Product tracking system

Abstract

The application provides a product tracking system, and relates to the technical field of product tracking. The product tracking system comprises a workbench, a conveying device and an encoder. The conveying device is arranged on the workbench and is configured to carry and convey products. The encoder is arranged on the conveying device. When the conveying device conveys the products to pass through a first detection position, the encoder reads an initial encoding value and sends the initial encoding value to the product tracking device. When the conveying device conveys the products to pass through a second detection position, the encoder reads a current encoding value and sends the current encoding value to the product tracking device. When the products are detected, the application can automatically screen out corresponding data of lost, missed and defective products, so that batch dislocation of the products during detection is avoided, and the accuracy and yield of product detection are improved.

Description

Technical Field

[0001] This application relates to the field of product tracking technology, and more particularly to a product tracking system. Background Technology

[0002] In automated production lines, accurate product tracking and inspection are crucial for improving production efficiency and ensuring product quality. Most existing product tracking systems can only achieve simple location tracking and cannot perform real-time quality inspection. Furthermore, while some inspection systems can detect product quality, they are often separate from the tracking system, leading to difficulties in data synchronization and impacting production efficiency and accuracy. Therefore, this application aims to provide a system that can both track product location and detect product quality in real time, thereby addressing the shortcomings of existing technologies. Utility Model Content

[0003] This application provides a product tracking system that can track the location of products in real time and detect product quality, while ensuring data synchronization and accuracy.

[0004] To achieve the above objectives, this application provides a product tracking system, comprising:

[0005] A conveying device is set on a workbench and configured to carry and convey products.

[0006] An encoder is mounted on a conveying device.

[0007] Specifically, when the conveying device transports the product past the first detection position, the encoder reads the initial encoding value and sends the initial encoding value to the product tracking device; when the conveying device transports the product past the second detection position, the encoder reads the current encoding value and sends the current encoding value to the product tracking device.

[0008] The product tracking system of this application embodiment can accurately and automatically filter out the corresponding data of lost, missed, and defective products during product inspection, thereby avoiding batch misalignment of products during inspection and improving the accuracy and yield of product inspection.

[0009] Other features and advantages of this application will be described in detail in the following detailed description section. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] To gain a more complete understanding of this application and its beneficial effects, the following description will be provided in conjunction with the accompanying drawings, wherein the same reference numerals in the following description denote the same parts.

[0012] Figure 1 This is the structure of a product tracking system according to an embodiment of this application.

[0013] Figure 2 This is the second structure of the product tracking system according to an embodiment of this application.

[0014] Figure 3 This is a schematic diagram of the imaging device according to an embodiment of this application.

[0015] Figure 4 This is a schematic diagram of the product tracking device according to an embodiment of this application.

[0016] Figure 5 This is a schematic diagram illustrating a scenario where the first and second sub-detection devices of the product in this application are lost.

[0017] Figure 6 This is a schematic diagram illustrating a scenario in which the product of this application embodiment is missed by the second sub-detection device.

[0018] Figure 7 This is a schematic diagram illustrating a scenario where the product in an embodiment of this application is missed by the first sub-detection device.

[0019] Figure 8 This is a schematic diagram illustrating a scenario in which the third and fourth sub-detection devices of the product in this application are lost during detection.

[0020] Figure 9 This is a schematic diagram illustrating a scenario where the product in an embodiment of this application is missed by the fourth sub-detection device.

[0021] Figure 10 This is a schematic diagram illustrating a scenario where the product in an embodiment of this application is missed by a third sub-detection device.

[0022] Figure 11 This is the overall flow of the product tracking method in this application embodiment.

[0023] Figure 12 This is the overall process for obtaining product testing results in this application embodiment.

[0024] Figure 13 This is a first example of a probability heatmap in an embodiment of this application.

[0025] Figure 14 This is a second example of a probability heatmap in an embodiment of this application.

[0026] Figure 15The structure of the image segmentation model in this application embodiment is shown.

[0027] Figure 16 This describes the training process of the image segmentation model in an embodiment of this application.

[0028] Figure 17 The following is a detailed flowchart of the product tracking method according to an embodiment of this application.

[0029] Figure 18 This is a structural diagram of a product tracking device according to an embodiment of this application.

[0030] Explanation of reference numerals in the attached figures:

[0031] 10-Conveying device; 101-Conveyor belt; 102-Glass turntable; 20-First detection device; 201-First sub-detection device; 202-Third sub-detection device; 30-Second detection device; 301-Second sub-detection device; 302-Fourth sub-detection device; 40-Product tracking device; 41-Image acquisition card; 42-Image segmentation model; 43-Vision processor; 431-ROI extraction module; 50-Imaging device; 501-First imaging device; 502-Second imaging device; 51-Illumination module; 52-Sensor component ; 53-Analog-to-Digital Conversion Module; 54-Image Capture Module; 60-Motion Mechanism; 70-Storage Bin; 80-Vibrating Feeder; 100-First Unloading Mechanism; 200-Second Unloading Mechanism; 300-Third Unloading Mechanism; 400-Workbench; 500-Encoder; 5001 First Encoder; 5002 Second Encoder; 600-Display Device; 700-GPU Accelerated Graphics Card; 801-Memory; 802-Processor; C-First Sub-Detection Position; D-Second Sub-Detection Position; E-Third Sub-Detection Position; F-Fourth Sub-Detection Position. Detailed Implementation

[0032] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the protection scope of this application.

[0033] Please see Figure 1 and Figure 2 , Figure 1 This illustration shows the structure of a product tracking system according to an embodiment of this application. Figure 2This illustration shows a second structure of the product tracking system according to an embodiment of this application. The product tracking system of this application includes a workbench 400, a conveying device 10, an encoder 500, and a product tracking device 40. The conveying device 10 is disposed on the workbench 400 and configured to carry and convey products; the encoder 500 is disposed on the conveying device 10; wherein, when the conveying device 10 conveys a product past a first detection position, the encoder 500 reads an initial encoding value and sends the initial encoding value to the product tracking device 40; when the conveying device 10 conveys a product past a second detection position, the encoder 500 reads the current encoding value and sends the current encoding value to the product tracking device 40. The product tracking device 40 is used to execute the product tracking method of the embodiment of the application to track and detect each product.

[0034] The product tracking system also includes a first detection device 20 and a second detection device 30. The first detection device 20 includes a first sub-detection device 201, and the second detection device 30 includes a second sub-detection device 301. The first detection position includes a first sub-detection position C, and the second detection position includes a second sub-detection position D. The conveying device 10 includes a conveyor belt 101.

[0035] The first sub-detection device 201 and the second sub-detection device 301 are respectively disposed on one side of the conveyor belt 101, and the second sub-detection position D is located downstream of the first sub-detection position C; wherein, the first sub-detection device 201 is used to send the first encoded value read by the first encoder 5001 to the product tracking device 40 when the conveyor belt 101 conveys the product past the first sub-detection position C; the second sub-detection device 301 is used to send the first current encoded value read by the first encoder 5001 to the product tracking device 40 when the conveyor belt 101 conveys the product past the second sub-detection position D.

[0036] For example, the product can be a sealing ring formed of liquid silicone rubber (LSR). Liquid silicone rubber is a liquid rubber material polymerized from silane monomers, possessing excellent heat resistance, cold resistance, and chemical resistance, and is commonly used in mold making, sealing materials, and medical devices. It is understood that the product can also be other types of industrial products, and this application embodiment does not specifically limit this. In some embodiments, the product tracking system further includes an imaging device 50, which is used to acquire an original image of the product based on the current shooting point at a first detection location.

[0037] For example, the imaging device 50 may include a first imaging device 501, which is disposed at a first sub-detection position C and faces the product.

[0038] Please see Figure 3 , Figure 3 This is a schematic diagram of the imaging device 50 according to an embodiment of this application. The imaging device 50 may include an illumination module 51, a sensor component 52, an analog-to-digital conversion module 53, and an image capture module 54 connected in sequence. The sensor component 52 may be a charge-coupled device (CCD) or a complementary metal-oxide-semiconductor (CMOS) device. The analog-to-digital conversion module 53 is an A / D converter.

[0039] Please see Figure 4 , Figure 4 This is a schematic diagram of the product tracking device 40 according to an embodiment of this application. The product tracking device 40 provided in this embodiment specifically includes an image acquisition card 41, an image segmentation model 42, and a vision processor 43. The image acquisition card 41 is connected to the imaging device 50 and is used to convert the output signal of the imaging device 50 into a digital signal. The image segmentation model 42 adopts a fully convolutional network model and is trained and validated using images of labeled sample products. The image segmentation model 42 is used to extract a probability heatmap of the original image. The vision processor 43 is connected to both the image acquisition card 41 and the image segmentation model 42. The vision processor 43 is used to determine the target region of interest corresponding to each product in the original image, and based on each target region of interest and the probability heatmap, to determine the detection result of each product. The detection result is used to characterize whether the product is defective.

[0040] For example, the vision processor 43 can be a host computer, a PLC (Programmable Logic Controller), or a robot, and this application embodiment does not specifically limit it.

[0041] In some embodiments, the visual processor 43 specifically includes:

[0042] ROI extraction module 431 is used to determine the target region of interest corresponding to each product in the original image;

[0043] The detection module is used to determine the detection result of each product based on the region of interest and probability heatmap of each target. The detection result is used to characterize whether the product is defective.

[0044] In some embodiments, the detection module is specifically used for:

[0045] Based on the probability heatmap and the region of interest for each target, extract the target region corresponding to each product;

[0046] The detection result for each product is determined based on the geometric features of each target region.

[0047] In some embodiments, the detection module is specifically used for:

[0048] The probability heatmap is binarized using a preset grayscale threshold to determine the first image;

[0049] Morphological filtering is applied to the first image to determine the second image;

[0050] Extract the target region corresponding to each target region of interest from the second image.

[0051] In some embodiments, the detection module is specifically used for:

[0052] If the geometric features of the target area meet the preset conditions, then it is determined whether the corresponding product is defective.

[0053] In some embodiments, the ROI extraction module 431 is specifically used for:

[0054] Determine the template image corresponding to the product to be tested. The template image includes the initial region of interest for each product.

[0055] Register the template image and the original image to determine the positioning deviation of the product under test;

[0056] The target region of interest is obtained by correcting each initial region of interest based on the positioning deviation.

[0057] In some embodiments, the ROI extraction module 431 is specifically used for:

[0058] Determine the positional information of the standard product outline in the template image;

[0059] Based on the positional information of the current contour of the product under test in the original image, and the positional information of the contour of the standard product, an affine transformation matrix is ​​determined to characterize the positioning deviation.

[0060] In some embodiments, the ROI extraction module 431 is specifically used for:

[0061] By keeping the size of each initial region of interest unchanged, the position of each initial region of interest is corrected based on the affine transformation matrix to obtain the target region of interest.

[0062] In some embodiments, the visual processor 43 is specifically used for:

[0063] The original image is input into the image segmentation model 42, and the probability heatmap returned by the image segmentation model 42 is received.

[0064] In some embodiments, the product tracking system further includes a display device 600 connected to the vision processor 43 for displaying the detection results for each product.

[0065] In some embodiments, the product tracking system further includes a GPU-accelerated graphics card 700 connected between the vision processor 43 and the image segmentation model 42 to accelerate the image processing of the vision processor 43 and the image segmentation model 42.

[0066] In some embodiments, the product tracking system further includes a storage bin 70 and a vibratory feeder 80 disposed on a workbench 400. The vibratory feeder 80 is mounted on one side of the storage bin 70, and the outlet of the storage bin 70 extends into the vibratory feeder 80. The storage bin 70 provides a stable supply of material. When the vibratory feeder 80 is low on material, a smart sensor detects the material status in the vibratory feeder 80. When material is low, the smart sensor reports the material (i.e., product) status, and the product tracking device 40 triggers the storage bin 70 to operate, sending material from the storage bin 70 to the vibratory feeder 80. If the material is insufficient, the smart sensor in the storage bin 70 will not detect any material and will send a feedback to the PLC. The PLC will control an indicator light to issue a material shortage message, reminding manual replenishment. The outlet of the vibratory feeder 80 is connected to a conveyor 10, which transports the material to the next process.

[0067] For example, the conveying device 10 includes a conveyor belt 101, a first detection position includes a first sub-detection position C, a second detection position includes a second sub-detection position D, a first detection device 20 includes a first sub-detection device 201 located at the first sub-detection position C, and a second detection device 30 includes a second sub-detection device 301 located at the second sub-detection position D. The conveyor belt 101 is used to sequentially transport products to the first sub-detection position C and the second sub-detection position D. The second sub-detection device 301 is located downstream of the first sub-detection device 201. The first sub-detection device 201 and the second sub-detection device 301 may be located on the same side or different sides of the conveyor belt 101, without particular limitation.

[0068] The encoder 500 includes a first encoder 5001. When the conveyor belt 101 transports the product past the first sub-detection position C, the first encoder 5001 reads the first encoded value and sends the first encoded value to the product tracking device 40. When the conveyor belt 101 transports the product past the second sub-detection position D, the first encoder 5001 reads the first current encoded value and sends the first current encoded value to the product tracking device 40.

[0069] In this embodiment of the application, the working process of the product tracking system includes:

[0070] Obtain the initial encoding sequence, which includes multiple initial encoding values. These multiple initial encoding values ​​are determined when each product passes through the first detection position.

[0071] Obtain the current code value. The current code value is determined when each product passes through the second detection position. Along the product conveying direction, the second detection position is located after the first detection position.

[0072] The distance difference between the current encoded value and each of the initial encoded values ​​in the initial encoded sequence is determined by comparing the current encoded value with each of the initial encoded values.

[0073] Based on the distance difference and the target distance, the tracking result of the corresponding product is determined. The target distance is the distance between the first detection position and the second detection position.

[0074] For example, conveyor belt 101 conveys the product to a first sub-detection position C. Upon sensing the product's arrival at sub-detection position C, first sub-detection device 201 generates a first coded signal and sends it to controller 40. Controller 40 responds to the first coded signal to obtain a first coded sequence, which includes multiple first coded values. Similarly, conveyor belt 101 conveys the product from sub-detection position C to a second sub-detection position D. Upon sensing the product's arrival at second sub-detection position D, second sub-detection device 301 generates a first current coded signal. Controller 40 executes the product tracking method of this embodiment based on the multiple first coded signals in the first coded sequence and the first current coded signal to determine the product tracking result.

[0075] Furthermore, the conveyor belt 101 transports the product to the first sub-detection position C. After sensing the product's arrival at the first sub-detection position C, the first sub-detection device 201 generates a positioning signal and sends it to the controller 40. It then controls the imaging device 50 to reach the current image capture point. Once the imaging device 50 reaches the current image capture point, the product tracking device 40 generates an image capture command to trigger the imaging device 50 to capture an image. The image capture command includes a point marker representing the shooting angle and position of the imaging device 50. The imaging device 50 captures the original image of the product at the current image capture point in response to the image capture command. The product tracking device 40 executes the product tracking method of this embodiment based on the first encoding sequence, the original image, and the first current encoding value to determine the product tracking result and detect whether the product is defective.

[0076] When determining the tracking result of a product, a first distance difference is determined between the first current code value and multiple first code values ​​in the first code sequence. Based on this first distance difference and the first target distance, the tracking result of the corresponding product is determined. The first target distance is the distance between the first sub-detection position C and the second sub-detection position D. If the first distance difference is not equal to the first target distance, the tracking of the corresponding product is determined to be abnormal. If the first distance difference is equal to the first target distance, the tracking of the corresponding product is determined to be normal.

[0077] The product tracking anomalies include situations where the product may be lost (i.e., missing) before reaching the second sub-detection position D, or missed being detected at the second sub-detection position D. Alternatively, the product may be missed being detected before the first sub-detection position C. The first sub-detection device 201 may include a visual detector; a missed detection at the first sub-detection position C could be due to a missed detection by the visual detector (e.g., the first imaging device 501 failing to take a picture). The second detection device 30 may be a blow-through photoelectric sensor; a missed detection at the second sub-detection position D could be due to a missed sensing by the blow-through photoelectric sensor, etc.

[0078] Normal product tracking conditions include: the product is detected when it reaches both the first sub-detection position C and the second sub-detection position D.

[0079] In some embodiments, the product tracking system further includes a first unloading mechanism 100. The first unloading mechanism 100 is located along the product conveying direction and between the second detection device 30 and the equipment in the next process step. After the controller 40 detects an abnormal product using the product tracking method of this embodiment based on the first encoded sequence, the original image, and the first current encoded value, the controller 40 issues a product abnormality command and controls the first unloading mechanism 100 to push the abnormal product out of the unloading port corresponding to the first unloading mechanism 100. The abnormal product includes products with product tracking abnormalities and products with abnormal detection results.

[0080] For example, after determining that the product tracking result is abnormal, the controller 40 issues a product abnormality instruction and controls the first unloading mechanism 100 to push the abnormal product out of the unloading port corresponding to the first unloading mechanism 100.

[0081] After confirming that the product tracking result is normal, the controller determines whether to reject the product based on the detection results of the products with normal tracking results. The detection results are used to characterize whether the product is defective. It is understood that when a product passes through the first sub-detection position C and is detected by the first sub-detection device 201, the controller 40 acquires the first code value of the product. If the product is simultaneously captured by the first imaging device 501, the first imaging device 501 transmits the original image of the product to the controller 40. At this time, the controller 40 analyzes the detection results based on the original image of the product and binds and saves the detection results with the first code value of the product. When the product passes through the first sub-detection position C and reaches the second sub-detection position D, the second sub-detection device 301 detects the product. The controller 40 reads the first current code value of the product and compares the first current code value with the first code value to obtain a first distance difference. If the first distance difference is equal to the first target distance, and the product detection result is defective, the controller 40 issues a product abnormality command and controls the first unloading mechanism 100 to push the abnormal product out of the unloading port corresponding to the first unloading mechanism 100.

[0082] If the first distance difference is equal to the first target distance, and the product test result is good, the product is transported to the next process for a second test to further determine whether the product is defective, thereby further improving the product yield.

[0083] In some embodiments, the conveying device 10 further includes a glass turntable 102, with one end of the conveyor belt 101 disposed at the discharge port of the vibrating plate 80 and the other end disposed at the inlet of the glass turntable 102.

[0084] In some embodiments, the first detection device 20 further includes a third sub-detection device 202, the second detection device 30 further includes a fourth sub-detection device 302, the first detection position further includes a third sub-detection position E, and the second detection position further includes a fourth sub-detection position F; the third sub-detection device 202 and the fourth sub-detection device 302 are disposed on the outer periphery of the glass turntable 102, and the fourth sub-detection device 302 is located downstream of the third sub-detection device 202. The third sub-detection device 202 is used to send the second encoded value read by the second encoder 5002 to the product tracking device 40 when the glass turntable 102 conveys the product past the third sub-detection position E; the fourth sub-detection device 302 is used to send the second current encoded value read by the second encoder 5002 to the product tracking device 40 when the glass turntable 102 conveys the product past the fourth sub-detection position F.

[0085] In some embodiments, the product tracking system further includes an imaging device 50, which may include a second imaging device 502, disposed at a third sub-detection position E and facing the product. A second forming device is used at the third sub-detection position E to acquire an original image of the product based on the current shooting point. The shooting principle of the second camera 90 is the same as that of the first imaging device 501, and will not be described in detail here. The controller 40 analyzes the original image captured by the second imaging device 502 to determine whether the current product is defective, classifying good and defective products to improve product yield.

[0086] In some embodiments, the product tracking system further includes a glass turntable 102. One end of the conveying device 10 is disposed at the discharge port of the vibratory feeder 80, and the other end is disposed at the inlet of the glass turntable 102.

[0087] In this embodiment, the product tracking system further includes the following steps: the glass turntable 102 transports the product to a third sub-detection position E. Upon sensing the product's arrival at the third sub-detection position E, the third detection device generates a second coded signal and sends this second coded signal to the controller 40. The controller 40 responds to the second coded signal to obtain a second coded sequence, which includes multiple second coded values. Similarly, the glass turntable 102 transports the product from the third sub-detection position E to a fourth sub-detection position F. Upon sensing the product's arrival at the fourth sub-detection position F, the fourth detection device generates a second current coded signal. The controller 40 executes the product tracking method of this embodiment based on the multiple second coded signals and the second current coded signal in the second coded sequence to determine the secondary tracking result of the product.

[0088] If the second distance difference is not equal to the second target distance, the corresponding product tracking is determined to be abnormal. The abnormal product may have been lost before reaching the third sub-detection position E or missed by the third sub-detection position E. If the second distance difference is equal to the second target distance, the corresponding product tracking is determined to be normal. The normal product is detected when it reaches both the third sub-detection position E and the fourth sub-detection position F. The controller 40 executes the product tracking method of this application embodiment based on multiple second encoding values ​​in the second encoding sequence, the original image, and the second current encoding value, and performs a second determination on whether the product is defective, so as to further improve the product yield.

[0089] The product tracking anomalies also include situations where the product might be lost (i.e., missing) before reaching the fourth sub-detection position F, or missed during detection at the fourth sub-detection position F. Alternatively, the product might be missed before reaching the third sub-detection position E. The third sub-detection device 202 includes a visual detector; a missed detection at the third sub-detection position E could be due to a missed detection by the visual detector (e.g., the second camera 90 missing an image). The fourth sub-detection device 302 can be a blow-through photoelectric sensor; a missed detection at the fourth sub-detection position F could be due to a missed sensing by the blow-through photoelectric sensor, etc.

[0090] Normal product tracking conditions include: the product is detected when it reaches both the third sub-detection position E and the fourth sub-detection position F.

[0091] In some embodiments, the product tracking system further includes a second unloading mechanism 200. The second unloading mechanism 200 is located on the outer periphery of the glass turntable 102. When the controller 40 determines a product to be good, the controller 40 issues a product normal command, controlling the second unloading mechanism 200 to push the good product out of the unloading port corresponding to the second unloading mechanism 200. Defective products will automatically flow to the next unloading port. Therefore, even if the controller 40 issues a product normal command, but the second unloading mechanism 200 does not execute the pushing command, the system will only misclassify good products as defective products, and will not misclassify defective products as good products, thereby further improving the product yield.

[0092] For example, after determining that the product tracking result is normal, the product flow is determined based on the detection result of the product with normal tracking result. The detection result is used to characterize whether there is a defect in the product. It can be understood that when the product passes through the third sub-detection position E and the third sub-detection device 202 detects the product, the controller 40 obtains the second code value of the product. If the product is also captured by the second camera 90, the second camera 90 transmits the original image of the product to the controller 40. At this time, the controller 40 analyzes the detection result based on the original image of the product and binds and saves the detection result with the second code value of the product. When the product passes through the third sub-detection position E and reaches the fourth sub-detection position F, the fourth sub-detection device 302 detects the product. The controller 40 reads the second current code value of the product and compares the second current code value with the second code value to obtain the second distance difference. If the second distance difference is equal to the second target distance and the product detection result is good, the controller 40 issues a product normal command and controls the second unloading mechanism 200 to push the good product out from the unloading port corresponding to the second unloading mechanism 200. If the second distance difference is equal to the second target distance, and the product test result is a defective product, the controller 40 controls the defective product to flow directly to the next discharge port for discharge.

[0093] To explain the product tracking principle more clearly, the following two examples illustrate the tracking principle when the product is being transported in the conveyor device 10.

[0094] Please see Figure 2 and Figure 3 , Figure 2 This is a schematic diagram of a scenario in which a product is detected as lost in the first sub-detection device 201, according to an embodiment of this application. Figure 3 This application embodiment illustrates a scenario where a product is detected by the first sub-detection device 201 but missed by the second sub-detection device 301. Wherein, Figure 2 and Figure 3 S1, S2, and S3 represent the various stages of product transport. S1 is the first stage, S2 is the second stage, and S3 is the third stage. The right-pointing arrow indicates the direction of product transport.

[0095] Specifically, the products include product 1, product 2, ..., product n. Assume:

[0096] Please see Figure 2 When product 1 passes the first sub-detection position C, encoder 500 reads the first encoded value and records it as C1. Also, product 1 is lost during transmission at the first sub-detection position C. Alternatively, please refer to... Figure 3 If product 1 is detected when it passes through the first sub-detection position C, but is missed when it passes through the second sub-detection position D (missed detection), then no record is made.

[0097] Please see Figure 2 and Figure 3 When product 2 passes through the first sub-detection position C, encoder 500 reads the first encoded value and records it as C2, and when product 2 passes through the second sub-detection position D, encoder 500 reads the first current encoded value and records it as D2.

[0098] Please see Figure 2 and Figure 3 When product n passes the first sub-detection position C, encoder 500 reads the first encoded value and records it as C. n When product n passes the second sub-detection position D, encoder 500 reads the first current encoded value and records it as D. n .

[0099] The first encoded values ​​obtained by the encoder 500 constitute the first encoded sequence, for example, C1, C2...C n .

[0100] For example, assuming the first target distance is L1 and D2 is the first current encoded value, the current first encoded sequence is C1 and C2. The first current encoded value D2 is compared one by one with C1 and C2 in the first encoded sequence, where the distance difference between D2 and C2 is |D2-C2|=L1, and the distance difference between D2 and C1 is |D2-C1|>L1. As can be seen from this, if product 1 shows a tracking anomaly and product 2 does not, the system automatically marks product 1 as a product with an abnormal tracking result and product 2 as a product with a normal tracking result. Based on the above judgment result, the controller 40 issues a tracking anomaly command for product 1, and pushes product 1 out through the first unloading mechanism 100 at the unloading port corresponding to the first unloading mechanism 100 (that is, to remove the abnormal product), and transfers product 2 to the next process (for example, to the glass turntable 102 for secondary detection). At the same time, the current parameters (C1, C2 in the first encoding sequence and the first current encoding value D2) are cleared to avoid a large amount of data affecting the system's operating speed.

[0101] Second example:

[0102] Please see Figure 4 , Figure 4 This application embodiment illustrates a scenario where the product is missed by the first sub-detection device 201. Wherein, Figure 4 S1, S2, and S3 represent the various stages of product transport. S1 is the first stage, S2 is the second stage, and S3 is the third stage. The right-pointing arrow indicates the direction of product transport.

[0103] The products include product 21, product 22, ..., product 2n. Assume:

[0104] When product 21 passes through the first sub-detection position C, if product 21 is missed (missed detection) at the first sub-detection position C, no record is made. When product 21 passes through the second sub-detection position D, encoder 500 reads the first current encoded value and records it as D. 21 .

[0105] When product 22 passes the first sub-detection position C, encoder 500 reads the first encoded value and records it as C. 21 When product 22 passes the second sub-detection position D, encoder 500 reads the first current encoding value and records it as D. 21 .

[0106] When product 2n passes the first sub-detection position C, encoder 500 reads the first encoded value and records it as C. 2n When product 2n passes the second sub-detection position D, encoder 500 reads the first current encoded value and records it as D. n1 .

[0107] The first encoded values ​​obtained by the encoder 500 constitute the first encoded sequence, for example, C 21 ......C2 n .

[0108] For example, assuming the distance to the first target is L2, D 21 If the first current encoded value is C, then the current first encoded sequence is C. 21 , set the first current encoded value D 21 With C in the first encoded sequence 21 Comparison, where D 21 and C 21 The distance difference between them is |D 21 -C 21 |=L2.

[0109] Therefore, it can be seen that product 21 exhibits a tracking anomaly, while product 22 does not. The system automatically marks product 21 as a product with an abnormal tracking result and product 22 as a product with a normal tracking result. Based on the above judgment, the controller 40 issues a tracking anomaly command for product 21, pushing product 21 out through the first unloading mechanism 100 at the unloading port corresponding to the first unloading mechanism 100 (that is, rejecting the abnormal product), and transferring product 22 to the next process (e.g., to the glass turntable 102 for secondary inspection). At the same time, the current parameter (C in the first encoding sequence) is... 21 and the first current encoded value D 21 Clean up the data to prevent large amounts of data from affecting the system's operating speed.

[0110] Third Example

[0111] See Figure 5 and Figure 6 , Figure 5 This is a schematic diagram illustrating a scenario in which a product is detected as lost in the third sub-detection device 202 according to an embodiment of this application. Figure 6 This application embodiment illustrates a scenario where a product is detected by a third sub-detection device 202 but missed by a fourth sub-detection device 302. Wherein, Figure 5 and Figure 6 S1, S2, and S3 represent the various stages of product transport. S1 is the first stage, S2 is the second stage, and S3 is the third stage. The right-pointing arrow indicates the clockwise or counterclockwise transport direction of the product.

[0112] Specifically, the products include product 31, product 32, ..., product 3n. Assume:

[0113] Please see Figure 5When product 31 passes the third sub-detection position E, encoder 500 reads the second encoded value and records it as C. 31 Furthermore, product 31 was lost during transmission at the third sub-detection position E, or, please refer to... Figure 6 If product 31 is detected when it passes the third sub-detection position E, but is missed when it passes the fourth sub-detection position F (missed detection), then no record is made.

[0114] Please see Figure 5 and Figure 6 When product 32 passes the third sub-detection position E, encoder 500 reads the second encoded value and records it as C. 32 When product 32 passes the fourth sub-detection position F, encoder 500 reads the second current encoding value and records it as D. 32 .

[0115] Please see Figure 2 and Figure 3 When product n passes the third sub-detection position E, encoder 500 reads the second encoded value and records it as C. 3n When product n passes the fourth sub-detection position F, encoder 500 reads the second current encoding value and records it as D. 3n .

[0116] The second encoded values ​​obtained by the encoder 500 constitute the second encoded sequence, for example, C 31 C 32 ......C 3n .

[0117] For example, assuming the distance to the second target is L3, D 32 If the second current encoded value is C, then the current second encoded sequence is C. 31 C 32 The second current encoded value D 32 respectively with C in the second coding sequence 31 C 32 A comparison was made one by one, among which D 32 and C 32 The distance difference between them is |D 32 -C 32 |=L,D 32 and C 31 The distance difference between them is |D 32 -C 31|>L3. It can be seen that product 31 exhibits a tracking anomaly, while product 32 does not. The system automatically marks product 31 as a product with an abnormal tracking result and product 32 as a product with a normal tracking result. Controller 40, based on the above judgment, issues a tracking normal command for product 32 and determines the product's flow direction based on the detection result of product 32. For example, if the detection result of product 32 is good, controller 40 issues a normal product command, controlling the second unloading mechanism 200 to push the good product out of the unloading port corresponding to the second unloading mechanism 200. If the detection result of product 32 is defective, controller 40 controls the defective product to flow directly to the next unloading port for discharge. Simultaneously, the current parameter (C in the second coding sequence) is... 31 C 32 And the second current encoded value D 32 Clean up the data to prevent large amounts of data from affecting the system's operating speed.

[0118] Fourth example

[0119] Please see Figure 7 , Figure 7 This application embodiment illustrates a scenario where the product is missed by the third sub-detection device 202. Wherein, Figure 7 S1, S2, and S3 represent the various stages of product transport. S1 is the first stage, S2 is the second stage, and S3 is the third stage. The right-pointing arrow indicates the direction of product transport.

[0120] The products include product 41, product 42, ..., product 4n. Assume:

[0121] When product 41 passes the third sub-detection position E, if product 41 is missed at the third sub-detection position E (missed detection), no record is made. When product 41 passes the fourth sub-detection position F, encoder 500 reads the second current encoding value and records it as D. 41 .

[0122] When product 42 passes the third sub-detection position E, encoder 500 reads the second encoded value and records it as C. 41 When product 42 passes the fourth sub-detection position F, encoder 500 reads the second current encoding value and records it as D. 41 .

[0123] When product 4n passes the third sub-detection position E, encoder 500 reads the second encoded value and records it as C. 41 When product 4n passes the fourth sub-detection position F, encoder 500 reads the second current encoded value and records it as D. 4n .

[0124] The second encoded values ​​obtained by the encoder 500 constitute the second encoded sequence, for example, C 41 ......C 4n .

[0125] For example, assuming the distance to the second target is L4, D 41 If the second current encoded value is C, then the current second encoded sequence is C. 41 The second current encoded value D 41 With C in the second encoded sequence 41 Comparison, where D 41 and C 41 The distance difference between them is |D 41 -C 41 |=L4.

[0126] Therefore, it can be seen that product 41 exhibits a tracking anomaly, while product 42 does not. The system automatically marks product 41 as a product with an abnormal tracking result and product 42 as a product with a normal tracking result. The controller 40, based on the above judgment, issues a tracking normal command for product 42 and determines the product's flow direction based on the detection result of product 42. For example, if the detection result of product 42 is a good product, the controller 40 issues a normal product command, controlling the second unloading mechanism 200 to push the good product out of the unloading port corresponding to the second unloading mechanism 200. If the detection result of product 42 is a defective product, the controller 40 controls the defective product to flow directly to the next unloading port for discharge. Simultaneously, the current parameter (C in the second encoding sequence) is... 41 C 42 And the second current encoded value D 42 Clean up the data to prevent large amounts of data from affecting the system's operating speed.

[0127] In other examples, where there is only one product on conveyor belt 1011, assuming the first target distance is L and D0 is the first current code value, since the product is missed at the first sub-detection position C, the product is directly identified as an abnormal product.

[0128] Therefore, this application can remove lost or missed products through this tracking method, thereby avoiding batch misalignment during product inspection and improving the accuracy and yield of product inspection.

[0129] The target distances L1, L2, L3, and L4 in the context can all be interval thresholds, meaning that the target distance is allowed to vary within a certain range. The specific interval thresholds can be set according to the actual situation, which will not be elaborated here.

[0130] Similarly, if the outer periphery of the glass turntable 102 is also equipped with a first detection device 20 and a second detection device 30, the tracking principle of the product when it is conveyed on the glass turntable 102 is the same as that of the product when it is conveyed on the conveying device 10, and will not be repeated here. The difference is that when the product is conveyed on the conveying device 10, the tracking distance is the straight-line distance between the two detection positions; when the product is conveyed on the glass turntable 102, the tracking distance is the circumference between the two detection positions.

[0131] The first encoded value and the first current encoded value in the context are distance values ​​that can be obtained by the linear encoder 500. The second encoded value and the second current encoded value are distance values ​​that can be obtained by the rotary encoder 500.

[0132] For example, the distance the product travels from the first sub-detection position C is t, and the distance the product travels from the second sub-detection position D is p. Here, Q1 is defined as the pulse difference between the product and the first sub-detection position C from the starting position, Q2 is defined as the pulse difference between the product and the second sub-detection position D from the first sub-detection position C, and K is the distance corresponding to a unit pulse. Then, t = Q1 * K and p = Q2 * K.

[0133] A linear encoder 500 is a sensor used to measure linear motion, providing precise position information by detecting movement along its length. Linear encoders 500 can include optical linear encoders, magnetic linear encoders, etc.

[0134] A rotary encoder 500 is a sensor used to measure rotational motion, providing position information by detecting the rotation angle of a shaft. The rotary encoder 500 may include one of the following: an optical rotary encoder 500, a magnetic rotary encoder 500, an incremental encoder 500, or an incremental encoder 500.

[0135] This application discloses a product tracking system that can accurately and automatically filter out data corresponding to lost, missed, or defective products during product testing, thereby preventing batch misalignment during testing and improving the accuracy and yield of product testing.

[0136] When tracking whether a product is defective, lost products, missed products, and defective products need to be automatically screened out in order to truly achieve the purpose of detection.

[0137] To address the aforementioned issues, this application provides a product tracking method applied to a product tracking device 40. By using the distance difference between the current encoded value and the initial encoded value, combined with the target distance, the tracking result of the corresponding product is determined. This allows for the removal of lost or missed products, preventing batch misalignment during product testing and thereby improving the accuracy and yield of product testing.

[0138] Please see Figure 11 , Figure 11 The overall flow of the product tracking method according to an embodiment of this application is illustrated. The product tracking method includes the following steps:

[0139] Step 201: Obtain the initial coding sequence and the detection results of the product.

[0140] The initial coding sequence includes multiple initial coding values, which are determined when each product passes through the first detection position.

[0141] Specifically, the initial encoded value can be collected by the encoder 500 set on the transmission device 10. The encoder 500 feeds back the collected initial encoded value to the product tracking device 40 and binds the initial encoded value with the product's detection result.

[0142] Please see Figure 12 , Figure 12 This illustration shows the overall process of obtaining product test results according to an embodiment of this application. Obtaining product test results includes the following steps:

[0143] Step 2011: Obtain the original image of the product.

[0144] Specifically, at the first detection location, the camera directly captures the original image of the product corresponding to the current photo-taking point.

[0145] Step 2012: Obtain the probability heatmap of the original image. The probability heatmap is used to characterize the probability distribution of the product at each pixel location. In other words, the probability heatmap refers to the predicted category probability of each pixel at a spatial location, i.e., the probability that each pixel at a spatial location is a product.

[0146] Please see Figure 13 and Figure 14 , Figure 13 This illustrates a first example of a probability heatmap in an embodiment of this application. Figure 14 This illustration demonstrates a second type of probability heatmap example according to an embodiment of this application. For example, the original images are shown in images a1 and a2, and the product is a defective liquid silicone sealant. The probability heatmaps are shown in images b1 and b2, where different grayscale values ​​of each pixel in images b1 and b2 represent different predicted probabilities of the product being a liquid silicone sealant. It is understood that the probability heatmap can concisely and effectively represent the probability distribution of the tested liquid silicone sealant at each pixel location.

[0147] In some embodiments, the original image can be input into a pre-built image segmentation model 42 to obtain a probability heatmap.

[0148] Please see Figure 15 , Figure 15 The structure of the image segmentation model 42 in this embodiment is illustrated. The image segmentation model 42 in this embodiment employs a fully convolutional network model and is trained and validated using labeled sample product images. The fully convolutional network model is also known as the FCN (Full Convolutional Network) model, which includes multiple convolutional layers C1 to C2. i And a deconvolutional layer T, and multiple convolutional layers C1 to C2. i The deconvolutional layer T is used to extract features from the input image B. It upsamples the extracted features to the same size as the input image B, resulting in the output image C. For example, the number of convolutional layers can be eight, with each layer having the following channel counts: 96, 256, 384, 384, 256, 4096, 4096, and 21. It should be noted that the specific number of convolutional layers and the number of channels in each layer can be set according to actual needs, and this embodiment does not impose specific limitations on this.

[0149] Understandably, the FCN model replaces all the fully connected layers of the traditional CNN (Convolutional Neural Networks) model with convolutional layers, and uses deconvolutional layers to upsample the output of the last convolutional layer, restoring it to the same size as the input image. This allows it to generate a prediction for each pixel while preserving the spatial information in the original input image. Finally, it performs pixel-by-pixel classification on the upsampled feature map, dividing the pixels in the image into different parts and assigning them different labels according to the set rules. Instead of outputting classification features, it generates a target probability heatmap.

[0150] Please see Figure 16 , Figure 16 The training process of the image segmentation model in this embodiment is illustrated. The image segmentation model 42 can be trained and validated using labeled images of sample products. Specifically, during training, multiple original images of sample products can be collected by taking pictures with a camera. Each original image is labeled, that is, each pixel of the original image is classified according to the product and labeled with different tags. Then, the labeled original images are used to train the image segmentation model 42, continuously iterating and optimizing the model parameters until the optimal model parameters are obtained, thus generating the trained image segmentation model 42.

[0151] The above scheme uses a deep learning-based image segmentation model 42 to transform various products in complex backgrounds into bright spots in grayscale backgrounds. This not only facilitates subsequent target extraction but also has high processing efficiency, which helps to improve production cycle time.

[0152] Step 2013: Determine the target region of interest for each product in the original image.

[0153] In machine vision and image processing, the region of interest (ROI) is the area that needs to be processed within an image, defined by shapes such as rectangles, circles, ellipses, or irregular polygons. Essentially, the ROI can represent the shape and size of a product.

[0154] In some embodiments, the target region of interest for each product in the original image can be determined in the following ways:

[0155] Step 1: Determine the template image corresponding to the product.

[0156] Specifically, template images can be pre-stored in the electronic device as ROI files. When needed, the device can retrieve the corresponding ROI file based on the product type to obtain the appropriate template image. It's understandable that when there are multiple product types, the electronic device can store multiple ROI files.

[0157] The template image includes a pre-determined initial region of interest for each product. These initial regions of interest are pre-labeled, stored, and displayed in the template image.

[0158] Step 2: Register the template image and the original image to determine the product's positioning deviation.

[0159] In some embodiments, the template image and the original image can be registered in the following ways:

[0160] Determine the positional information of the standard product contour in the template image. Based on the positional information of the current contour of the product to be tested in the original image, and the positional information of the standard product contour, determine the affine transformation matrix to characterize the positioning deviation.

[0161] Specifically, the standard product outline refers to the outline of the product in the template image. The current outline refers to the outline of the product to be tested in the original image. Position information can include the position and rotation angle of the corresponding outline in the corresponding image; that is, the position information of the standard product outline can include the position and rotation angle of the standard product outline in the template image, and the position information of the current outline can include the position and rotation angle of the current outline in the original image.

[0162] For example, based on the position and rotation angle of the current contour in the original image, and the position and rotation angle of the standard product contour in the template image, the affine transformation matrix can be obtained, which can be specifically represented by the following formula (1):

[0163]

[0164] In formula (1), A is the affine transformation matrix, M is the matrix representing the rotation and scaling of the coordinate axes, and B is the coordinate axis translation matrix.

[0165] Step 4: Correct each initial region of interest based on the positioning deviation to obtain the target region of interest.

[0166] In some embodiments, the size of each initial region of interest can be kept constant, and the position of each initial region of interest can be corrected based on the affine transformation matrix to obtain the target region of interest.

[0167] For example, the position and rotation angle of the initial region of interest can be corrected based on the calculated affine transformation matrix to obtain the target region of interest, wherein the size of the target region of interest is the same as the size of the corresponding initial region of interest.

[0168] Specifically, the position and rotation angle of the initial region of interest can be corrected using the following formula (2):

[0169]

[0170] In formula (2), X' and Y' are the coordinates of the target region of interest, X and Y are the coordinates of the initial region of interest, M is the matrix representing the rotation and scaling of the coordinate axes, and B is the coordinate axis translation matrix.

[0171] The above technical solution allows for the creation of independent template images for each product, with corresponding regions of interest (ROIs) on the template images. By correcting the product contour deviation between the template images and the real-time captured original images, the position of the ROI is obtained for each product in the original image. This enables the ROI to change with the product's position, thus keeping the position of the ROI relative to the product constant. The entire solution not only avoids product feature extraction but also reduces interference from unstable factors during detection, improving detection stability.

[0172] Step 2014: Determine the detection result for each product based on each target region of interest and the probability heatmap.

[0173] In some embodiments, step 2014 can be implemented in the following manner:

[0174] Step 1: Based on the probability heatmap and the region of interest for each target, extract the target region corresponding to each product.

[0175] In some examples, the target area for each product can be extracted based on the probability heatmap in the following ways:

[0176] The first step is to perform binarization segmentation on the probability heatmap using a preset grayscale threshold to determine the first image.

[0177] Specifically, the preset grayscale threshold can be set based on the grayscale difference between the foreground and background in the probability heatmap. After setting the grayscale threshold, pixels with grayscale values ​​greater than or equal to the preset grayscale threshold can be identified as foreground pixels, and all of their grayscale values ​​are assigned a value of 1. These foreground pixels form the segmented target part. Pixels with grayscale values ​​less than the preset grayscale threshold can be identified as background pixels, and all of their grayscale values ​​are assigned a value of 0. These background pixels form the background part.

[0178] The second step is to perform morphological filtering on the first image to determine the second image.

[0179] Specifically, after segmenting the target portion, morphological operators can be used for computation, such as morphological automatic hole-filling algorithms. This can better suppress noise or remove burrs, improving detection accuracy.

[0180] The third step is to extract the target region corresponding to each target region of interest from the second image.

[0181] Specifically, the target regions corresponding to each target region of interest in the second image are extracted to facilitate subsequent analysis.

[0182] By processing the probability heatmap using the above technical solution and extracting the target region corresponding to each product based on the target region of interest, each product in a complex background can be transformed into a bright spot in a grayscale background and the target can be accurately extracted. This can meet the target recognition requirements under complex backgrounds, multiple targets, and multiple features. The detection procedure is relatively simple and the detection stability is high.

[0183] Step two: Determine the detection result for each product based on the geometric features of each target area.

[0184] In some embodiments, if the geometric features of the target area meet preset conditions, it is determined that the corresponding product is defective.

[0185] In some examples, geometric features may include at least one of area, length, width, roundness, rectangularity, and aspect ratio. Preset conditions may include thresholds corresponding to the geometric features.

[0186] For example, taking a circular product as an example, and using area to represent geometric features, where the area can be calculated using the grayscale value of each pixel in the target area. If the area of ​​the target area is less than a preset threshold, then it is determined that the product corresponding to that target area is defective. If the area of ​​the target area is greater than or equal to the preset threshold, then it is determined that the product corresponding to that target area is not defective.

[0187] The above technical solution does not have special requirements for product color consistency, product-background contrast, or product outline clarity during processing. This can improve the situation where the detection accuracy of product defects is easily affected by factors such as background and lighting, thereby enhancing the stability of product defect detection.

[0188] Step 202: Obtain the current encoded value.

[0189] The current code value is determined when each product passes through the second detection position, which is located after the first detection position along the product conveying direction.

[0190] Specifically, the current encoded value can be acquired by the encoder 500 set on the transmission device 10, and the encoder 500 will feed back the acquired current encoded value to the controller.

[0191] Step 203: Determine the distance difference based on the current encoded value and the first encoded value.

[0192] For example, the products include product 1, product 2, ..., product n. Assume:

[0193] For example, please see Figure 2 When product 1 passes the first sub-detection position C, encoder 500 reads the first encoded value and records it as C1. Also, product 1 is lost during transmission at the first sub-detection position C. Alternatively, please refer to... Figure 3 If product 1 is detected when it passes through the first sub-detection position C, but is missed when it passes through the second sub-detection position D (missed detection), then no record is made.

[0194] Please see Figure 2 and Figure 3 When product 2 passes through the first sub-detection position C, encoder 500 reads the first encoded value and records it as C2, and when product 2 passes through the second sub-detection position D, encoder 500 reads the first current encoded value and records it as D2.

[0195] Please see Figure 2 and Figure 3 When product n passes the first sub-detection position C, encoder 500 reads the first encoded value and records it as C. n When product n passes the second sub-detection position D, encoder 500 reads the first current encoded value and records it as D. n .

[0196] Step 204: Determine the tracking result for the corresponding product based on the distance difference and the target distance.

[0197] The target distance is the distance between the first detection position and the second detection position.

[0198] In some embodiments, step 204 can be implemented in the following manner:

[0199] First, if the distance difference is not equal to the target distance, identify the corresponding product tracking anomaly.

[0200] For example, if the distance difference is greater than the target distance, it is determined that the product corresponding to the first coded value was lost before reaching the second detection position, or it is determined that the product corresponding to the first coded value was missed by the second detection position when it reached the second detection position.

[0201] If the distance difference is less than the target distance, it is determined that the product corresponding to the second code value was missed by the first detection position when it arrived at the first detection position.

[0202] The above examples all illustrate situations where product tracking is abnormal. Once a product tracking abnormality is determined, the abnormal product can be removed.

[0203] Second, if the distance difference is equal to the target distance, the tracking of the corresponding product is confirmed to be normal.

[0204] Among them, the test results of tracking normal products are used to determine whether a product should be rejected, and the test results are used to characterize whether there are defects in the product.

[0205] For example, please see Figure 1 When a product passes through the first sub-detection position C and is detected by the first sub-detection device 201, the controller 40 acquires the first code value of the product. If the product is simultaneously captured by the first imaging device 501, the first imaging device 501 transmits the original image of the product to the controller 40. At this time, the controller 40 analyzes the detection result based on the original image of the product and binds and saves the detection result with the first code value of the product. After passing through the first sub-detection position C, when the product reaches the second sub-detection position D, the second sub-detection device 301 detects the product. The controller 40 reads the first current code value of the product and compares the first current code value with the first code value to obtain a first distance difference. If the first distance difference is equal to the first target distance, and the product detection result is a defective product, the controller 40 issues a product abnormality command and controls the first unloading mechanism 100 to push the abnormal product out of the unloading port corresponding to the first unloading mechanism 100.

[0206] If the first distance difference is equal to the first target distance, and the product test result is good, the product is transported to the next process for a second test to further determine whether the product is defective, thereby further improving the product yield.

[0207] After confirming that the product tracking result is normal, the product flow is determined based on the detection results of the products with normal tracking results. The detection results are used to characterize whether there are defects in the products. It is understood that when a product passes through the third sub-detection position E and is detected by the third sub-detection device 202, the controller 40 acquires the second code value of the product. If the product is simultaneously captured by the second camera 90, the second camera 90 transmits the original image of the product to the controller 40. At this time, the controller 40 analyzes the detection results based on the original image of the product and binds and saves the detection results with the second code value of the product. When the product passes through the third sub-detection position E and reaches the fourth sub-detection position F, the fourth sub-detection device 302 detects the product. The controller 40 reads the second current code value of the product and compares the second current code value with the second code value to obtain a second distance difference. If the second distance difference is equal to the second target distance, and the product detection result is good, the controller 40 issues a product normal command and controls the second unloading mechanism 200 to push the good product out of the unloading port corresponding to the second unloading mechanism 200. If the second distance difference is equal to the second target distance, and the product test result is a defective product, the controller 40 controls the defective product to flow directly to the next discharge port for discharge.

[0208] Please see Figure 17 , Figure 17This illustration shows the specific flow of the product tracking method according to an embodiment of this application. In some examples, after executing the product tracking method according to an embodiment of this application, the product tracking device 40 can send the detection result of each product to the upper control system of the product tracking device 40 to complete the detection of the current shooting point. In practical applications, depending on the size of the product being photographed, the detection accuracy, etc., it may be necessary to take pictures once or multiple times. When the first detection device 20 senses that the product has arrived at the shooting point, the encoder 500 records the first encoding value and sends it to the upper control system. The upper control system can send the shooting point to the product tracking device 40 so that the product tracking device 40 controls the motion mechanism 60 to drive the camera to the shooting point in sequence and trigger the camera to take the original image of the product to be tested. Then, the product tracking device 40 continues to perform AI (Artificial Intelligence) detection on the original image of the product to generate a probability heatmap of the original image, and analyzes the current shooting point to obtain the initial region of interest corresponding to each product at the corresponding shooting point. The template image of the product to be tested at the shooting point is used to register the original image, and the position of the initial region of interest is corrected to obtain the target region of interest. The product tracking device 40 analyzes the probability heatmap in conjunction with the target region of interest to obtain the detection result for each product. It then binds the detection result of each product to a first coded value and generates an initial coded sequence. When the second detection device 30 senses a product arriving at the discharge position, the encoder 500 records the current coded value and sends it to the upper control system. The upper control system calculates the distance difference between the initial and current coded values ​​and compares this distance difference with the target distance. If the distance difference equals the target distance, it is determined that the current product has not been misaligned and the detection result corresponding to the coded value is the detection result for the current product. Finally, based on the detection result of the current product, it is determined whether to remove it, thus completing the tracking and detection of the current product using the product tracking method of this embodiment.

[0209] It is understood that the product tracking method provided in this application embodiment transforms various products in complex backgrounds into bright spots in grayscale backgrounds through deep learning-based image segmentation methods, and determines the product detection results by combining the target region of interest. This can accurately extract the target and reduce the interference of unstable factors. When performing product detection, it is not easily affected by factors such as product loss or missed detection, and can ensure detection accuracy under various conditions, thereby improving the stability of product tracking and detection.

[0210] Accordingly, please refer to Figure 18 , Figure 18This illustration shows the structure of a product tracking device according to an embodiment of this application. This application also provides a product tracking device, including a memory 801 and a processor 802. The memory 801 stores a program; the processor 802 executes the program stored in the memory 801; when the program stored in the memory 801 is executed, the processor 802 executes the product tracking method provided in this application embodiment.

[0211] It is understood that the product tracking device of this application embodiment is not easily affected by factors such as product loss or missed detection when performing product testing, and can ensure the detection accuracy under various conditions, has high detection stability, and facilitates product quality control.

[0212] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0213] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0214] The embodiments, implementation methods, and related technical features of this application can be combined and substituted for each other without conflict.

[0215] The above are merely preferred embodiments of this application and are not intended to limit this application in any way. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this application without departing from the scope of the technical solution of this application shall still fall within the scope of the technical solution of this application.

Claims

1. A product tracking system, characterized in that, include: Workbench (400); A conveying device (10) is disposed on the worktable (400) and is configured to carry and convey products. An encoder (500) is disposed on the conveying device (10); When the product is conveyed by the conveying device (10) past the first detection position, the encoder (500) reads the first encoding value and sends the first encoding value to the product tracking device (40); when the product is conveyed by the conveying device (10) past the second detection position, the encoder (500) reads the current encoding value and sends the current encoding value to the product tracking device (40).

2. The product tracking system according to claim 1, characterized in that, Also includes: An imaging device (50) is used to acquire the original image of the product and send it to the product tracking device (40).

3. The product tracking system according to claim 2, characterized in that, The product tracking device (40) includes: Image acquisition card (41), which is connected to the imaging device (50) and is used to convert the output signal of the imaging device (50) into a digital signal; Image segmentation model (42) is used to extract the probability heatmap of the original image; A visual processor (43) is connected to the image acquisition card (41) and the image segmentation model (42) respectively. The visual processor (43) is used to determine the target region of interest corresponding to each product in the original image, so as to determine the detection result of each product based on each target region of interest and the probability heatmap. The detection result is used to characterize whether the product is defective.

4. The product tracking system according to claim 3, characterized in that, Also includes: A display device (600) is connected to the vision processor (43) for displaying the detection results.

5. The product tracking system according to claim 3, characterized in that, Also includes: A GPU-accelerated graphics card (700) is connected between the vision processor (43) and the image segmentation model (42).

6. The product tracking system according to claim 1, characterized in that, The conveying device (10) includes a conveyor belt (101) and a glass turntable (102), the glass turntable (102) being disposed downstream of the conveyor belt (101) to receive products conveyed from the conveyor belt (101).

7. The product tracking system according to claim 6, characterized in that, The encoder (500) includes a first encoder (5001) and a second encoder (5002), the first encoder (5001) being disposed on the conveyor belt (101) and the second encoder (5002) being disposed on the glass turntable (102); The first detection position includes a first sub-detection position (C) and a third sub-detection position (E), and the second detection position includes a second sub-detection position (D) and a fourth sub-detection position (F); When the product is conveyed by the conveyor belt (101) past the first sub-detection position (C), the first encoder (5001) reads the first encoded value and sends the first encoded value to the product tracking device (40); and when the product is conveyed by the conveyor belt (101) past the second sub-detection position (D), the first encoder (5001) reads the first current encoded value and sends the first current encoded value to the product tracking device (40); When the product is conveyed by the glass turntable (102) past the third sub-detection position (E), the second encoder (5002) reads the second encoding value and sends the second encoding value to the product tracking device (40); and when the product is conveyed by the glass turntable (102) past the fourth sub-detection position (F), the second encoder (5002) reads the second current encoding value and sends the second current encoding value to the product tracking device (40).

8. The product tracking system according to claim 6, characterized in that, It also includes a storage bin (70) and a vibratory feeder (80) disposed on the workbench (400). The vibratory feeder (80) is installed on one side of the storage bin (70), and the discharge port of the storage bin (70) extends into the vibratory feeder (80). (102) One end of the conveyor belt is located at the outlet of the vibrating plate (80), and the other end is located at the inlet of the glass turntable (102).

9. The product tracking system according to claim 8, characterized in that, Also includes: The first unloading mechanism (100) is located on one side of the conveyor belt (101) along the conveying direction of the product and is used to screen defective products.

10. The product tracking system according to claim 9, characterized in that, Also includes: The second unloading mechanism (200) is disposed on the worktable (400) and located on the outer periphery of the glass turntable (102) for screening good products.

Images