A customs port screening system and method

Abstract

The application discloses a customs port screening system and method, comprising a feeding module for feeding grain samples into a screening module, a material screening module for preliminarily screening the samples based on a four-stage screening technology, separating materials with different particle sizes and densities, an identification and sample retention module for collecting sample images and performing species identification and saving voucher specimens through a YOLO neural network technology, and a data analysis module for performing data analysis on screened data and classifying, uploading and updating the database; the application realizes high-precision sorting of the samples by using a multi-stage screening technology and a multi-channel physical sorting technology, controls an intelligent mechanical arm to complete retention labeling of different samples by using an image recognition technology, and finally realizes full-automatic processing of the entry grain samples.

Description

Technical Field

[0001] This invention relates to the field of customs supervision and material sorting technology, and in particular to a customs port screening system and method. Background Technology

[0002] With increasingly frequent international trade, customs checkpoints, as important channels for the entry and exit of goods, are particularly crucial in controlling the quality of imported grains and cereals. In order to ensure the quality and safety of grains during storage and import, it is necessary to accurately sort and extract grain samples.

[0003] The existing customs technology for sorting and extracting grain samples has improved the efficiency and accuracy of sorting and extracting grain samples, which can, to a certain extent, ensure food security and improve customs supervision efficiency. However, it lacks a multi-system collaborative sorting scheme to complete the entire process of automatic processing of a batch of original samples to be inspected, from initial screening to sample classification, fine impurity filtration, automatic identification and storage of sieved samples, and finally data uploading to the platform and risk prediction. A large amount of manual operation is still required to inspect the current customs entry samples, which is troublesome and time-consuming. Therefore, this invention proposes a customs port screening system and method to solve the problems existing in the prior art. Summary of the Invention

[0004] To address the aforementioned problems, the present invention aims to propose a customs port screening system and method. This system and method utilizes multi-level screening technology and multi-stage physical sorting technology to achieve high-precision sorting of samples, and uses image recognition technology to control an intelligent robotic arm to complete the retention and labeling of different samples, ultimately achieving fully automated processing of imported grain samples.

[0005] To achieve the objectives of this invention, the invention is implemented through the following technical solution: a customs port screening system, comprising a feeding module, a material screening module, an identification and sampling module, and a data analysis module. The feeding module is used to feed grain samples into the screening module. The material screening module performs preliminary screening of the samples based on four-stage sieving technology, separating materials with different particle sizes and densities. The identification and sampling module is used to acquire sample images and perform species-level identification and preserve voucher specimens using YOLO neural network technology. The data analysis module is used to perform data analysis on the screened data and perform classification, uploading, and database updates.

[0006] A further improvement is that the feeding module includes a main conveyor belt feeding submodule, an auxiliary conveyor belt circulation submodule, and a waste and impurity disposal submodule. The main conveyor belt feeding submodule is used to transport the sample from the inlet to the screening area. The auxiliary conveyor belt circulation submodule is used to ensure the continuous flow of the sample during the screening process. The waste and impurity disposal submodule is used to dispose of the screened waste and impurities.

[0007] Further improvements are made in that: the material screening module includes a four-stage molecular screening module, a physical sorting sub-module, and a color sorting extraction sub-module. The four-stage molecular screening module uses four layers of sieves with different particle sizes to perform four-stage screening based on a size particle size screener. The physical sorting sub-module further sorts the material based on physical methods such as centrifugal screening, vibration negative pressure suction, acoustic standing wave airflow chromatography, and electrostatic adsorption cleaning. The color sorting extraction sub-module uses a color sorter to screen and extract pure grain particles based on color differences.

[0008] Further improvements are made in the following aspects: The physical sorting submodule includes a centrifugal screening unit, a vibration negative pressure suction unit, an acoustic standing wave airflow chromatography unit, and an electrostatic adsorption cleaning unit. The centrifugal screening unit uses centrifugal force to separate particles of different sizes from the oversize material after the three-layer and four-layer sieving of the four-stage sieve molecular module. The vibration negative pressure suction unit uses mechanical vibration and negative pressure suction technology to further separate and collect particles of different densities. The acoustic standing wave airflow chromatography unit uses acoustic standing waves of different frequencies and airflow chromatography technology to accurately separate particles of different sizes again. The electrostatic adsorption cleaning unit uses an electrostatic generator to generate electrostatic adsorption to remove impurities from the sample surface.

[0009] A further improvement is that the identification and retention module includes a sample image acquisition submodule and a neural network recognition submodule. The sample image acquisition submodule acquires image information of the sample based on a high-resolution camera, and the neural network recognition submodule identifies the sample targets in the image based on the YOLO neural network, identifies the species level, and then labels and saves the data.

[0010] Further improvements are made in that: the data analysis module includes a data classification submodule, a PTZ data upload submodule, and a database update submodule. The data classification submodule preprocesses the collected data and classifies the sample data and impurity data according to the identification results. The PTZ data upload submodule is used to upload the classified data to the PTZ. The database update submodule is used to update the sample database and impurity database according to the classification results.

[0011] A further improvement is that the data classification submodule includes a grain sample data storage unit and an impurity sample data storage unit. The grain sample data storage unit is used to store the screened grain sample data, and the impurity sample data storage unit is used to store the separated and screened impurity sample data.

[0012] A further improvement is that the database update submodule includes an impurity sample database update unit and a valid sample database update unit. The impurity sample database update unit is used to store impurity sample data identified outside the original database, and the valid sample database update unit is used to store valid sample data identified outside the original database.

[0013] A screening method for a customs port screening system includes the following steps: Step 1: Four-stage screening process. The sample to be screened is transported to the four-stage screening module through the feeding module. It is screened through four layers of sieves with different particle sizes to screen out the straw and screen out and retain samples of large weeds and pests. Step 2: Centrifugal screening. The material remaining after the three-layer and four-layer screening in Step 1 is sent to the centrifugal screening unit and slowly centrifuged at a speed of 10-100 rpm to obtain the near-center target, the middle target, and the edge target. Step 3: Vibration and negative pressure suction treatment. The targets on the edge of the turntable after centrifugation and screening in Step 2 are mechanically vibrated at different frequencies and amplitudes and collected using different suction forces to obtain high-frequency, medium-frequency and low-frequency targets. Step 4: Acoustic standing wave airflow tomography. For each of the three types of targets extracted after vibration negative pressure absorption in Step 3, resonant screening is performed using sound waves of different frequencies. At the peak position of the standing wave formed by the sound waves, jets of different speeds are collected using air jet pipes to obtain high-frequency acoustic wave screened materials, medium-frequency acoustic wave screened materials, and low-frequency acoustic wave screened materials. Step 5: Electrostatic adsorption cleaning treatment. The multiple groups of materials obtained in steps 1 to 4 are laid flat on the conveyor belt. Each group of materials is electrostatically adsorbed and cleaned by an energized electrostatic generator to remove tiny impurities from the surface. After cleaning, the materials are identified and retained by the identification and retention module to obtain the impurity-removed materials. Step 6: Color sorting and extraction processing. The processed materials are separated according to their different characteristics using a color sorter. Finally, image recognition is performed to complete the screening and sorting of the materials.

[0014] The beneficial effects of this invention are as follows: Through a multi-stage sorting process, from coarse sorting to fine sorting, this invention gradually removes various impurities such as weed seeds, straw, and pests from imported grains and forages, effectively ensuring the purity of imported grains and forages and preventing harmful substances from entering the country with the goods and posing a potential threat to the domestic ecological environment and agricultural production. Simultaneously, the system also has sample extraction and preservation functions, enabling the accurate extraction of representative grain and forage samples, providing strong evidence for subsequent quality testing and quarantine analysis, and assisting customs authorities in carrying out efficient and accurate supervision work, thus safeguarding national food security and ecological security. Attached Figure Description

[0015] Figure 1 This is a system architecture diagram of the present invention.

[0016] Figure 2 This is a schematic diagram of the system flow of the present invention.

[0017] Figure 3 This is a schematic diagram of the four-stage screening principle of the present invention.

[0018] Figure 4 This is a schematic diagram of the YOLO neural network of this invention.

[0019] Figure 5 This is a schematic diagram illustrating the data analysis principle of the present invention.

[0020] Figure 6 This is a schematic diagram illustrating the centrifugal screening principle of the present invention.

[0021] Figure 7 This is a schematic diagram illustrating the principle of vibration negative pressure absorption in this invention.

[0022] Figure 8 This is a schematic diagram illustrating the principle of acoustic standing wave airflow tomography in this invention.

[0023] Figure 9 This is a schematic diagram illustrating the electrostatic adsorption cleaning principle of the present invention. Detailed Implementation

[0024] To enhance understanding of the present invention, the present invention will be further described in detail below with reference to embodiments. These embodiments are only used to explain the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0025] Example 1 according to Figures 1-5 As shown, this embodiment provides a customs port screening system, including a feeding module, a material screening module, an identification and sampling module, and a data analysis module.

[0026] The feeding module is used to feed grain samples into the screening module. It includes a main conveyor belt feeding submodule, an auxiliary conveyor belt circulation submodule, and a waste and impurity disposal submodule. The main conveyor belt feeding submodule is used to transport the sample from the inlet to the screening area. The auxiliary conveyor belt circulation submodule is used to ensure the continuous flow of the sample during the screening process. The waste and impurity disposal submodule is used to dispose of the waste and impurities screened out.

[0027] The material screening module performs preliminary screening of samples based on four-stage sieving technology, separating materials with different particle sizes and densities. It includes a four-stage sieving molecular module, a physical sorting sub-module, and a color sorting extraction sub-module. The four-stage sieving molecular module uses a scale particle size screen to perform four-stage sieving using four layers of sieves with different particle sizes. The physical sorting sub-module further sorts the materials using physical methods such as centrifugal screening, vibration negative pressure suction, acoustic standing wave airflow chromatography, and electrostatic adsorption cleaning. The color sorting extraction sub-module uses a color sorter to extract pure grain particles by screening based on color differences.

[0028] Four-stage sieving module: The sample enters the size distribution sieve through the inlet. A rotating rake spreads the sample evenly, ensuring that every particle has a chance to pass through the sieve openings, improving sieving efficiency. The size distribution sieve then uses four layers of sieves with different particle sizes for four-stage sieving, removing straw and retaining large weeds and pests. The sieve uses vibration technology, causing the spread sample to vibrate in a jumping motion on the sieve mesh. During this movement, particles smaller than the sieve mesh size fall off, while particles larger than the mesh size remain on the mesh. The four layers of sieve mesh sizes decrease from top to bottom. After sieving, the material from each layer is conveyed out of the sieve by a conveyor device, resulting in four stages of sieving: Primary screening: The material that passes through the screen consists of straw, large weeds, and pests and diseases; the material that passes through the screen enters the secondary screening. Secondary screening: The material that passes through the sieve is whole beans and weed seeds that are 5mm or larger; the material that passes through the sieve enters the tertiary screening. Third-stage screening: The material that passes through the sieve is incomplete beans and weed seeds with a size of 3mm-5mm, and the material that passes through the sieve enters the fourth-stage screening; Four-stage screening: The material on the screen consists of weed seeds and impurities of 0.5mm-3mm, while the material under the screen enters the waste conveying module.

[0029] This multi-stage sieving process effectively removes most impurities, improves sample purity, and provides high-quality input for subsequent steps such as centrifugation and mechanical vibration aspiration. A schematic diagram of the four-stage sieving process is attached to the instruction manual. Figure 3 As shown.

[0030] Color sorting extraction submodule: The oversize material after secondary particle size sieving is arranged in a row as a sample and enters the color sorter to separate soybeans from other impurities. The impurities and the oversize material from the third and fourth sieves are then centrifuged and processed for image recognition. The color sorter separates and screens materials according to their different characteristics through optical illumination, and has the advantages of high sorting efficiency, accurate grading, good product quality, and low damage rate.

[0031] Color sorting is a crucial step in automated sorting systems, used to separate soybeans from other impurities, ensuring their purity and quality. In color sorting, the oversize material from the secondary particle size distribution is lined up as a sample and enters the color sorter. The color sorter uses optical equipment to separate and screen materials based on their different characteristics. Through high-precision optical sensors and image processing technology, the color sorter analyzes the color, shape, and texture of each particle, effectively separating soybeans from other impurities. The separated impurities, along with the oversize material from the tertiary and quaternary sieves, undergo subsequent processing steps such as centrifugation, screening, and image recognition. Color sorters offer advantages such as high sorting efficiency, accurate grading, good product quality, and low damage rate, effectively improving sorting accuracy and efficiency. This color sorting process ensures the purity and quality of soybeans, providing high-quality samples for customs supervision and subsequent processing.

[0032] The physical sorting submodule includes a centrifugal screening unit, a vibratory negative pressure suction unit, an acoustic standing wave airflow chromatography unit, and an electrostatic adsorption cleaning unit. The centrifugal screening unit uses centrifugal force to separate particles of different sizes from the oversize material after the three- and four-layer sieving of the four-stage sieve molecular module. The vibratory negative pressure suction unit uses mechanical vibration and negative pressure suction technology to further separate and collect particles of different densities. The acoustic standing wave airflow chromatography unit uses acoustic standing waves of different frequencies and airflow chromatography technology to accurately separate particles of different sizes again. The electrostatic adsorption cleaning unit uses an electrostatic generator to generate electrostatic adsorption to remove impurities from the sample surface.

[0033] The identification and retention module is used to acquire sample images and perform species-level identification and preservation of credential specimens using YOLO neural network technology. It includes a sample image acquisition submodule and a neural network recognition submodule. The sample image acquisition submodule acquires image information of the sample based on a high-resolution camera, and the neural network recognition submodule identifies the sample targets in the image based on the YOLO neural network, identifies the species level, and then labels and saves the data.

[0034] Specifically, after electrostatic adsorption cleaning, multiple sets of target samples and primary sieve residues are laid flat on a target tray. Artificial intelligence target detection algorithms are used to identify the weed seeds in the samples at the species level. A robotic arm then aspirates and dispenses the sample, labels it, and preserves it. The target detection algorithm can accurately identify and classify different types of weed seeds, ensuring the accuracy and integrity of the samples.

[0035] Image recognition and credential specimen retention are key steps in automated sorting systems, used for species-level identification of weed seeds in samples and for preserving credential specimens. During image recognition, multiple sets of target samples and primary sieve residues, after electrostatic adsorption cleaning, are laid flat on a target tray, and images of the samples are captured using a high-resolution camera. These images are then transmitted to an artificial intelligence target detection algorithm, which uses a deep learning model to perform species-level identification of weed seeds in the samples. The target detection algorithm can accurately identify and classify different types of weed seeds, ensuring the accuracy and integrity of each sample. After identification, the samples are picked up and dispensed by a robotic arm. Each sample is labeled and stored in a dedicated storage container for subsequent analysis and verification.

[0036] This image recognition and credential specimen retention method ensures that each sample has a detailed record and backup, providing strong support for customs supervision and scientific research. The YOLO neural network principle used for image recognition is as shown in the attached instruction manual. Figure 4 As shown.

[0037] The data analysis module is used to analyze, classify, upload, and update the database of the filtered data. It includes a data classification submodule, a PTZ data upload submodule, and a database update submodule. The data classification submodule preprocesses the collected data and classifies the sample data and impurity data according to the identification results. The PTZ data upload submodule is used to upload the classified data to the PTZ. The database update submodule is used to update the sample database and impurity database according to the classification results.

[0038] The data classification submodule includes a grain sample data storage unit and an impurity sample data storage unit. The grain sample data storage unit is used to store the screened grain sample data, and the impurity sample data storage unit is used to store the separated and screened impurity sample data.

[0039] The database update submodule includes an impurity sample database update unit and a valid sample database update unit. The impurity sample database update unit is used to store impurity sample data identified outside the original database, and the valid sample database update unit is used to store valid sample data identified outside the original database.

[0040] The detailed workflow of the data analysis module is as follows: (1) Data preprocessing: Before the processing results of the identification module are transmitted to the data analysis module, the data needs to be preprocessed, including data cleaning, format conversion and standardization, to ensure the quality and consistency of the data.

[0041] (2) Sample and impurity classification: The preprocessed results will be transmitted to the data analysis module via a file transfer protocol system, and the sample and impurity data will be classified according to the identification results of the identification module. The classification results will be used for subsequent updates and storage of sample and impurity data.

[0042] (3) Cloud Platform Data Upload: The categorized data needs to be uploaded to the cloud platform. This typically involves the following steps: Assess the data type and size to select a suitable cloud service provider.

[0043] Ensure you have a stable internet connection with sufficient bandwidth.

[0044] Classify and organize the data, and choose a suitable upload tool, such as web upload, dedicated client software, or command-line tool.

[0045] When uploading sensitive data, encrypt the data.

[0046] (4) Database Update: Update the sample and impurity databases based on the classification results. This includes inserting new records, updating existing records, or deleting outdated records. The update process needs to ensure data consistency and integrity, and concurrent operations are handled through transaction management.

[0047] (5) Result Verification and Feedback: After data upload and database update are completed, the results are verified to ensure data accuracy. The system provides a feedback mechanism to keep users informed of the data processing status and to intervene when necessary.

[0048] (6) Monitoring and Maintenance: The data analysis module has monitoring functions to track potential problems during data processing. Regular maintenance helps optimize system performance and ensures the efficiency and stability of the data processing flow.

[0049] Through the above process, the data analysis module can effectively process multiple types of data, classify samples and impurities, and ensure accurate data upload and timely database updates. The data analysis principles are detailed in the attached instruction manual. Figure 5 As shown.

[0050] Example 2 according to Figures 6-9 As shown, this embodiment provides theoretical support for a physical sorting submodule in a customs port screening system, as well as an explanation of the principle of the YOLO neural network.

[0051] Centrifugation Screening Unit: The material remaining after three or four layers of sieving enters the centrifugation screening unit for slow centrifugation. The rotating disc has a rough surface and is controlled by a motor to rotate at a constant speed, generating centrifugal force. The rotation speed is controlled within the range of 10-100 rpm. Under the action of centrifugal force, the sample on the disc surface will move towards the edge of the disc. The roughness of the sample surface, the sample morphology, and the sample density all affect its radial distribution on the disc. Finally, the sample is distributed radially to areas with different centrifugation radii. If the disc radius used is r, the targets are divided into three categories according to the radius after centrifugation: Proximal targets (0-1 / 3r): Targets with low density, rough surface, small particle size and irregular shape, such as impurities; Intermediate targets (1 / 3r-2 / 3r): targets with moderate density, relatively smooth surface, moderate particle size and relatively regular shape, such as light weed seeds and incomplete forage samples; Edge targets (2 / 3r-r): targets with higher density, smooth surface, larger particle size and round and regular shape, such as round weed seeds.

[0052] After centrifugation, the annular baffle is lowered to divide the turntable area into three zones. The pneumatic cleaning module is then activated to send the samples from the three zones into three containers along the three discharge ports and then transport them along the conveyor belt to the next sorting module.

[0053] This centrifugal sorting method effectively separates particles of different densities, providing clearly categorized samples for further mechanical vibration and aspiration. Centrifugal sorting not only improves sorting accuracy but also reduces the complexity of subsequent processing, ensuring that each category of sample receives targeted treatment. A schematic diagram of the centrifugal sorting principle is attached to the instruction manual. Figure 6 As shown.

[0054] Centrifugal screening is based on Stokes' law at low Reynolds numbers and the balance of centrifugal force and friction. Particle Reynolds Number The Reynolds number is a dimensionless number that describes the motion of particles in a fluid and can help determine whether the fluid flow is laminar or turbulent. The formula for calculating the particle Reynolds number is: in It is the density of the gas inside the centrifuge. It is the diameter of the sample particles. It is the velocity of the sample particles relative to the gas. This refers to the dynamic viscosity of the gas inside the centrifuge. Calculating the particle Reynolds number allows us to determine whether the flow is laminar, transitional, or turbulent, which is crucial for fluid dynamics research and applications. The Reynolds number is directly proportional to the sample's velocity. In this system, centrifugation is slow, resulting in a low Reynolds number, indicating laminar airflow within the centrifuge. The sample's velocity at equilibrium can be categorized into two types: one for lightweight targets that float during centrifugation. In laminar air, the velocity of this type of target can be calculated using Stokes' law. in It is the density of the sample particles. The viscosity of the gas inside the centrifuge. The radius of the sample particle, This is due to gravitational acceleration. On the other hand, the velocity of the rotating sample is affected by the radius of rotation R and the rotational speed. Influence: .

[0055] At a constant rotational speed, the greater the density, the faster the speed and the larger the radius of rotation, placing it on the outer edge of the turntable. Another type of target moves on a rough turntable during centrifugation; its speed can be given by the centrifugal force formula. Friction is primarily determined by the roughness of the target surface, and is defined as follows: g, and thus the velocity required for balance can be obtained. The rougher the sample surface, the greater the speed required for balance, and the less likely it is to move and stay near the center of the turntable. On the other hand, a smooth surface cannot be balanced at a lower speed and will move to the edge of the turntable.

[0056] Vibration Negative Pressure Suction Unit: Targets at the edge of the rotating disc after centrifugal screening are mechanically vibrated at different frequencies and amplitudes, and collected using varying suction forces. The vibration frequency is controlled between 0-30Hz, the amplitude between 0-5mm, and the negative pressure suction force between 1-4kPa. Different combinations of frequency, amplitude, and suction force create three modes, classifying the targets into three categories: Mode 1: High-frequency vibration, low suction, suitable for light impurities; Mode 2: Medium-frequency vibration, medium suction, suitable for medium-density weed seeds; Mode 3: Low-frequency vibration, high suction, suitable for heavy impurities.

[0057] Mechanical vibration and suction technology are key steps in automated sorting systems for further separating and collecting particles of different densities. During mechanical vibration, the target particles at the edge of the turntable are laid flat on a vibrating platform that vibrates at different frequencies and amplitudes. High-frequency vibration is suitable for lightweight impurities, which are easily separated and collected by a low-suction suction device. Medium-frequency vibration is suitable for medium-density weed seeds, which are separated and collected by a medium-suction suction device. Low-frequency vibration is suitable for heavy impurities, which are separated and collected by a high-suction suction device.

[0058] This combination of mechanical vibration and suction allows for the precise separation of particles of different densities, ensuring that each type of sample is effectively processed. This technology not only improves sorting efficiency but also reduces human intervention, increasing the system's automation level. A diagram illustrating the vibration-negative pressure suction principle is attached to the instruction manual. Figure 7 As shown.

[0059] Acoustic standing wave airflow tomography unit: After vibration negative pressure absorption, three types of targets are identified. Each type of target is screened using resonance with sound waves of different frequencies. At the peaks of the standing waves formed by the sound waves, air jets of different velocities are collected using air jet pipes. The sound wave frequency range is controlled between 10-40MHz, and the air jet velocity is controlled between 0-10m / s. Each group of targets is divided into three categories according to different sound wave frequencies and jet velocities: High-frequency sonic jet: suitable for small weed seeds; Medium-frequency acoustic jet: suitable for medium-sized weed seeds; Low-frequency sonic high-speed jet: suitable for larger weed seeds.

[0060] Acoustic standing wave airflow chromatography is an advanced sorting method that utilizes the resonant properties of sound waves to further screen particles of different sizes. During sound wave vibration, a sound generator produces sound waves of different frequencies, forming standing waves within a cavity. The sample resonates under the influence of the high-frequency sound waves and concentrates at the peaks of the standing waves, thus being separated. At each peak, airflow with varying suction forces is used to collect the separated particles.

[0061] This combination of acoustic vibration and airflow chromatography allows for the precise separation of particles of different sizes, ensuring that each type of sample is effectively processed. This technology not only improves sorting accuracy but also reduces the complexity of subsequent processing, thereby increasing the overall efficiency of the system. The principle of acoustic standing wave airflow chromatography is explained in the attached instruction manual. Figure 8 As shown.

[0062] The principle of standing waves within the cavity serves as the theoretical support for this system.

[0063] In a standing wave airflow tomography unit, when the sound waves generated by the sound generator propagate within the cavity, the presence of reflected waves causes the wave in space to have both forward and backward propagating components. These components superimpose at certain locations to form a fixed waveform called a standing wave. Within these waveforms, some points have an amplitude that is always zero, called troughs, while others have amplitudes that superimpose to their maximum value, called antinodes. The condition for the formation of a standing wave is two waves with the same frequency and amplitude but propagating in opposite directions. The mathematical explanation for its formation is as follows: Consider a sound wave propagating along the positive x-axis, whose expression is: The expression for the reflected wave is: The total displacement after the two waves are superimposed is: From this equation, we can deduce that the spatial location of the antinodes of a standing wave is determined by the cosine function. The decision, its maximum value appears in At odd multiples of, that is: because Substituting the values, we can obtain the position x of the antinode, which satisfies the condition. That is, the antinode appears at an odd multiple of one-quarter of the wavelength. , , At positions such as these, the crests and troughs of two waves propagating in opposite directions meet, causing the amplitudes to overlap and forming antinodes.

[0064] Electrostatic Adsorption Cleaning Unit: Dozens of samples, classified as described above, are laid flat on a conveyor belt and passed through an electrostatic generator. The samples, carrying static electricity, are then fed into the electrostatic adsorption cleaning device. Passing through conveyor belts with voltage applied to both sides, the tiny dust particles carrying static electricity are adsorbed onto the conveyor belts. Different voltages are applied to the sides of the conveyor belts in different sections, with the voltage range controlled between 5kV and 35kV. The current between the conveyor belts ranges from 0 to 400μA, and the spacing between the charged conveyor belts is 5mm to 50mm. Each sample group undergoes electrostatic adsorption cleaning. Afterwards, dust particles are removed from the conveyor belts using a dust cleaning brush head, ensuring the conveyor belts are reused. The cleaned samples are left on the original conveyor belts and enter the subsequent identification and sampling module.

[0065] Electrostatic adsorption technology can effectively remove minute impurities from sample surfaces, preparing samples for image recognition and voucher specimen preservation, thereby improving sample purity and image recognition accuracy. The principle of electrostatic adsorption cleaning is as described in the attached instruction manual. Figure 9 As shown.

[0066] Electrostatic induction, conductivity, charge attraction, and van der Waals forces serve as the theoretical basis for this module.

[0067] electrostatic induction Electrostatic induction plays a crucial role in electrostatic adsorption cleaning. When impurities such as weeds in grains and cereals approach an electrostatic field, electrostatic induction occurs. Electrostatic induction refers to the redistribution of free charges within a conductor when a charged object approaches it, under the influence of the electric field. This results in one end of the conductor carrying one type of charge, while the other end carries a different charge, with equal magnitudes and opposite signs. The formula is: in The amount of induced charge on the conductor. This represents the charge of the external charged body.

[0068] electrical conductivity Electrical conductivity is a physical quantity that describes the ability of a material to conduct electricity, and it also has a certain impact on the sorting of impurities in grains and cereals. The formula for electrical conductivity is: in Electrical conductivity, measured in Siemens units per meter (S / m); The current flowing through a conductor is measured in amperes (A). Electric field strength is expressed in volts per meter (V / m).

[0069] Charge attraction force During the sorting process, electrostatic attraction helps to adsorb impurities onto specific collection devices. Electrostatic attraction refers to the interaction force between charged bodies, and its formula is Coulomb's law: in It is the force of electric attraction, and its unit is Newton (N). is the Coulomb constant, with a value of 9.0 × 10⁻⁶. N• / ; and These are the charges of the two charged bodies, both in coulombs (C). The distance between two charged bodies is measured in meters (m).

[0070] van der Waals Van der Waals forces are weak interactions between neutral molecules or atoms, and they play an auxiliary role in the adsorption and sorting of impurities in grains and cereals. The formula for van der Waals forces is: in The van der Waals force is measured in Newtons (N). This is the van der Waals constant, whose value depends on the type of molecule or atom; It is the distance between molecules or atoms, and the unit is meter (m).

[0071] The principle of YOLO neural network: Grid division: YOLO first divides the input image into a fixed-size S×S grid. Each grid cell is responsible for detecting targets within its designated region. For example, in YOLOv1, the original image is divided into a 7×7 grid.

[0072] Bounding box prediction: For each grid cell, the YOLO algorithm predicts multiple bounding boxes. These bounding boxes are represented by their center coordinates (x, y), width (w), and height (h). Each bounding box also includes a confidence score, indicating whether an object is contained within the box and its accuracy.

[0073] Object classification: For each bounding box, YOLO uses a classifier to predict the object's category. Convolutional neural networks (CNNs) are typically used to extract features, and fully connected layers are used for classification.

[0074] Confidence assessment: Each bounding box also predicts a confidence score, which represents the probability that an object exists in the bounding box and the accuracy of the bounding box.

[0075] Non-maximum suppression (NMS): For each category, non-maximum suppression is used to remove overlapping bounding boxes. The bounding box with the highest confidence is selected, and bounding boxes with high overlap with it are discarded.

[0076] The YOLO network output is an S×S×(5×B+C) tensor, where S is the grid size; B is the number of bounding boxes predicted per grid; and C is the number of classes.

[0077] For each bounding box, YOLO predicts the following parameters: x, y: Relative coordinates of the center point of the bounding box, ranging from 0 to 1; w,h: Width and height of the bounding box, normalized relative to the size of the entire image; Confidence: The probability that the bounding box contains an object and the accuracy of the bounding box prediction.

[0078] Example 3 This embodiment provides an application example of the screening method of a customs port screening system.

[0079] Suppose we have a batch of grain samples containing straw, weeds, weed seeds, diseased or rotten beans, insect carcasses and eggs, sand, and other impurities. The automated sorting system described above will be used to sort this batch of samples to remove impurities and extract pure grains.

[0080] (1) Spread the grain and straw samples evenly on a vibrating sieving platform and improve the purity of the samples through four-stage sieving. The first stage of sieving removes straw and large weeds, the second stage separates beans and weed seeds of 5 mm and above, the third stage removes incomplete particles of 3 mm-5 mm, and the fourth stage removes small weed seeds and impurities of 0.5 mm-3 mm. This process ensures that the samples are evenly distributed and improves sieving efficiency.

[0081] (2) After three or four layers of sieving, the 3mm-5mm incomplete beans and weed seeds are put into a rotary centrifuge. The centrifugal force is used to divide the particles into three categories according to density: the near-center target is the impurities with higher density, the middle target is the weed seeds with moderate density, and the edge target is the light impurities with lower density, such as insect corpses and insect eggs.

[0082] (3) Place the three types of targets after centrifugation and screening into the vibration platform respectively. High frequency vibration is suitable for light impurities (such as insect corpses and insect eggs), which are collected by a low suction device; medium frequency vibration is suitable for medium density weed seeds, which are collected by a medium suction device; low frequency vibration is suitable for heavy impurities (such as sand), which are collected by a high suction device.

[0083] (4) The target seeds absorbed by mechanical vibration are laid flat on a vibration platform, and resonance screening is performed using sound waves of different frequencies and amplitudes. High-frequency sound waves separate small weed seeds, medium-frequency sound waves separate medium-sized seeds, and low-frequency sound waves separate larger weed seeds. The separated particles are collected by airflows with different suction forces.

[0084] (5) After classification, multiple groups of samples are laid out on a conveyor belt and electrostatic adsorption is used to clean and remove tiny impurities from the surface, improving purity and identification accuracy. The cleaned samples are photographed by a high-resolution camera, and the images are transmitted to an AI target detection algorithm for weed seed identification. Finally, the samples are picked up, packaged, and labeled by a robotic arm and stored for subsequent analysis.

[0085] (6) After secondary screening, whole beans and weed seeds enter the color sorter. The color, shape and texture are analyzed by optical sensors to accurately separate soybeans from impurities. The separated impurities are centrifuged and screened together with other sieve materials, and then subjected to subsequent image recognition processing to improve sorting efficiency and quality.

[0086] Through the above multi-step sorting and processing, customs can effectively remove impurities such as straw, weeds, weed seeds, diseased beans, insect carcasses and eggs, and sand from grain samples, extracting pure grain and ensuring the quality and safety of imported grain.

[0087] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A customs port screening system, characterized in that: The system includes a feeding module, a material screening module, an identification and sampling module, and a data analysis module. The feeding module is used to feed grain samples into the material screening module, which includes a main conveyor belt feeding submodule, an auxiliary conveyor belt circulation submodule, and a waste and impurity disposal submodule. The material screening module performs preliminary screening of the samples based on four-stage screening technology to separate materials with different particle sizes and densities. The identification and sampling module is used to acquire sample images and perform species-level identification and preserve voucher specimens using YOLO neural network technology. The data analysis module is used to perform data analysis on the screened data and perform classification, uploading, and database updates. The material screening module includes a four-stage molecular sieve module, a physical sorting sub-module, and a color sorting extraction sub-module. The four-stage molecular sieve module uses four layers of sieves with different particle sizes to perform four-stage sieving based on a size particle size screener. The color sorting extraction sub-module uses a color sorter to extract pure grain particles by screening based on color differences. The physical sorting submodule includes a centrifugal screening unit, a vibratory negative pressure suction unit, an acoustic standing wave airflow chromatography unit, and an electrostatic adsorption cleaning unit. The centrifugal screening unit uses centrifugal force to separate particles of different sizes from the oversize material after the three and four layers of screening in the four-stage sieve molecular module. The vibratory negative pressure suction unit uses mechanical vibration and negative pressure suction technology to further separate and collect particles of different densities. The acoustic standing wave airflow chromatography unit uses acoustic standing waves of different frequencies and airflow chromatography technology to accurately separate particles of different sizes again. The electrostatic adsorption cleaning unit uses an electrostatic generator to generate electrostatic adsorption to remove impurities from the sample surface. The identification and retention module includes a sample image acquisition submodule and a neural network recognition submodule. The sample image acquisition submodule acquires image information of the sample based on a high-resolution camera, and the neural network recognition submodule identifies the sample targets in the image based on the YOLO neural network, identifies the species level, and then labels and saves the data.

2. The customs port screening system according to claim 1, characterized in that: The main conveyor belt feeding submodule is used to transport the sample from the inlet to the screening area, the auxiliary conveyor belt circulation submodule is used to ensure the continuous flow of the sample during the screening process, and the waste and impurity disposal submodule is used to dispose of the screened waste and impurities.

3. The customs port screening system according to claim 1, characterized in that: The data analysis module includes a data classification submodule, a PTZ data upload submodule, and a database update submodule. The data classification submodule preprocesses the collected data and classifies the sample data and impurity data according to the identification results. The PTZ data upload submodule is used to upload the classified data to the PTZ. The database update submodule is used to update the sample database and impurity database according to the classification results.

4. A customs port screening system according to claim 3, characterized in that: The data classification submodule includes a grain sample data storage unit and an impurity sample data storage unit. The grain sample data storage unit is used to store the screened grain sample data, and the impurity sample data storage unit is used to store the separated and screened impurity sample data.

5. A customs port screening system according to claim 3, characterized in that: The database update submodule includes an impurity sample database update unit and a valid sample database update unit. The impurity sample database update unit is used to store impurity sample data identified outside the original database, and the valid sample database update unit is used to store valid sample data identified outside the original database.

6. A screening method for a customs port screening system according to any one of claims 1-5, characterized in that, Includes the following steps: Step 1: Four-stage screening process. The sample to be screened is transported to the four-stage screening module through the feeding module. It is screened through four layers of sieves with different particle sizes to screen out the straw and screen out and retain samples of large weeds and pests. Step 2: Centrifugal screening. The material remaining after the three-layer and four-layer screening in Step 1 is sent to the centrifugal screening unit and slowly centrifuged at a speed of 10-100 rpm to obtain the near-center target, the middle target, and the edge target. Step 3: Vibration and negative pressure suction treatment. The edge targets after centrifugation and screening in Step 2 are mechanically vibrated at different frequencies and amplitudes and collected using different suction forces to obtain high-frequency screening targets, medium-frequency screening targets and low-frequency screening targets. Step 4: Acoustic standing wave airflow tomography. For each of the three types of targets extracted after vibration negative pressure absorption in Step 3, resonant screening is performed using sound waves of different frequencies. At the peak position of the standing wave formed by the sound waves, jets of different speeds are used to collect the materials, resulting in high-frequency acoustic wave screened materials, medium-frequency acoustic wave screened materials, and low-frequency acoustic wave screened materials. Step 5: Electrostatic adsorption cleaning treatment. The multiple groups of materials obtained in steps 1 to 4 are laid flat on the conveyor belt. Each group of materials is electrostatically adsorbed and cleaned by an energized electrostatic generator to remove tiny impurities from the surface. After cleaning, the materials are identified and retained by the identification and retention module to obtain the impurity-removed materials. Step 6: Color sorting and extraction processing. The processed materials are separated according to their different characteristics using a color sorter. Finally, image recognition is performed to complete the screening and sorting of the materials.

Images