Detection device and method for detecting mould in food

Abstract

The invention discloses a detection device and method for detecting mould in food, and relates to the technical field of food safety detection.The detection device comprises a device shell, a detection probe and a sample analysis system, and the sample analysis system comprises an image acquisition unit, a feature extraction unit, a feature comparison unit and a sampling delimiting unit; a sample image on a sample disc is obtained through a camera, a category feature recognition model is constructed based on a convolutional neural network, feature extraction is performed on a sample analysis image through the feature recognition model to obtain sample category features, and meanwhile, a sample standard comparison image corresponding to the sample analysis image is obtained; and carrying out mold point searching on the sample analysis image, selecting the mold point with the largest area as a sampling point, further planning the moving track of the multi-axis robot, adjusting a servo motor according to the sampling range so as to adjust the distance between cutting knives, and adjusting the steering degree of a first driving motor so as to ensure the final sample quality.

Description

Technical Field

[0001] The present invention relates to the technical field of food safety detection, and in particular to a detection device and method for detecting mold in food. Background Art

[0002] Mycotoxins are secondary toxic metabolites produced by fungi, which are commonly found in feed and food. They harm humans or animals mainly through the food supply chain, including causing common inflammatory reactions such as gastrointestinal diseases, changing macrophage phenotypes and DNA damage, and even being highly carcinogenic. They can also harm the reproductive system of humans or animals and have serious effects on children's growth and development. Mycotoxins are contaminated in a wide range, including crops, fresh aquatic products, and milk and coffee used by people in daily life.

[0003] The prior art is to sample food, then put the food sample in a test tube, and then allow the probe to enter the test tube and contact the food for testing. During the testing process, in order to ensure the accuracy of the test data, the probe needs to be replaced. However, in the prior art, food mold testing generally requires manual sampling, and secondary contamination may occur during the sampling process. In addition, the distribution of mold in the food is uneven. When sampling, it is easy to miss the location of mold distribution through manual visual observation, resulting in inaccurate sampling and thus affecting the results of mold testing.

[0004] In view of the above technical defects, a solution is now proposed. Summary of the invention

[0005] The purpose of the present invention is to extract features from sample analysis images through a feature recognition model to obtain sample category features, search for mold points in the sample analysis images, and select the mold points with the largest area as sampling points, further plan the movement trajectory of the multi-axis robot, adjust the servo motor according to the sampling range to adjust the distance between the cutting knives, and at the same time adjust the steering degree of the first drive motor to ensure the final sample quality.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a detection device for detecting mold in food, comprising a device shell, a detection probe and a sample analysis system, a detection port is opened on one side of the device shell, a sealing door panel is movably connected to the inner wall of the detection port, a sample table is fixedly provided on the inner wall of the device shell, a camera is fixedly provided on the top surface of the sample table, a multi-axis robot is fixedly provided at the center position of the device shell, a sampling mechanism is fixedly provided on the end surface of the output end of the multi-axis robot, a steering mechanism is fixedly provided on the inner wall of the device shell, a crushing cylinder is fixedly provided on the end surface of the steering mechanism, a horizontal bracket is fixedly provided on the inner wall of the device shell above the crushing cylinder, a crushing mechanism is fixedly provided on the bottom surface of the horizontal bracket, a first telescopic rod is fixedly provided on the bottom surface of the horizontal bracket, and the detection probe is fixedly provided on the bottom surface of the first telescopic rod.

[0007] Furthermore, the sampling mechanism includes a first drive motor, which is fixedly mounted on the end surface of the output end of the multi-axis robot. A horizontal plate is fixedly mounted on the bottom surface of the first drive motor. Two horizontal tracks located in the same straight line are opened on the bottom surface of the horizontal plate. Cutting components are fixedly mounted on the inner walls of the two horizontal tracks.

[0008] Furthermore, the cutting assembly includes a cutting knife and a servo motor, the servo motor is fixedly mounted on the outer surface of the horizontal track, a ball screw is connected to the outer side of the output end of the servo motor, the ball screw extends to the interior of the horizontal track, a slide is movably connected to the inner wall of the horizontal track, the slide is movably mounted on the outer surface of the ball screw, an extension shaft is fixedly mounted on the bottom end surface of the slide, and the cutting knife is fixedly mounted on the bottom end surface of the extension shaft.

[0009] Furthermore, the cross section of the cutting knife is a quarter ring structure, the bottom end surface of the cutting knife is provided with a cutting surface, and the two cutting knives are arranged opposite to each other.

[0010] Furthermore, the steering mechanism includes a steering motor, which is fixed to the bottom end of the inner wall of the device casing, and a steering support shaft is fixed to the end surface of the output end of the steering motor, and a rotating base is fixed to the end surface of the steering support shaft, and the grinding cylinder is detachably connected to the top surface of the rotating base.

[0011] Furthermore, the crushing mechanism includes a second electric telescopic rod, which is fixed to the bottom surface of the horizontal bracket, and a second driving motor is fixed to the bottom surface of the second electric telescopic rod. The output end of the second driving motor is connected to a driving support shaft, and the bottom surface of the driving support shaft is connected to a lateral pressure shaft, and the bottom surface of the lateral pressure shaft is fixed to a grinding hammer head, and the grinding hammer head is in active contact with the inner wall of the sample bottle.

[0012] Further, the sample analysis system includes an image acquisition unit, a feature extraction unit, a feature comparison unit and a sampling delineation unit;

[0013] The image acquisition unit is used to acquire the sample image on the sample plate through a camera, and obtain a clear sample analysis image after preprocessing the sample image, and send the sample analysis image to the feature extraction unit;

[0014] After the feature extraction unit obtains the sample analysis image, it obtains the sample standard comparison image corresponding to the sample analysis image from the network database, builds a category feature recognition model based on the convolutional neural network, extracts features from the sample analysis image through the feature recognition model to obtain sample category features, and simultaneously obtains the sample standard comparison image corresponding to the sample analysis image, and sends the sample standard comparison image and the sample analysis image to the feature comparison unit;

[0015] After the feature comparison unit obtains the sample standard comparison image and the sample analysis image, it searches for mold spots on the sample analysis image, selects the mold spot with the largest area as the sampling point, establishes a plane coordinate system based on the sample disk, obtains the real-time coordinates of the sampling point, defines the sampling range based on the distribution density of the mold spots and the thickness of the sample, and sends the real-time coordinates of the sampling point and the sampling range to the sampling definition unit;

[0016] The sampling demarcation unit is used to obtain and process the real-time coordinates of the sampling point and the sampling range, obtain the real-time coordinates of the end of the multi-axis robot and the real-time coordinates of the sampling point to plan the moving trajectory of the multi-axis robot, adjust the servo motor according to the sampling range to adjust the distance between the cutting knives, and adjust the steering degree of the first drive motor to ensure the final sample quality.

[0017] The present invention also provides a method for using a detection device for detecting mold in food, comprising the following steps:

[0018] Step 1: Get the sample to be tested. The area of ​​the sample to be tested is smaller than the cross section of the sample table but larger than the minimum cutting range of the cutting knife. Open the sealing door on the right side and place the sample to be tested on the sample table. Close the sealing door.

[0019] Step 2: Use a camera to take a picture of the sample to be tested on the sample stage and send it to the sample analysis system. After the sample picture is analyzed by the sample analysis system, the moving trajectory of the multi-axis robot is planned. The multi-axis robot is controlled by the central controller to move the sampling component above the sample stage so that the cutting knife is aligned with the sampling point position.

[0020] Step 2: According to the sampling range, the servo motor is adjusted to adjust the distance between the cutting knives. The central controller controls the multi-axis robot to drive the cutting knives downward to cut into the sample. After the cutting knives rise, the rotation angle of the first driving motor is adjusted to adjust the cutting position of the cutting knives. The circular target sample is obtained by multiple reciprocating cutting.

[0021] Step 3: After the last cutting is completed, the servo motor drives the two cutting knives to move relative to each other. After the target sample is clamped, the central controller controls the multi-axis robot to drive the target sample to move above the crushing cylinder. The central controller controls the servo motor to drive the two cutting knives to move in the opposite direction, and the target sample falls into the crushing cylinder.

[0022] Step 4: The central controller controls the second electric telescopic rod to drive the grinding hammer to move downward until the outer surface of the grinding hammer contacts the test sample and the inner wall of the crushing cylinder, and the second driving motor drives the grinding hammer to rotate in the crushing cylinder to grind the test sample into powder;

[0023] Step 5: The central controller controls the steering motor to drive the rotating base to rotate a fixed angle, so that the crushing cylinder moves to the bottom of the detection probe. The central controller controls the first electric telescopic rod to drive the detection probe to move into the crushing cylinder until it contacts the crushed test sample, completing the automated mold detection process.

[0024] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0025] 1. The detection device for detecting mold in food uses a camera to take a picture of a sample to be detected on a sample table and sends it to a sample analysis system. After the sample analysis system analyzes the sample picture, the moving trajectory of the multi-axis robot is planned. The multi-axis robot is controlled by a central controller to move the sampling component above the sample table so that the cutting knife is aligned with the sampling point position. The servo motor is adjusted according to the sampling range to adjust the distance between the cutting knives. The central controller controls the multi-axis robot to drive the cutting knife downward to cut into the sample. After the cutting knife rises, the cutting position of the cutting knife is adjusted by adjusting the rotation angle of the first drive motor. A circular target sample is obtained by multiple reciprocating cutting to complete the automatic sampling process.

[0026] 2. The detection device for detecting mold in food drives two cutting knives to move relative to each other through a servo motor. After clamping the target sample, the central controller controls the multi-axis robot to drive the target sample to move above the crushing cylinder. The central controller controls the servo motor to drive the two cutting knives to move in the opposite direction, and the target sample falls into the crushing cylinder. The second electric telescopic rod is controlled to drive the grinding hammer to move downward until the outer surface of the grinding hammer contacts the detection sample and the inner wall of the crushing cylinder. The grinding hammer is driven by the second drive motor to rotate in the crushing cylinder to grind the detection sample to powder, thereby completing the automatic sample preparation process.

[0027] 3. The detection device for detecting mold in food obtains the sample image on the sample plate through a camera, and builds a category feature recognition model based on a convolutional neural network. The feature recognition model extracts the features of the sample analysis image to obtain the sample category features, and simultaneously obtains the sample standard comparison image corresponding to the sample analysis image, searches for mold points in the sample analysis image, and selects the mold point with the largest area as the sampling point. The real-time coordinates of the end of the multi-axis robot and the real-time coordinates of the sampling point are obtained to plan the moving trajectory of the multi-axis robot, and the servo motor is adjusted according to the sampling range to adjust the distance between the cutting knives, and the steering degree of the first drive motor is adjusted to ensure the final sample quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 The overall external structure schematic diagram of the present invention is shown;

[0029] Figure 2 The overall internal structure schematic diagram of the present invention is shown;

[0030] Figure 3 The external structure schematic diagram of the sampling mechanism of the present invention is shown;

[0031] Figure 4 The schematic diagram of the structure of the sample analysis system of the present invention is shown;

[0032] Legend: 1. Device housing; 2. Sealing door panel; 3. Sample table; 4. Camera; 5. Multi-axis robot; 6. Crushing cylinder; 7. Horizontal bracket; 8. First telescopic rod; 9. First drive motor; 10. Horizontal plate; 11. Horizontal track; 12. Servo motor; 13. Ball screw; 14. Slide; 15. Extension support shaft; 16. Cutting knife; 17. Steering motor; 18. Steering support shaft; 19. Rotating base; 20. Second electric telescopic rod; 21 Second drive motor; 22. Drive support shaft; 23. Lateral pressure shaft; 24. Grinding hammer; 25. Detection probe. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] Embodiment 1:

[0035] like Figure 1-3 As shown, a detection device for detecting mold in food includes a device shell 1, a detection probe 25 and a sample analysis system. A detection port is opened on one side of the device shell 1, and a sealing door panel 2 is movably connected to the inner wall of the detection port. A sample table 3 is fixedly provided on the inner wall of the device shell 1, and a camera 4 is fixedly provided on the top surface of the sample table 3. A multi-axis robot 5 is fixedly provided at the center position of the device shell 1, and a sampling mechanism is fixedly provided on the end surface of the output end of the multi-axis robot 5. A steering mechanism is fixedly provided on the inner wall of the device shell 1, and a crushing cylinder 6 is fixedly provided on the end surface of the steering mechanism. A horizontal bracket 7 is fixedly provided on the inner wall of the device shell 1 above the crushing cylinder 6, and a crushing mechanism is fixedly provided on the bottom surface of the horizontal bracket 7. A first telescopic rod 8 is fixedly provided on the bottom surface of the horizontal bracket 7, and the detection probe 25 is fixedly provided on the bottom surface of the first telescopic rod 8.

[0036] The sampling mechanism includes a first drive motor 9, which is fixed on the end surface of the output end of the multi-axis robot 5. A horizontal plate 10 is fixed on the bottom surface of the first drive motor 9. Two horizontal rails 11 located in the same straight line are opened on the bottom surface of the horizontal plate 10. Cutting components are fixed on the inner walls of the two horizontal rails 11.

[0037] The cutting assembly includes a cutting knife 16 and a servo motor 12. The servo motor 12 is fixedly mounted on the outer surface of the horizontal track 11. A ball screw 13 is connected to the outer side of the output end of the servo motor 12. The ball screw 13 extends to the inside of the horizontal track 11. A slide 14 is movably connected to the inner wall of the horizontal track 11. The slide 14 is movably mounted on the outer surface of the ball screw 13. An extension support shaft 15 is fixedly mounted on the bottom end surface of the slide 14. The cutting knife 16 is fixedly mounted on the bottom end surface of the extension support shaft 15.

[0038] The cross section of the cutting blade 16 is a quarter ring structure, and a cutting surface is provided on the bottom surface of the cutting blade 16 . The two cutting blades 16 are arranged opposite to each other.

[0039] The steering mechanism includes a steering motor 17, which is fixed to the bottom end of the inner wall of the device casing. A steering support shaft 18 is fixed to the end surface of the output end of the steering motor 17, and a rotating base 19 is fixed to the end surface of the steering support shaft 18. The grinding cylinder 6 is detachably connected to the top surface of the rotating base 19.

[0040] The crushing mechanism includes a second electric telescopic rod 20, which is fixed on the bottom end surface of the horizontal bracket 7, and a second driving motor is fixed on the bottom end surface of the second electric telescopic rod 20, and the output end of the second driving motor is connected to a driving support shaft 21 of the second driving motor; 22, the driving support shaft 21 of the second driving motor; the bottom end surface of 22 is connected to a lateral pressing shaft 23, and a grinding hammer head 24 is fixed on the bottom end surface of the lateral pressing shaft 23, and the grinding hammer head 24 is in active contact with the inner wall of the sample bottle.

[0041] The working principle is as follows: obtain a sample to be tested, the area of ​​the sample to be tested is smaller than the cross section of the sample stage 3 but larger than the minimum cutting range of the cutting knife 16, open the right sealing door panel 2 and place the sample to be tested on the sample stage 3, and close the sealing door panel 2;

[0042] The camera 4 takes a picture of the sample to be tested on the sample stage 3 and sends it to the sample analysis system. After the sample analysis system analyzes the sample picture, the moving trajectory of the multi-axis robot 5 is planned. The central controller controls the multi-axis robot 5 to move the sampling assembly to the top of the sample stage 3 so that the cutting knife 16 is aligned with the sampling point position.

[0043] According to the sampling range, the servo motor 12 is adjusted to adjust the distance between the cutting knives 16. The central controller controls the multi-axis robot 5 to drive the cutting knives 16 to cut into the sample downward. After the cutting knives 16 rise, the cutting position of the cutting knives 16 is adjusted by adjusting the rotation angle of the first driving motor 9. The circular target sample is obtained by multiple reciprocating cutting.

[0044] After the last cutting is completed, the servo motor 12 drives the two cutting knives 16 to move relative to each other, and after the target sample is clamped, the central controller controls the multi-axis robot 5 to drive the target sample to move above the crushing cylinder 6, and the central controller controls the servo motor 12 to drive the two cutting knives 16 to move in the opposite direction, and the target sample falls into the crushing cylinder 6;

[0045] The central controller controls the second electric telescopic rod 20 to drive the grinding hammer 24 to move downward until the outer surface of the grinding hammer 24 contacts the test sample and the inner wall of the crushing cylinder 6, and the second driving motor drives the grinding hammer 24 to rotate in the crushing cylinder 6 to grind the test sample into powder;

[0046] The central controller controls the steering motor 17 to drive the rotating base 19 to rotate a fixed angle, so that the crushing cylinder 6 moves to under the detection probe 25. The central controller controls the first electric telescopic rod to drive the detection probe 25 to move into the crushing cylinder 6 until it contacts the crushed test sample, completing the automated mold detection process.

[0047] The central controller model is S7-200CN, which belongs to the disclosed prior art. Its position can be adjusted according to actual conditions, so it is not shown in the figure.

[0048] Embodiment 2:

[0049] like Figure 4 As shown, a detection device for detecting mold in food includes a device housing 1, a detection probe 25 and a sample analysis system, wherein the sample analysis system includes an image acquisition unit, a feature extraction unit, a feature comparison unit and a sampling delineation unit;

[0050] The image acquisition unit is used to acquire the sample image on the sample plate through the camera 4, and obtain a clear sample analysis image after preprocessing the sample image, and send the sample analysis image to the feature extraction unit;

[0051] After the feature extraction unit obtains the sample analysis image, it obtains the sample standard comparison image corresponding to the sample analysis image from the network database, builds a category feature recognition model based on the convolutional neural network, extracts features from the sample analysis image through the feature recognition model to obtain sample category features, and simultaneously obtains the sample standard comparison image corresponding to the sample analysis image, and sends the sample standard comparison image and the sample analysis image to the feature comparison unit;

[0052] The specific process of finding the sample standard comparison image is as follows:

[0053] S101, obtaining a plurality of sample standard comparison images of different food categories and initial samples of corresponding food category characteristics based on a network database as training data, and dividing the training data into three parts: a training set, a validation set, and a test set in a ratio of 8:1:1, wherein each sample is a food image, and a label corresponding to the sample indicates a category characteristic of the food image;

[0054] S102, modify the last fully connected layer of the convolutional neural network, keep the input unchanged, set the output to image data that has similarity with the reference data set, initialize the weights of the last layer, use the gradient descent algorithm for learning, retrain the entire network, and obtain an image recognition model;

[0055] S103, dividing the initial training set into multiple batches, and optimizing a portion of the image data selected from each batch of the initial training set, constructing an optimized training set based on the optimized image data and the unselected image data as a test set, evaluating the model effect, and outputting image information that is similar to the reference data set as a sample standard comparison image.

[0056] After the feature comparison unit obtains the sample standard comparison image and the sample analysis image, it searches for mold spots on the sample analysis image, selects the mold spot with the largest area as the sampling point, establishes a plane coordinate system based on the sample disk, obtains the real-time coordinates of the sampling point, defines the sampling range based on the distribution density of the mold spots and the thickness of the sample, and sends the real-time coordinates of the sampling point and the sampling range to the sampling definition unit;

[0057] The sampling demarcation unit is used to obtain and process the real-time coordinates of the sampling point and the sampling range, obtain the real-time coordinates of the end of the multi-axis robot 5 and the real-time coordinates of the sampling point to plan the movement trajectory of the multi-axis robot 5, adjust the servo motor 12 according to the sampling range to adjust the distance between the cutting knives 16, and adjust the steering degree of the first drive motor 9 to ensure the final sample quality.

[0058] The specific process of obtaining the real-time coordinates of the sampling points and the sampling range is as follows:

[0059] S201, obtaining a sample analysis image, wherein the sample analysis image includes a sample plane image and a sample side image, and performing image edge and contrast enhancement on the sample plane image by using a differential Laplace enhancement algorithm to obtain an enhanced image corresponding to the sample plane image;

[0060] S202, extracting image features from the enhanced image using a local binary pattern algorithm to obtain an image texture feature set, wherein the image texture features include a mold spot feature, a discoloration feature, and an abnormal texture roughness feature;

[0061] S203, establishing a plane coordinate system based on the sample disk, overlapping the sample plane image with the plane coordinate system, and then visualizing the image texture features, converting the image texture features into pigment points and displaying them on the plane coordinate system to obtain mold points, calculating and comparing the areas of the mold points, taking the mold point with the largest area as the sampling point, and obtaining the real-time coordinates of the sampling point based on the center point of the sampling point;

[0062] S204, obtaining multiple sample side images, calculating the disparity map through a stereo matching algorithm, obtaining the sample thickness data H0, and calculating the sampling range R based on the preset detection sample volume value V:

[0063] The present invention obtains the sample image on the sample tray through the camera 4, and constructs a category feature recognition model based on the convolutional neural network, extracts the features of the sample analysis image through the feature recognition model to obtain the sample category features, and simultaneously obtains the sample standard comparison image corresponding to the sample analysis image, searches for mold points in the sample analysis image, and selects the mold point with the largest area as the sampling point, obtains the real-time coordinates of the end of the multi-axis robot 5 and the real-time coordinates of the sampling point to plan the moving trajectory of the multi-axis robot 5, adjusts the servo motor 12 according to the sampling range to adjust the distance between the cutting knives 16, and adjusts the steering degree of the first drive motor 9 to ensure the final sample quality.

[0064] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by technical personnel in this field for each group of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0065] The above formulas are all dimensionless and numerical calculations. The formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0066] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A detection device for detecting mold in food, comprising a device housing (1), a detection probe (25) and a sample analysis system, characterized in that: A detection port is provided on one side of the device housing (1), and a sealing door panel (2) is movably connected to the inner wall of the detection port. A sample stage (3) is fixedly provided on the inner wall of the device housing (1), and a camera (4) is fixedly provided on the top surface of the sample stage (3). A multi-axis robot (5) is fixedly provided at the center of the device housing (1), and a sampling mechanism is fixedly provided on the end surface of the output end of the multi-axis robot (5). A steering mechanism is fixedly provided on the inner wall of the device housing (1), and a crushing cylinder (6) is fixedly provided on the end surface of the steering mechanism. A horizontal bracket (7) is fixedly provided on the inner wall of the device housing (1) above the crushing cylinder (6), and a crushing mechanism is fixedly provided on the bottom surface of the horizontal bracket (7). A first telescopic rod (8) is fixedly provided on the bottom surface of the horizontal bracket (7), and the detection probe (25) is fixedly provided on the bottom surface of the first telescopic rod (8).

2. The detection device for detecting mold in food according to claim 1, characterized in that: The sampling mechanism comprises a first drive motor (9), the first drive motor (9) is fixedly mounted on the end surface of the output end of the multi-axis robot (5), a horizontal plate (10) is fixedly mounted on the bottom surface of the first drive motor (9), two horizontal rails (11) located in the same straight line are arranged on the bottom surface of the horizontal plate (10), and cutting components are fixedly mounted on the inner walls of the two horizontal rails (11).

3. The detection device for detecting mold in food according to claim 2, characterized in that: The cutting assembly comprises a cutting blade (16) and a servo motor (12), wherein the servo motor (12) is fixedly arranged on the outer surface of a horizontal track (11), a ball screw (13) is connected to the outer side of an output end of the servo motor (12), the ball screw (13) extends to the inside of the horizontal track (11), a slide table (14) is movably connected to the inner wall of the horizontal track (11), the slide table (14) is movably sleeved on the outer surface of the ball screw (13), an extension support shaft (15) is fixedly arranged on the bottom end surface of the slide table (14), and the cutting blade (16) is fixedly arranged on the bottom end surface of the extension support shaft (15).

4. The detection device for detecting mold in food according to claim 3, characterized in that: The cross section of the cutting knife (16) is a quarter ring structure, the bottom end surface of the cutting knife (16) is provided with a cutting surface, and the two cutting knives (16) are arranged opposite to each other.

5. The detection device for detecting mold in food according to claim 1, characterized in that: The steering mechanism comprises a steering motor (17), the steering motor (17) is fixedly arranged at the bottom end of the inner wall of the device housing, a steering support shaft (18) is fixedly arranged on the end surface of the output end of the steering motor (17), a rotating base (19) is fixedly arranged on the end surface of the steering support shaft (18), and the grinding cylinder (6) is detachably connected to the top surface of the rotating base (19).

6. The detection device for detecting mold in food according to claim 1, characterized in that: The pulverizing mechanism comprises a second electric telescopic rod (20), the second electric telescopic rod (20) is fixedly arranged on the bottom end surface of the horizontal support (7), a second driving motor is fixedly arranged on the bottom end surface of the second electric telescopic rod (20), an output end of the second driving motor is connected to a driving support shaft (22), a bottom end surface of the driving support shaft (22) is connected to a lateral pressure shaft (23), a bottom end surface of the lateral pressure shaft (23) is fixedly arranged with a grinding hammer head (24), and the grinding hammer head (24) is in active contact with the inner wall of the sample bottle.

7. The detection device for detecting mold in food according to claim 1, characterized in that: The sample analysis system comprises an image acquisition unit, a feature extraction unit, a feature comparison unit and a sampling delineation unit; The image acquisition unit is used to acquire a sample image on the sample plate through a camera (4), and obtain a clear sample analysis image after preprocessing the sample image, and send the sample analysis image to the feature extraction unit; After the feature extraction unit obtains the sample analysis image, it obtains the sample standard comparison image corresponding to the sample analysis image from the network database, builds a category feature recognition model based on the convolutional neural network, extracts features from the sample analysis image through the feature recognition model to obtain sample category features, and simultaneously obtains the sample standard comparison image corresponding to the sample analysis image, and sends the sample standard comparison image and the sample analysis image to the feature comparison unit; After the feature comparison unit obtains the sample standard comparison image and the sample analysis image, it searches for mold spots on the sample analysis image, selects the mold spot with the largest area as the sampling point, establishes a plane coordinate system based on the sample disk, obtains the real-time coordinates of the sampling point, defines the sampling range based on the distribution density of the mold spots and the thickness of the sample, and sends the real-time coordinates of the sampling point and the sampling range to the sampling definition unit; The sampling demarcation unit is used to obtain and process the real-time coordinates of the sampling point and the sampling range, obtain the real-time coordinates of the end of the multi-axis robot (5) and the real-time coordinates of the sampling point to plan the movement trajectory of the multi-axis robot (5), adjust the servo motor (12) according to the sampling range to adjust the distance between the cutting knives (16), and adjust the steering degree of the first drive motor (9) to ensure the final sample quality.

8. The detection device for detecting mold in food according to claim 1, characterized in that: The specific process of finding the sample standard comparison image is as follows: S101, obtaining a plurality of sample standard comparison images of different food categories and initial samples of corresponding food category characteristics based on a network database as training data, and dividing the training data into three parts: a training set, a validation set, and a test set in a ratio of 8:1:1, wherein each sample is a food image, and a label corresponding to the sample indicates a category characteristic of the food image; S102, modify the last fully connected layer of the convolutional neural network, keep the input unchanged, set the output to image data that has similarity with the reference data set, initialize the weights of the last layer, use the gradient descent algorithm for learning, retrain the entire network, and obtain an image recognition model; S103, dividing the initial training set into multiple batches, and optimizing a portion of the image data selected from each batch of the initial training set, constructing an optimized training set based on the optimized image data and the unselected image data as a test set, evaluating the model effect, and outputting image information that is similar to the reference data set as a sample standard comparison image.

9. The method for using the detection device for detecting mold in food according to claim 1, characterized in that: The following steps are involved: Step 1: Obtain a sample to be tested, the area of ​​the sample to be tested is smaller than the cross section of the sample table (3) but larger than the minimum cutting range of the cutting knife (16), open the right sealing door panel (2), place the sample to be tested on the sample table (3), and close the sealing door panel (2); Step 2: Using a camera (4) to take a picture of a sample to be tested on the sample stage (3) and sending it to a sample analysis system; after the sample analysis system analyzes the sample picture, a moving trajectory of the multi-axis robot (5) is planned; and the multi-axis robot (5) is controlled by a central controller to move the sampling assembly to the top of the sample stage (3) so that the cutting knife (16) is aligned with the sampling point position; Step 2: According to the sampling range, the servo motor (12) is adjusted to adjust the distance between the cutting knives (16); the central controller controls the multi-axis robot (5) to drive the cutting knives (16) to cut downward into the sample; after the cutting knives (16) rise, the rotation angle of the first driving motor (9) is adjusted to adjust the cutting position of the cutting knives (16); and a circular target sample is obtained by reciprocating cutting for multiple times; Step 3: After the last cutting is completed, the servo motor (12) drives the two cutting knives (16) to move relative to each other, and after the target sample is clamped, the central controller controls the multi-axis robot (5) to drive the target sample to move above the crushing cylinder (6), and the central controller controls the servo motor (12) to drive the two cutting knives (16) to move in the opposite direction, and the target sample falls into the crushing cylinder (6); Step 4: The central controller controls the second electric telescopic rod (20) to drive the grinding hammer (24) to move downward until the outer surface of the grinding hammer (24) contacts the test sample and the inner wall of the crushing cylinder (6), and the second driving motor drives the grinding hammer (24) to rotate in the crushing cylinder (6) to grind the test sample into powder; Step 5: The central controller controls the steering motor (17) to drive the rotating base (19) to rotate at a fixed angle, so that the crushing cylinder (6) moves to below the detection probe (25). The central controller controls the first electric telescopic rod to drive the detection probe (25) to move into the crushing cylinder (6) until it contacts the crushed detection sample, thereby completing the automated mold detection process.

Images