A continuous sorting method for defective cored lotus seeds based on online sound recognition

Abstract

The present invention discloses a method for continuously sorting defective cored lotus seeds based on online sound recognition, comprising the steps of placing and adjusting the cored lotus seeds, continuously spraying a curtain-like micro-pressure detection airflow from a detection airflow nozzle, continuously moving the cored lotus seeds through the detection airflow using a roller chain, collecting the sound response signals generated by the cored lotus seeds as they pass through the detection airflow in real time online using a microphone, detecting and classifying the sound response signals using a sound recognition program, and finally performing sorting by a sorting actuator. The present invention utilizes a micro-pressure detection airflow to excite the cored lotus seeds to generate a sound response signal, which minimally disturbs the lotus seeds. The roller chain can transport the cored lotus seeds unconstrained and continuously through the detection airflow, enabling online continuous sorting. Furthermore, the use of a detection airflow nozzle with a flattened outlet cross-section accommodates the sorting of cored lotus seeds of varying sizes, effectively resolving the problem of prior art techniques that require pre-processing operations such as compacting the lotus seeds and adjusting the nozzle position for centering, resulting in inability to perform continuous sorting. This fundamentally improves sorting efficiency.

Description

Technical Field

[0001] The present invention relates to the field of material classification based on sound recognition, in particular to a method for continuously sorting defective cored lotus seeds based on online sound recognition. Background Art

[0002] Lotus seeds are used medicinally and edible, possessing considerable economic value. Re-coring lotus seeds is a crucial process in lotus seed production, used to remove the core and eliminate bitterness. A common method involves drilling a through hole along the axis of the lotus seed to allow the core and its debris to escape. However, during this process, the drilled core hole often becomes misaligned with the axis of the lotus seed, resulting in defective cored lotus seeds. This prevents the core from being fully removed, leading to problems such as ingredient mixing, mildew, and reduced appearance quality. Sorting defective cored lotus seeds can effectively improve the quality of lotus seed products.

[0003] The difference between defective and good cored lotus seeds lies primarily in whether the core hole is coaxial with the axis of the lotus seed. Manual sorting offers good sorting quality, but it also presents issues such as fatigue, sanitation hazards, and low efficiency. Regarding automated sorting, patent document CN116944068B discloses a method for sorting defective cored lotus seeds based on airflow excitation and audio recognition. This technical solution first uses two rollers to accommodate the cored lotus seeds and adjust the lotus seed's posture so that its axis is parallel to the roller's axis. A pressure block is then used to press the lotus seed to secure it. The position of the air needle is then adjusted so that its axis is coaxial with the lotus seed's axis to achieve centering. The air needle is then directed to spray an excitation airflow toward the lotus seed's end face. Due to the different orientations of the core hole, the airflow acts differently on good and defective cored lotus seeds, generating differential sound signals. Based on this, an audio classification program is used to complete defective product sorting and detection.

[0004] However, this technical solution has the following limitations: 1. During sorting and testing, both the air needle and the lotus seed must be stationary, and the lotus seed must be compressed. Therefore, it can only operate intermittently, preventing continuous online sorting, which limits sorting efficiency. 2. Before sorting and testing, the air needle axis must be aligned with the lotus seed axis to meet the requirements for sorting and testing lotus seeds of different sizes. This additional air needle position adjustment process significantly increases sorting time and further reduces sorting efficiency. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a method for continuously sorting defective cored lotus seeds based on online sound recognition, which further improves the work efficiency of sorting defective cored lotus seeds through continuous operation.

[0006] The technical solution adopted by the present invention is: a method for continuously sorting defective cored lotus seeds based on online sound recognition, comprising the following steps:

[0007] Step 1: Deploy the roller chain and place the cored lotus seeds to be sorted in the link space of the roller chain;

[0008] Step 2: Adjust the posture of the cored lotus seeds so that their axis is parallel to the axis of the roller of the roller chain;

[0009] Step 3: Deploy a detection airflow nozzle and place it on one side of the roller chain so that its axis is parallel to the axis of the roller of the roller chain; the outlet cross-section of the detection airflow nozzle is flat; the outlet cross-section covers the height variation range of the axis of the cored lotus seeds on the roller chain in the height direction to accommodate the sorting of cored lotus seeds of different sizes;

[0010] Step 4: causing the detection airflow nozzle to continuously eject the detection airflow; causing the roller chain carrying the cored lotus seeds to continuously move and pass through the detection airflow;

[0011] Step 5: Using a microphone, collect the acoustic response signals generated by the cored lotus seeds passing through the test airflow in real time. Because the first cored hole on a qualified cored lotus seed is coaxial with the axis of the seed, the test airflow can pass through it. However, the second cored hole on a defective cored lotus seed is skewed from the axis of the seed, preventing the test airflow from passing through it. Therefore, there is a difference between the acoustic response signals corresponding to the qualified cored lotus seed and the acoustic response signals corresponding to the defective lotus seed.

[0012] Step 6: Use the sound recognition program to detect the sound response signal online, determine the category of the currently detected lotus seeds, and make a sorting decision. The sorting actuator then removes the defective products in the chain space to complete the sorting.

[0013] As a further optimization of this solution, in the above-mentioned continuous sorting method of defective cored lotus seeds based on online sound recognition, the sound recognition program is a deep learning model for automatically extracting the features of the sound response signal and classifying it.

[0014] As a further optimization of this solution, in the above-mentioned continuous sorting method of defective lotus seeds based on online sound recognition, the deep learning model adopts a one-dimensional convolutional neural network to directly identify and classify the one-dimensional sound response signal; the deep learning model can also adopt a two-dimensional convolutional neural network to identify and classify the two-dimensional time-frequency spectrum formed after the one-dimensional sound response signal is converted into time and frequency.

[0015] As a further optimization of this solution, in the above-mentioned continuous sorting method of cored lotus seeds based on online sound recognition, the parameter values ​​selected for detecting the airflow are based on the standard that the airflow does not destroy the posture of the cored lotus seeds in the chain space.

[0016] As a further optimization of this solution, in the above-mentioned method for continuous sorting of defective cored lotus seeds based on online sound recognition, the pressure of the detection airflow is less than or equal to 0.03 MPa.

[0017] As a further optimization of this solution, in the above-mentioned continuous sorting method of defective cored lotus seeds based on online sound recognition, the moving speed of the roller chain is less than or equal to 114.3 mm / s.

[0018] As a further optimization of this solution, in the above-mentioned continuous sorting method of defective cored lotus seeds based on online sound recognition, the specific shape of the outlet cross-section of the detection airflow nozzle is one of a rectangle, an oblong, and an ellipse; the width dimension of the outlet cross-section is smaller than the height dimension; and the detection airflow is a curtain-like airflow.

[0019] As a further optimization of this solution, in the above-mentioned continuous sorting method of defective cored lotus seeds based on online sound recognition, grooves are made on the circumferential surface of the rollers of the roller chain, and the chain link space is located between the grooves of adjacent rollers for accommodating a cored lotus seed.

[0020] As a further optimization of this solution, in the above-mentioned continuous sorting method of defective lotus seeds based on online sound recognition, the defective product removal method adopted by the sorting actuator is one of air blowing, hitting, sweeping, air suction, and clamping.

[0021] Compared with the prior art, the present invention has the following beneficial effects: 1. A roller chain is used to carry cored lotus seeds, and the sorting and detection of the cored lotus seeds is completed during the continuous movement of the chain without pausing, thereby effectively improving the efficiency of the sorting operation; 2. A micro-pressure detection airflow is used to excite the cored lotus seeds to produce a sound response signal, and the posture of the lotus seeds will not be destroyed during this process, so there is no need to press the lotus seeds, creating conditions for continuous sorting; 3. A detection airflow nozzle with a flat outlet cross-section is used to continuously spray a curtain-shaped micro-pressure detection airflow, thereby adapting to the sorting of cored lotus seeds of different sizes, so there is no need to adjust the nozzle position in advance, streamlining the process, and creating conditions for continuous sorting. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is the continuous sorting principle diagram of the present invention;

[0023] Figure 2 This is a comparison diagram of good and defective cored lotus seeds of the present invention when they pass through the detection airflow;

[0024] Figure 3 is a schematic diagram of a detection airflow nozzle of the present invention;

[0025] Figure 4 This invention Figure 3 AA cross-section in;

[0026] Figure 5 It is a schematic diagram of the curtain-shaped detection airflow of the present invention adapted to lotus seeds of different sizes.

[0027] The main components in the figure include: 1 roller chain; 1-1 roller; 1-1-1 groove; 1-2 chain link space; 2 detection airflow nozzle; 2-1 detection airflow; 2-2 outlet section; 3 cored lotus seeds; 3-1 good product; 3-1a first cored hole; 3-2 defective product; 3-2a second cored hole; 4 microphone; 4-1 sound response signal. DETAILED DESCRIPTION

[0028] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings.

[0029] like Figure 1-Figure 5 As shown, the present invention is a continuous sorting method for defective cored lotus seeds based on online sound recognition, and the continuous sorting method comprises the following steps:

[0030] Step 1: Deploy the roller chain 1 and place the cored lotus seeds 3 to be sorted in the link space 1-2 of the roller chain 1;

[0031] The roller 1-1 of the roller chain 1 is provided with a groove 1-1-1 on its circumferential surface, and the link space 1-2 is located between the grooves 1-1-1 of adjacent rollers 1-1 and is used to accommodate a cored lotus seed 3;

[0032] In a specific implementation, the cored lotus seeds 3 are placed by one of automatic feeding and manual feeding;

[0033] Step 2: Adjust the posture of the cored lotus seeds 3 so that their axis is parallel to the axis of the roller 1-1 of the roller chain 1;

[0034] In a specific implementation, the posture of the cored lotus seeds 3 is adjusted by causing the adjacent rollers 1-1 to rotate in the same direction, thereby driving the cored lotus seeds 3 in the link space 1-2 between the two to rotate;

[0035] Step 3: Deploy the detection airflow nozzle 2 and place it on one side of the roller chain 1, with its axis parallel to the axis of the roller 1-1 of the roller chain 1; the outlet cross-section 2-2 of the detection airflow nozzle 2 is flat; the outlet cross-section 2-2 covers the height variation range of the axis of the cored lotus seeds 3 on the roller chain 1 in the height direction to accommodate the sorting of cored lotus seeds 3 of different sizes;

[0036] When implementing it specifically, Figure 3-Figure 5As shown, the specific shape of the outlet cross section 2-2 of the detection airflow nozzle 2 is one of a rectangle, an oblong, and an ellipse; the width dimension of the outlet cross section 2-2 is smaller than the height dimension; the detection airflow 2-1 is a curtain-like airflow;

[0037] Step 4: The detection airflow nozzle 2 continuously ejects the detection airflow 2-1; the roller chain 1 carrying the cored lotus seeds 3 continuously moves and passes through the detection airflow 2-1;

[0038] The parameter values ​​selected for the detection airflow 2-1 are based on the standard that the detection airflow 2-1 does not destroy the posture of the cored lotus seeds 3 in the chain link space 1-2; in specific implementation, the air pressure of the detection airflow 2-1 is less than or equal to 0.03 MPa, and the moving speed of the roller chain 1 is less than or equal to 114.3 mm / s;

[0039] Step 5: Using microphone 4, collect online and real-time acoustic response signals 4-1 generated when the cored lotus seeds 3 pass through the test airflow 2-1. Because the first cored hole 3-1a on the acceptable cored lotus seeds 3-1 is coaxial with the axis of the cored lotus seeds 3, the test airflow 2-1 can pass through the first cored hole 3-1a. However, the second cored hole 3-2a on the defective cored lotus seeds 3-2 is tilted from the axis of the cored lotus seeds 3, and the test airflow 2-1 cannot pass through the second cored hole 3-2a. Therefore, there is a difference between the acoustic response signals 4-1 corresponding to the acceptable cored lotus seeds 3-1 and the acoustic response signals 4-1 corresponding to the defective cored lotus seeds 3-2.

[0040] Step 6: Use the sound recognition program to detect the sound response signal 4-1 online, determine the category of the coreless lotus seeds 3 being detected, and make a sorting decision. Then, the sorting actuator removes the defective products 3-2 in the chain link space 1-2 to complete the sorting.

[0041] In a specific implementation, the sound recognition program is a deep learning model for automatically extracting features of the sound response signal 4-1 and classifying it;

[0042] The deep learning model uses a one-dimensional convolutional neural network to directly identify and classify the one-dimensional sound response signal 4-1. The deep learning model can also use a two-dimensional convolutional neural network to identify and classify the two-dimensional time-frequency spectrum formed after the one-dimensional sound response signal 4-1 is converted into a time-frequency spectrum.

[0043] The defective product rejection method adopted by the sorting actuator is one of air blowing, hitting, sweeping, air suction, and clamping.

Claims

1. A method for continuously sorting defective cored lotus seeds based on online sound recognition, characterized in that: The following steps are involved: Step 1: Deploy the roller chain and place the cored lotus seeds to be sorted in the link space of the roller chain; Step 2: Adjust the posture of the cored lotus seeds so that their axis is parallel to the axis of the roller of the roller chain; Step 3: Deploy a detection airflow nozzle and place it on one side of the roller chain so that its axis is parallel to the axis of the roller of the roller chain; the outlet cross-section of the detection airflow nozzle is flat; the outlet cross-section covers the height variation range of the axis of the cored lotus seeds on the roller chain in the height direction to accommodate the sorting of cored lotus seeds of different sizes; Step 4: causing the detection airflow nozzle to continuously eject the detection airflow; causing the roller chain carrying the cored lotus seeds to continuously move and pass through the detection airflow; Step 5: Using a microphone, collect the acoustic response signals generated by the cored lotus seeds passing through the test airflow in real time. Because the first cored hole on a qualified cored lotus seed is coaxial with the axis of the seed, the test airflow can pass through it. However, the second cored hole on a defective cored lotus seed is skewed from the axis of the seed, preventing the test airflow from passing through it. Therefore, there is a difference between the acoustic response signals corresponding to the qualified cored lotus seed and the acoustic response signals corresponding to the defective lotus seed. Step 6: Use the sound recognition program to detect the sound response signal online, determine the category of the currently detected lotus seeds, and make a sorting decision. The sorting actuator then removes the defective products in the chain space to complete the sorting.

2. The method for continuously sorting defective cored lotus seeds based on online sound recognition according to claim 1, characterized in that: The sound recognition program is a deep learning model that is used to automatically extract features of sound response signals and classify them.

3. The method for continuously sorting defective cored lotus seeds based on online sound recognition according to claim 2, characterized in that: The deep learning model uses a one-dimensional convolutional neural network to directly identify and classify one-dimensional sound response signals; the deep learning model can also use a two-dimensional convolutional neural network to identify and classify two-dimensional time-frequency spectrograms formed after time-frequency conversion of one-dimensional sound response signals.

4. The method for continuously sorting defective cored lotus seeds based on online sound recognition according to claim 1, characterized in that: The parameter values ​​selected for the detection airflow are based on the standard that the detection airflow does not destroy the posture of the cored lotus seeds in the chain link space.

5. The method for continuously sorting defective cored lotus seeds based on online sound recognition according to claim 4, characterized in that: The air pressure of the detected airflow is less than or equal to 0.03 MPa.

6. The method for continuously sorting defective cored lotus seeds based on online sound recognition according to claim 1, wherein: The moving speed of the roller chain is less than or equal to 114.3 mm / s.

7. The method for continuously sorting defective cored lotus seeds based on online sound recognition according to claim 1, characterized in that: The specific shape of the outlet cross section of the detection airflow nozzle is one of a rectangle, an oblong, and an ellipse; the width dimension of the outlet cross section is smaller than the height dimension; and the detection airflow is a curtain-shaped airflow.

8. The method for continuously sorting defective cored lotus seeds based on online sound recognition according to claim 1, characterized in that: Grooves are formed on the circumferential surfaces of the rollers of the roller chain, and the chain link space is located between the grooves of adjacent rollers and is used to accommodate a cored lotus seed.

9. The method for continuously sorting defective cored lotus seeds based on online sound recognition according to claim 1, characterized in that: The defective product rejection method adopted by the sorting actuator is one of air blowing, hitting, sweeping, air suction, and clamping.

Images