Integrated detection and sorting equipment for internal and external defects of walnut kernels based on multispectral imaging

Abstract

The invention provides walnut kernel internal and external defect integrated detection and separation equipment based on multispectral imaging, and relates to the technical field of food safety processing, the equipment comprises a detection bin, and a material distribution system which drives a screen frame to reciprocate to realize uniform material distribution of walnut kernels is arranged in the bin; the top operation system can move on the top XY sliding rail and integrates feeding, first multispectral imaging and defective product removing and cleaning functions of a multi-axis mechanical arm; the second multispectral imaging device can move on the bottom track; and a fixed temperature control unit. The problems that irregular materials are difficult to automatically distribute, and the detection precision of internal hidden chemical defects is low are solved, full-process automation from feeding to detection to sorting and self-cleaning is achieved, and the automatic sorting machine has the advantages of being high in detection precision, high in sorting efficiency, high in integration degree, stable in operation and the like.

Description

Technical Field

[0001] This invention belongs to the field of food safety processing technology, specifically relating to an integrated detection and sorting device for internal and external defects of walnut kernels based on multispectral imaging. Background Technology

[0002] Walnuts are a popular and highly nutritious food, and quality control is a core aspect of their production and processing, directly impacting their commercial value and food safety. According to the requirements of the HACCP (Hazard Analysis and Critical Control Points) food safety management system, effective analysis and control of various potential hazards are essential during the production of dried walnuts and related products to ensure the safety of the final product.

[0003] The HACCP plan clearly identifies several key hazards in the walnut kernel production process: Biological hazards, such as mold caused by microbial contamination and insect infestation, are considered significant raw material defects. Mold not only affects the sensory quality of products, but the mycotoxins it produces also pose a serious threat to human health. Therefore, HACCP plans to include microbial indicators such as mold as key acceptance and process control items. Chemical hazards can occur when oils oxidize and become rancid due to improper storage or processing (i.e., "oil spoilage"). The HACCP program aims to control these hazards by strictly monitoring hygiene indicators such as peroxide value and acid value.

[0004] Physical hazards, such as shell fragments, pebbles, and other foreign objects, must also be removed to ensure product purity.

[0005] However, current technologies for controlling these hazards still have limitations. Walnut kernel quality control largely relies on manual sorting, which is not only inefficient and labor-intensive, but also lacks consistent sorting standards for defects with only minor differences in color or shape, resulting in poor reliability. In particular, early mold, insect infestation, or rancidity occurring inside the product are completely invisible to the human eye.

[0006] While some automated testing technologies have made progress, they still cannot fully meet the HACCP system's control requirements for critical hazards: visible light technology cannot penetrate the product surface and is ineffective against internal biological hazards such as mold and insect infestation; while indirect inference methods based on physical parameters such as weight and vibration cannot directly identify chemical composition changes caused by issues like oil rancidity. Therefore, there is an urgent need in this field for an automated device capable of efficiently, accurately, and comprehensively detecting both internal and external defects in walnut kernels. This device should not only identify external flaws but, more importantly, directly detect internal biological and chemical deterioration related to HACCP critical control points, thereby fundamentally ensuring product quality and food safety. Summary of the Invention

[0007] To overcome the problems in the background technology, this invention develops an integrated detection and sorting device for internal and external defects of walnut kernels based on multispectral imaging technology, aiming to solve the technical challenges of uniformly distributing irregular walnut kernel materials, simultaneously detecting internal and external defects, and efficiently and accurately sorting them.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solution: The walnut kernel internal and external defect integrated detection and sorting equipment based on multispectral imaging includes a detection chamber, a material feeding system, a top operation system, a bottom detection system, a temperature control unit, and a central controller, all installed within the detection chamber. The fabric system includes a support platform installed on the ground of the testing chamber, a sieve frame slidably installed on the support platform, and a drive mechanism for driving the sieve frame to reciprocate. A transparent testing plate is horizontally installed inside the sieve frame, and an electrically controlled baffle is hinged to the outlet end of the sieve frame. The top operation system includes an XY slide rail installed on the top wall of the detection chamber corresponding to the fabric system, a feeding device installed on the XY slide rail and moving along the slide rail, a first multispectral imaging device, a rejection mechanism and a cleaning mechanism. The first multispectral imaging device and the rejection mechanism are jointly installed on the XY sliding base of an XY slide rail. The rejection mechanism is a multi-axis robotic arm. The bottom detection system includes a detection track installed on the ground of the detection chamber and located below the transparent detection plate, and a second multispectral imaging device installed on a movable base of the detection track; The temperature control unit includes an infrared heating lamp assembly installed on the inner wall of the detection chamber; The central controller is installed outside the detection chamber and is electrically connected to the fabric feeding system, the top operation system, the bottom detection system, and the temperature control unit to coordinate and control the reciprocating fabric feeding motion of the screen frame, the movement and operation of the top operation module and the bottom detection system; The central controller receives and processes multispectral image data from the first and second multispectral imaging devices. By extracting and comparing reflectance or absorptivity information of multiple preset spectral bands in the image data, and comparing the analysis results with the defect spectral feature database preset by the central controller, the internal and external defects of the walnut kernel are identified and classified.

[0009] Furthermore, the feeding device is a hopper installed on a movable base of the XY slide rail. The side wall of the detection chamber is provided with a feeding window for the feeding device to enter and exit. The feeding device is equipped with a light shield that cooperates with the feeding window. The feeding path of the feeding device corresponds to the transparent detection plate.

[0010] Furthermore, the cleaning mechanism includes a telescopic rod connected to the movable base of the XY slide rail, and a scraper installed at the lower end of the telescopic rod to cooperate with the sieve frame and the transparent detection plate.

[0011] Furthermore, the testing chamber is provided with a premium product frame located below the baffle of the sieve frame, and a defective product frame and a scrap product frame located on the side of the carrying platform. The testing chamber is equipped with a door corresponding to the premium product frame, the defective product frame and the scrap product frame.

[0012] The integrated detection method for internal and external defects in walnut kernels based on multispectral imaging technology, executed using the aforementioned equipment, includes the following steps: S1. Feeding and spreading: Control the feeding device in the top operation system to move above the transparent detection plate to feed the material. Then, drive the screen frame to reciprocate until the walnut kernels are evenly spread on the transparent detection plate to form a single layer. S2, First scan: Control the first multispectral imaging device in the top working system and the second multispectral imaging device in the bottom detection system to simultaneously scan the walnut kernels on the stationary transparent detection plate and acquire the first multispectral image data; S3. Temperature control processing: The temperature control unit is activated to perform a preset duration of temperature control processing on the walnut kernels on the transparent detection plate. S4. Second scan: Control the first multispectral imaging device and the second multispectral imaging device to perform synchronous scanning again to acquire second multispectral image data; S5. Analysis and Classification: Extract reflectance or absorptivity information of multiple preset spectral bands from the first and second multispectral image data, calculate the spectral feature change of each walnut kernel caused by temperature control treatment, and compare the change and original spectral features with a preset defect spectral feature database to identify the internal and external defects of each walnut kernel and classify them as superior, second-grade or waste products, while recording their position coordinates. S6. Defective Product Removal: Control the removal mechanism in the top operation system to remove the walnut kernels classified as substandard and waste products one by one according to their position coordinates, and put them into the corresponding collection boxes respectively. S7. Good Product Collection: After the defective products are removed, open the electronically controlled baffle at the outlet end of the sieve frame and control the cleaning mechanism in the top operation system to sweep all the remaining superior products on the transparent detection plate into the superior product frame below.

[0013] Furthermore, step S also includes using image data from the first multispectral imaging device to correct image errors introduced by the transparent detection plate in the second multispectral imaging device.

[0014] Furthermore, in step S8, the defect identification process is based on calculating the amount of spectral feature change between the first multispectral image data and the second multispectral image data.

[0015] Furthermore, after the good product collection step is completed, a self-cleaning step S8 is also included: controlling the scraper of the cleaning mechanism to clean the surface of the transparent detection plate, and simultaneously controlling the robotic arm of the rejection mechanism to perform auxiliary cleaning operations.

[0016] The beneficial effects of this invention are: 1. This invention integrates all working units into a closed testing chamber, effectively isolating it from interference from external ambient light, dust, and airflow, providing a stable and clean prerequisite for high-precision multispectral imaging; at the same time, this closed design also facilitates the control of internal temperature and humidity and prevents cross-contamination during processing, meeting the stringent requirements of the HACCP system for environmental control in the production process. 2. The multispectral imaging technology of this invention can acquire spectral fingerprint information in the near-infrared band, which is invisible to the human eye. Different chemical substances, such as mold metabolites and rancid fatty acids, have unique absorption and reflection characteristics for light of specific wavelengths. This invention, through temperature-controlled excitation-differential analysis, applies a brief thermal excitation through a temperature control unit and compares the multispectral scan data before and after. This method can greatly amplify the subtle changes in spectral characteristics caused by abnormal moisture due to mold (biological hazards) or by rancidity of oils (chemical hazards), elevating the detection dimension from "morphology" to "physicochemical reaction characteristics." It can accurately identify hidden internal defects that cannot be detected, providing strong technical support for the monitoring of key control points in the HACCP system. 3. The reciprocating material spreading system of the present invention solves the problem of single-layer spreading of irregular materials in a gentle and efficient manner. By driving the transparent detection plate to reciprocate, the material is spread quickly and evenly by using gentle vibration and inertial force, which provides a foundation for subsequent accurate imaging and rejection operations. 4. The spectral analysis capability of this invention enables it to accurately distinguish between "superior grade", "second-grade" with only appearance defects and "waste" with safety hazards, realizing multi-level intelligent sorting, which facilitates enterprises to reprocess and reuse second-grade products, ensuring food safety while maximizing the value of raw materials; 5. The equipment of this invention integrates an automated self-cleaning cycle, which can automatically clean and verify the effect of the core component transparent detection plate after batch operation, ensuring the long-term stability of detection accuracy, reducing manual maintenance, and meeting the high reliability requirements of industrial continuous production. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall external structure of this application; Figure 2 This is a schematic diagram of the overall internal structure of this application; Figure 3 This is a bottom view of the internal structure of this application; Figure 4 This is a side view of the internal structure of this application; Figure 5 This is a schematic diagram of the top-mounted operating system structure of this application; Figure 6 This is a schematic diagram of the fabric system structure of this application; Figure 7 This is a side view of the bottom detection system of this application; Figure 8 This is a flowchart of the process flow of this application; Figure 9 This is a block diagram of the control system of this application.

[0018] Explanation of reference numerals in the attached figures: 001-Detection chamber, 002-Fabrication system, 003-Top operation system, 004-Bottom detection system, 005-Temperature control unit, 006-Central controller, 011-Premium product frame, 012-Defective product frame, 013-Scrap product frame, 014-Bin door, 021-Bearing platform, 022-Screen frame, 023-Drive mechanism, 024-Transparent detection plate, 025-Baffle, 031-XY slide rail, 032-Feeding device, 321-Light shield, 033-First multispectral imaging device, 034-Rejection mechanism, 035-Cleaning mechanism, 351-Telescopic rod, 352-Scraper, 041-Detection track, 042-Second multispectral imaging device. Detailed Implementation

[0019] To make the objectives, technical solutions, and beneficial effects of the present invention clearer, the preferred embodiments of the present invention will be described in detail below to facilitate understanding by those skilled in the art.

[0020] like Figure 1 , Figure 2 As shown, the present invention provides an integrated detection and sorting device for internal and external defects of walnut kernels based on multispectral imaging technology. The main body of the device is a closed detection chamber 001, which forms a relatively independent, clean and light-proof working space. The detection chamber 001 is equipped with a central controller 006 for human-computer interaction and system operation, such as an industrial PC (IPC) with a built-in high-performance graphics card (GPU), and can perform high-speed image data preprocessing in real time through an FPGA coprocessor card.

[0021] Fabric System 002 like Figure 3 , Figure 4 , Figure 6 , Figure 7 As shown, the fabric system 002 includes a support platform 021 installed on the ground of the detection chamber 001, a screen frame 022 slidably installed on the support platform 021, a drive mechanism 023 that drives the screen frame 022 to reciprocate, a transparent detection plate 024 horizontally installed inside the screen frame 022, and an electrically controlled baffle 025 hinged to the outlet end of the screen frame 022. The fabric spreading system 002 is the foundation for automating the processing of irregular walnut kernels. The support platform 021 serves as a stable base and is equipped with linear guide rails. The sieve frame 022 is mounted on the guide rails via pulleys or sliders. In this embodiment, the drive mechanism 023 is preferably a crank-connecting rod mechanism consisting of a servo motor, an eccentric wheel, and a connecting rod. The central controller 006 can precisely control the speed and angle of the servo motor to adjust the frequency and amplitude of the reciprocating motion of the sieve frame 022 to adapt to walnut kernels of different sizes and friction coefficients, achieving the best single-layer spreading effect and ensuring that the walnut kernels do not overlap or pile up. The transparent detection plate 024 is preferably made of high-transmittance, low-dispersion quartz glass or surface-hardened acrylic plate to reduce optical interference to the bottom imaging.

[0022] Top XY Operation System 003 and Bottom Detection System 004 like Figure 2-5 As shown, the top operation system 003 includes an XY slide rail 031 installed on the top wall of the detection chamber 001 corresponding to the fabric system 002, a feeding device 032 installed on the XY slide rail 031 and moving along the slide rail, a first multispectral imaging device 033, a rejection mechanism 034 and a cleaning mechanism 035. The first multispectral imaging device 033 and the rejection mechanism 034 are jointly installed on a sliding base of an XY slide rail 031. like Figure 2 , Figure 3 , Figure 4 , Figure 7 As shown, the bottom detection system 004 includes a detection track 041 installed on the ground of the detection chamber 001 and located below the transparent detection plate 024, and a first multispectral imaging device 033 installed on the movable base of the detection track 041. like Figure 5 As shown, the XY slide rail 31 of the top XY working system 003 is a synchronous belt module or ball screw module driven by a high-precision servo motor to achieve rapid and accurate positioning of each working unit. In this embodiment, the first multispectral imaging device 33 and the second multispectral imaging device 042 are preferably a pushbroom hyperspectral camera or a filter wheel multispectral imaging system. Among them, the pushbroom hyperspectral camera has a spectral range covering 400nm to 1700nm, covering the visible light (VIS) and short-wave near-infrared (SWIR) bands. The light source used in conjunction with it is a linearly arranged broadband quartz halogen lamp, which ensures that uniform and high-intensity linear illumination is provided to the walnut kernel at the moment the camera pushbrooms to collect data. The multispectral imaging system mainly includes a high-sensitivity monochrome area array industrial camera, a high-speed rotating filter wheel, and a broadband illumination source. The high-sensitivity monochrome area array industrial camera is used to capture images. The high-speed rotating filter wheel is mounted in front of the camera lens and has multiple different narrowband filters installed on it. The central controller 006 can precisely control the rotation angle of the filter wheel, so that different filters are switched sequentially into the optical path. The broadband illumination source, such as a halogen lamp or xenon lamp array, provides uniform illumination covering all target wavelengths. The filters mounted on the aforementioned filter wheel, whose key wavelengths are pre-selected based on the spectral characteristics of internal and external defects in walnut kernels, include: A 550nm filter is used to obtain green channel information and assess whether the color is normal. The 675nm filter is responsive to chlorophyll and other substances, and can be used to identify some plant impurities. 970nm or 1450nm filters: strong absorption peaks for water, making them exceptionally sensitive to moisture in moldy areas; 1200nm or 1720nm filters: characteristic absorption peaks of oils and fats, responding to changes in fatty acid composition caused by oil deterioration; A full-pass or broadband filter: used to acquire high-resolution morphological images.

[0023] The rejection mechanism 34 is preferably a high-speed SCARA (horizontal multi-joint) robot or Delta (parallel) robot, whose end effector is a customized mechanical gripper or a multi-suction cup vacuum generator, which can selectively drive one or more mechanical grippers or suction cups to generate negative pressure according to the position and shape of the defective product, so as to perform precise and gentle gripping.

[0024] Temperature control unit 005 like Figure 3 As shown, the temperature control unit 005 includes an infrared heating lamp assembly installed on the inner wall of the detection chamber 001; in this embodiment, it is an adjustable power quartz halogen infrared heating lamp array. The central controller 006 can precisely control its heating power and irradiation time, such as rapidly and uniformly raising the surface temperature of the walnut kernel by 5-10 degrees Celsius within 2-3 seconds. This brief and gentle heating is sufficient to stimulate the differences in physicochemical reactions in defective parts without damaging the quality of good walnut kernels.

[0025] To ensure the long-term stable operation, high-precision control, and operational safety of the equipment of this invention, the equipment actually needs to be equipped with a variety of auxiliary conventional sensors and standard components. These auxiliary components include, but are not limited to: photoelectric sensors, proximity switches, or mechanical limit switches for determining the origin and travel limit positions of moving parts such as the XY slide rail 031, detection track 041, and screen frame 022; rotary or linear encoders for providing closed-loop position and speed feedback for the servo motors driving each motion axis; temperature sensors for monitoring the actual heating temperature of the temperature control unit 005 to achieve precise closed-loop control; and door magnetic sensors or safety light curtains linked with the door 014 to ensure operational safety. The integration of these auxiliary sensors with the central controller 006 is a conventional technical means in this field. The specific selection and arrangement of these sensors can be conventionally designed by those skilled in the art based on the core concept of this invention without any creative effort.

[0026] Equipment working method The present invention achieves fully automated operation through programmed control of the central controller 006: S1. Feeding and Spreading: The central controller 006 first controls the feeding device 032 of the top XY operating system 003 to move to the feeding area of ​​the transparent detection plate 024, and the pneumatic hopper opens to complete the feeding; then, the drive mechanism 023 of the spreading system 002 is started, so that the screen frame 022 reciprocates at a preset frequency of 5-15Hz and an amplitude of 10-30mm for 20-30s, and the walnut kernels can be evenly spread. After that, the central controller 006 stops the drive mechanism 023, and the screen frame 022 comes to a complete stop. S2, First Scan (Baseline Data Acquisition): With the sieve frame 022 stationary, the central controller 006 issues the first scan command. The first multispectral imaging device 33 in the top working system 003, driven by its XY slide rail (31), begins to move linearly along the X-axis (or Y-axis) at a constant speed. At the same time, its built-in push-broom camera begins to acquire images of the upper surface of the walnut kernels on the transparent detection plate 024 row by row. Meanwhile, the second multispectral imaging device 042 in the bottom detection system 004, on its detection track 041, begins to move linearly in strict synchronization, in the same direction, and at the same speed as the first multispectral imaging device 33, acquiring images of the lower surface of the walnut kernels. After the scan is completed, two original hyperspectral data cubes corresponding to the upper and lower surfaces are generated respectively. This data is stored as a baseline state. S3. Temperature control: After scanning, the sieve frame (22) and walnut kernels remain stationary. The central controller 006 starts the temperature control unit 005 fixed on the top wall of the chamber. Its infrared heating lamp group briefly and gently heats the walnut kernels, causing a small and uniform increase in surface temperature to stimulate the spectral response differences of internal defects. Step S4, Second Scan (Stress Data Acquisition): After the temperature control process is completed, the central controller 006 immediately issues a second scan command; the first multispectral imaging device 33 and the second multispectral imaging device 042 completely repeat the scanning process of step S2, and perform a complete synchronous scan of the upper and lower surfaces of the walnut kernel in the thermally excited state again at the same speed, path and parameters. After the scan is completed, two more hyperspectral data cubes are generated, and this data is stored as the stress state. S5. Analysis and Classification: Data preprocessing: First, geometric correction and image segmentation are performed on the two hyperspectral data cubes to identify each individual walnut kernel and its pixel region; Feature extraction and vector construction: From the spectral curve corresponding to each walnut kernel pixel, the reflectance values ​​of key bands are extracted, focusing on the characteristic absorption peaks around 1450nm related to moisture, around 1200nm and 1700nm related to oil, and in the 970nm-1100nm range related to protein and cellulose structure; at the same time, the change in reflectance of the second scan data relative to the first scan data in these characteristic bands is calculated. To improve computational efficiency, the system first uses methods such as Principal Component Analysis (PCA) or Continuous Projection Algorithm (SPA) to select several characteristic bands (λ1, λ2, ..., λ) from hundreds of bands that contribute the most to defect classification. n For each walnut kernel's ROI, the system calculates its average reflectance in each characteristic band and constructs two core feature vectors: The original spectral eigenvector V_orig=[R_before(λ1),R_before(λ2),...,R_before(λ)] n )]; The differential spectral eigenvector V_diff=[|R_after(λ1)-R_before(λ2)|,|R_after(λ1)-R_before(λ2)|,...,|R_after(λ1)-R_before(λ2)| n )-R_before(λ n )|]; Finally, these two vectors are combined into a final feature vector V_final=[V_orig,V_diff] for information fusion; Defect identification and classification: The final feature vector V_final is input into a classification model pre-trained with a large number of samples (such as Support Vector Machine (SVM) or Convolutional Neural Network (CNN)). The model has a built-in defect spectral feature database, which pre-stores a large number of verified superior, substandard (e.g., attached shells) and waste (e.g., different degrees of mold and oiliness) samples under different temperature control conditions. The model compares the feature vector extracted in real time with the standard features in the database and outputs the classification result (superior, substandard, waste) and confidence score for each walnut kernel. Error correction: In this step, the controller compares the images from the upper and lower cameras. If a low reflectivity area with a suspected defect exists in the lower camera image, but also has similar spectral characteristics in the corresponding transparent background area of ​​the upper camera, the system determines that the defect originates from the transparent detection plate 024 itself (such as stains or scratches) and ignores the signal to avoid misjudgment. S6, Defective Product Removal: The central controller 006 plans the movement path of the robotic arm of the removal mechanism 34 according to the coordinate list of defective products (substandard products and waste products) output by S5, and picks up the defective products one by one with the highest efficiency and puts them into the substandard product box 012 and the waste product box 013 respectively. S7. Good Product Collection: After rejection, the baffle 025 at the outlet of the sieve frame 022 is opened by electronic control, and the scraper 352 of the cleaning mechanism 35 sweeps all the remaining superior products on the plate into the superior product frame 011 below. S8. Self-cleaning: After the batch operation is completed, the scraper of the cleaning mechanism 35 first sweeps the large particles of debris to the corner, and then the robotic arm of the rejection mechanism 34 changes the end tool or uses a suction cup to pick up the debris to the debris box or suck it away; finally, the imaging system can perform a quick scan of the empty plate and determine whether repeated cleaning is needed through a cleanliness algorithm.

[0027] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made to it in form and detail without departing from the scope defined by the claims of the present invention.

Claims

1. A multispectral imaging-based integrated detection and sorting device for internal and external defects in walnut kernels, characterized in that: It includes a detection chamber (001), a fabric distribution system (002), a top operation system (003), a bottom detection system (004), a temperature control unit (005), and a central controller (006) installed in the detection chamber (001). The fabric system (002) includes a support platform (021) installed on the ground of the testing chamber (001), a sieve frame (022) slidably installed on the support platform (021), and a drive mechanism (023) for driving the sieve frame (022) to reciprocate. A transparent testing plate (024) is horizontally installed inside the sieve frame (022), and an electrically controlled baffle (025) is hinged to the outlet end of the sieve frame (022). The top operation system (003) includes an XY slide rail (031) installed on the top wall of the detection chamber (001) corresponding to the fabric system (002), a feeding device (032) installed on the XY slide rail (031) and moving along the slide rail, a first multispectral imaging device (033), a rejection mechanism (034) and a cleaning mechanism (035). The first multispectral imaging device (033) and the rejection mechanism (034) are jointly installed on the XY sliding base of an XY slide rail (031). The rejection mechanism (034) is a multi-axis robotic arm. The bottom detection system (004) includes a detection track (041) installed on the ground of the detection chamber (001) and located below the transparent detection plate (024), and a second multispectral imaging device (042) installed on a movable base of the detection track (041); The temperature control unit (005) includes an infrared heating lamp assembly installed on the inner wall of the detection chamber (001); The central controller (006) is installed outside the detection chamber (001) and is electrically connected to the fabric distribution system (002), the top operation system (003), the bottom detection system (004), and the temperature control unit (005) to coordinate the control of the reciprocating fabric distribution motion of the screen frame (022), the movement and operation of the top operation module and the bottom detection system (004); The central controller (006) receives and processes the multispectral image data of the first multispectral imaging device (033) and the second multispectral imaging device (042). By extracting and comparing the reflectance or absorptivity information of multiple preset spectral bands in the image data, and comparing the analysis results with the defect spectral feature database preset by the central controller (006), the internal and external defects of the walnut kernel are identified and classified.

2. The device according to claim 1, characterized in that: The feeding device (032) is a hopper installed on the movable base of the XY slide rail (031). The side wall of the detection chamber (001) is provided with a feeding window for the feeding device (032) to enter and exit. The feeding device (032) is provided with a light shield (321) that cooperates with the feeding window. The feeding path of the feeding device (032) corresponds to the transparent detection plate (024).

3. The device according to claim 1, characterized in that: The cleaning mechanism (035) includes a telescopic rod (351) connected to the movable base of the XY slide rail (031), and a scraper (352) installed at the lower end of the telescopic rod (351) to cooperate with the sieve frame (022) and the transparent detection plate (024).

4. The device according to claim 1, characterized in that: The testing chamber (001) has a premium product frame (011) located below the baffle (025) of the sieve frame (022), and a defective product frame (012) and a scrap product frame (013) located on the side of the carrying platform (021). The testing chamber (001) is equipped with a door (014) corresponding to the premium product frame (011), the defective product frame (012) and the scrap product frame (013).

5. A method for integrated detection of internal and external defects in walnut kernels based on multispectral imaging technology, characterized in that, The device according to any one of claims 1-4 performs the following steps: S1. Feeding and spreading: Control the feeding device (032) in the top operation system (003) to move above the transparent detection plate (024) to feed the material. Then, drive the screen frame (022) to reciprocate until the walnut kernels are evenly spread on the transparent detection plate (024) to form a single layer. S2, First scan: Control the first multispectral imaging device (033) in the top operation system (003) and the second multispectral imaging device (042) in the bottom detection system (004) to simultaneously scan the walnut kernels on the stationary transparent detection plate (024) and acquire the first multispectral image data; S3. Temperature control process: The temperature control unit (005) is activated to perform a preset time temperature control process on the walnut kernels on the transparent detection plate (024); S4. Second scan: Control the first multispectral imaging device (033) and the second multispectral imaging device (042) to perform synchronous scanning again to acquire second multispectral image data; S5. Analysis and Classification: Extract reflectance or absorptivity information of multiple preset spectral bands from the first and second multispectral image data, calculate the spectral feature change of each walnut kernel caused by temperature control treatment, and compare the change and original spectral features with a preset defect spectral feature database to identify the internal and external defects of each walnut kernel and classify them as superior, second-grade or waste products, while recording their position coordinates. S6. Defective product rejection: Control the rejection mechanism (034) in the top operation system (003) to remove the walnut kernels classified as substandard and waste products one by one according to the position coordinates and put them into the corresponding collection boxes respectively; S7. Good Product Collection: After the defective products are removed, open the electronically controlled baffle (025) at the outlet end of the sieve frame (022) and control the cleaning mechanism (035) in the top operation system (003) to sweep all the remaining superior products on the transparent detection plate (024) into the superior product frame below.

6. The integrated inspection method for internal and external defects of walnut kernels according to claim 5, characterized in that: Step S5 further includes using image data from the first multispectral imaging device (033) to correct image errors introduced by the transparent detection plate (024) in the second multispectral imaging device (042).

7. The integrated inspection method for internal and external defects of walnut kernels according to claim 5, characterized in that: In step S5, the defect identification process is based on calculating the amount of spectral feature change between the first multispectral image data and the second multispectral image data.

8. The method according to claim 5, characterized in that: After the good product collection step is completed, a self-cleaning step S8 is also included: controlling the scraper of the cleaning mechanism (035) to clean the surface of the transparent detection plate (024), and simultaneously controlling the robotic arm of the rejection mechanism (034) to perform auxiliary cleaning operations.

Images