, multiple mass values ​​are continuously sampled and the median value is taken. Then, the tare weight of the corresponding contour tray that is pre-stored is deducted to obtain the mass of the ginseng. The completed 3D point cloud is reconstructed using a closed surface, and the volume of the ginseng is calculated. The density of the ginseng is determined based on its mass and volume. The average value of the density of the three rings is calculated, and the ginseng density and the average ring density are converted into density score and ring score respectively according to the same normalization rule. The average value of the density score and the ring score is used as the firmness score of the ginseng flesh.

6. The intelligent identification method for ginseng quality according to claim 5, characterized in that: In step S5, the morphological parameters, spectral characteristics, flesh firmness, and quality are fused together, specifically as follows: S561. Compare the maximum radial dimension of the ginseng body with the preset specification limit. If the maximum radial dimension is greater than the specification limit, the ginseng is classified into the large specification category; if the maximum radial dimension is less than or equal to the specification limit, the ginseng is classified into the small specification category. S562. Convert the integrity of the rootlets, spectral characteristics, and quality into dimensionless scores according to the pre-stored normalization upper and lower limits, and use the determined flesh firmness score as the dimensionless score corresponding to the flesh firmness; weight and fuse each dimensionless score according to the fusion weight corresponding to the ginseng's specification category to obtain the fusion score; among them, in the fusion weight corresponding to the large specification category, the weight of the rootlet integrity is greater than the weight of the spectral characteristics; in the fusion weight corresponding to the small specification category, the weight of the spectral characteristics is greater than the weight of the rootlet integrity. S563. Determine the initial quality grade of ginseng based on the fusion score and the pre-stored fusion score grade threshold. S564. Verify the initial quality level based on the integrity of the hair roots to determine the final quality level; wherein, when the integrity of the hair roots is lower than a preset integrity threshold, the initial quality level is downgraded by one level to obtain the final quality level; when the integrity of the hair roots is not lower than the preset integrity threshold, the initial quality level is determined as the final quality level.

7. The intelligent identification method for ginseng quality according to claim 6, characterized in that: Step S5, which controls the air nozzles in the pneumatic sorting module corresponding to the target sorting bin according to the final quality level, includes: S571. After obtaining the final quality grade of ginseng, call the pre-stored correspondence between the final quality grade and the sorting warehouse, determine the target sorting warehouse corresponding to the final quality grade, and assign the high quality grade to the first sorting warehouse, the medium quality grade to the second sorting warehouse, and the low quality grade to the third sorting warehouse. S572. Determine the target sorting bin for ginseng based on the final quality grade of ginseng and the corresponding relationship, and generate a sorting instruction carrying the target sorting bin information; S573. Based on the conveying relationship between the ginseng-containing contour tray and the target sorting bin along the conveying path, determine the execution time of the sorting instruction, and when the execution time is reached, send the sorting instruction to the pneumatic sorting module, which will then blow the ginseng into the target sorting bin.

8. The intelligent identification method for ginseng quality according to claim 6, characterized in that: After determining the final quality grade of ginseng in step S5, the process also includes establishing and tracing quality records, specifically: The final quality grade, morphological parameters, spectral characteristics, flesh firmness and quality of ginseng are linked and stored with the number of the molded tray and the batch to which it belongs, forming a quality file for each ginseng. The system calculates the percentage of ginseng that has been graded in the same batch, with the final quality grades being high, medium, and low. When the number of ginseng that has been graded within the statistical period reaches the minimum number of samples stored in advance, and the percentage of low-quality ginseng increases sequentially within at least two consecutive statistical periods, a batch quality warning is generated. It also retrieves the quality records of ginseng that has been graded in the same batch based on batch information, realizing full traceability of a single ginseng from self-testing and grading to warehousing.

9. The intelligent identification method for ginseng quality according to claim 7, characterized in that: The pneumatic sorting module blows ginseng into the target sorting bin, and the specific execution process includes the following steps: S901. Using the weighing station exit as the trigger reference position, determine the remaining conveying time for the ginseng to move from the trigger reference position to the target sorting warehouse based on the distance along the conveying path between the trigger reference position and the target sorting warehouse and the conveying speed of the conveying device. The sum of the response delay of the air nozzle corresponding to the target sorting bin, the drift time corresponding to the ginseng mass, and the time correction amount corresponding to the target sorting bin is used as the trigger advance amount. When the remaining conveying time reaches the trigger advance amount, the air nozzle corresponding to the target sorting bin is triggered. S902. Compare the quality of the ginseng with the average quality of the same batch of ginseng. If the quality of the ginseng is less than the average quality, the ginseng is determined to be in the first quality grade; otherwise, the ginseng is determined to be in the second quality grade. Read the corresponding drift time and blowing pressure from the pre-stored pneumatic parameter calibration table according to the quality grade. S903. A photoelectric sensor is installed at the entrance of the sorting bin to detect the actual landing point of the ginseng relative to the center of the target sorting bin entrance, and the correction direction is determined based on the landing point deviation between the actual landing point and the center of the target sorting bin entrance. S904. When the ginseng landing point deviation corresponding to the same target nozzle is too far back in at least two consecutive sorting processes, the time correction amount corresponding to the target nozzle is increased by a preset correction step size; when the ginseng landing point deviation corresponding to the same target nozzle is too far forward in at least two consecutive sorting processes, the time correction amount corresponding to the target nozzle is decreased by a preset correction step size. S905. The time correction amount corresponding to each target air nozzle is accumulated independently; when the corrected actual landing point returns to the allowable deviation range of the target sorting warehouse entrance center, the current time correction amount is maintained. When the direction of subsequent landing point deviation is opposite to the direction of cumulative correction, the time correction amount is backed back to the initial value by one correction step.

10. A ginseng quality intelligent identification system, wherein the system implements the ginseng quality intelligent identification method according to any one of claims 1 to 9, characterized in that: It includes a conveying device, at least two contoured trays spaced apart on the conveying device, a three-dimensional shape acquisition device, an image acquisition device, a spectral acquisition device and a weighing device arranged sequentially along the conveying path of the conveying device, a positioning mechanism set at the spectral acquisition station, a grading processing module, a pneumatic sorting module, at least two pairs of photoelectric sensors set at the entrance of each sorting bin and a storage module. The contoured tray is used to support the entire ginseng and keep it in a fixed position. The three-dimensional morphology acquisition device is used to acquire three-dimensional point clouds of ginseng and extract morphological parameters; the image acquisition device is used to acquire ginseng epidermal images registered with the three-dimensional point clouds; the positioning mechanism is used to limit the shape-following tray during the spectral acquisition process; the spectral acquisition device is used to acquire the near-infrared spectrum of ginseng and obtain spectral characteristics; the weighing device is used to obtain the mass of ginseng. The grading module is used to determine the ginseng density based on its mass and volume, and to determine the flesh firmness based on the ginseng density and ring density. It also integrates the morphological parameters, spectral characteristics, flesh firmness, and mass to determine the final quality grade of the ginseng. The pneumatic sorting module includes air nozzles corresponding to the sorting bins, used to blow the ginseng into the corresponding sorting bins according to the final quality grade. The photoelectric sensor is used to detect the actual landing point of the ginseng. The storage module stores the tartrate weight of the conformal tray, saponin characteristic bands, normalized upper and lower limits, pneumatic parameter calibration tables, and quality records.

Images