A product quality detection method and related system

Abstract

The application provides a product quality detection method, applied to the field of artificial intelligence (AI) technology, comprising: acquiring a to-be-detected image, identifying the to-be-detected image, obtaining at least one first detected target and a first reference target in the to-be-detected image, determining a first relative position of the at least one first detected target relative to the first reference target, acquiring a second relative position of at least one second detected target in a template image relative to a second reference target in the template image, performing quality detection on a to-be-detected product according to the first relative position and the second relative position, and obtaining a quality detection result. The method introduces reference targets in the to-be-detected image and the template image respectively, compares the relative positions of the detected targets in the to-be-detected image and the detected targets in the template image relative to the reference targets in the respective images, realizes quality detection, thereby avoiding image registration, reducing the false detection rate, and improving the robustness.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence (AI) technology, and in particular to a product quality inspection method, a product quality inspection system, a computing device cluster, a computer-readable storage medium, and a computer program product. Background Technology

[0002] With the continuous development of artificial intelligence (AI) technology, especially deep learning (DL) technology, product quality inspection based on deep learning has become a trend in the industrial manufacturing field. Product quality inspection based on deep learning refers to using cameras to acquire images instead of the human eye, and then using deep learning algorithms to compile quality inspection standards for specific products into image processing rules, thereby replacing human experience to achieve automated quality inspection.

[0003] Product quality inspection based on deep learning typically involves two parts: offline processing and online processing. Offline processing involves selecting a qualified product as a template, capturing a template image using a camera, and then using a deep learning algorithm to identify the target objects in the template image. This process obtains the outline of the target object and its position in the reference coordinate system of the template image, recording the number of target objects and their positions. Online processing involves capturing images of the product to be inspected to generate an image for inspection. A deep learning algorithm then identifies the outline of the target objects in the image and their positions in the reference coordinate system. Next, an affine transformation is performed on the image to align its reference coordinate system with that of the template image, thus achieving registration. Finally, the target objects in the image and the template image are matched. A successful match indicates the product is qualified; an unsuccessful match indicates the product is unqualified.

[0004] However, in some scenarios, such as product offset or rotation, the existing technology has a high false detection rate for product quality inspection. Summary of the Invention

[0005] This application provides a product quality inspection method. This method introduces reference targets into both the image to be inspected and the template image, and compares the relative positions of the target in the image to be inspected and the target in the template image with respect to the reference targets in their respective images. This achieves quality inspection by reducing registration and thus solving the problem of inaccurate affine transformation due to product offset or rotation, thereby reducing the false detection rate. This application also provides a product quality inspection system, a computing device cluster, a computer-readable storage medium, and a computer program product corresponding to the above method.

[0006] Firstly, this application provides a product quality inspection method. This method can be executed by a product quality inspection system. In some embodiments, the product quality inspection system can be a software system, and a computing device or cluster of computing devices executes the program code of the software system to perform the product quality inspection method. In other embodiments, the product quality inspection system can also be a hardware system for performing quality inspection on products, such as an industrial control computer, a server, etc. This application uses a product quality inspection system as a software system as an example for illustration.

[0007] Specifically, the product quality inspection system acquires an image to be inspected, which is an image obtained by taking pictures of the product to be inspected. Then, the product quality inspection system identifies the image to be inspected and obtains at least one target to be inspected in the image (in this application, the target to be inspected in the image is also referred to as the first target to be inspected) and a reference target in the image to be inspected (in some cases, also referred to as the first reference target). The system determines the relative position of at least one first target to be inspected relative to the first reference target (for convenience, it can also be referred to as the first relative position). The product quality inspection system also acquires the relative position of at least one target to be inspected in a template image (also referred to as the second target to be inspected) relative to the reference target (i.e., the second reference target) in the template image (also referred to as the second relative position). Based on the first relative position and the second relative position, the system performs quality inspection on the product to be inspected and obtains the quality inspection result.

[0008] This method introduces reference targets into both the target image and the template image, and compares the relative positions of the detected target in the target image and the template image with respect to the reference targets in their respective images to achieve quality detection. It eliminates the need for template registration and affine transformation, thus avoiding the problem of inaccurate affine transformations caused by slight product offsets or rotations in the camera's field of view, reducing the false detection rate and demonstrating good robustness. Furthermore, the absence of affine transformation significantly reduces computational load and shortens detection time, meeting production requirements.

[0009] In some possible implementations, when the image to be detected includes multiple first detected targets, the product quality inspection system can perform quality inspection on the product to be detected based on the number of first detected targets in the image to be detected, the number of second detected targets in the template image, and the first relative position and the second relative position, and obtain the quality inspection result.

[0010] Specifically, the product quality inspection system can first determine whether the number of first detected targets in the image to be inspected is consistent with the number of second detected targets in the template image. If they are consistent, the system continues to match the first relative position with the second relative position, and obtains the quality inspection result based on the matching result. A matching result indicating that the first relative position and the second relative position match indicates that the product to be inspected is qualified; a matching result indicating that the first relative position and the second relative position do not match indicates that the product to be inspected is unqualified. If they are inconsistent, the product to be inspected can be directly identified as unqualified.

[0011] In this way, some unqualified products can be screened out in advance (for example, products whose quantity of the first inspection target is inconsistent with the quantity of the second inspection target), which improves the efficiency of quality inspection and meets production needs.

[0012] In some possible implementations, the product quality inspection system can match the first relative position with the second relative position to obtain a matching result. When the matching result indicates that the first relative position and the second relative position do not match, the product quality inspection system determines that the product to be inspected is a defective product.

[0013] This method can perform quality inspection on products by performing simple matching operations on relative positions. Since it does not require a large number of calculations, it improves the efficiency of quality inspection, meets production needs, and can achieve quality inspection without the need for high-performance hardware, thus reducing inspection costs.

[0014] In some possible implementations, the product quality inspection system can determine the coordinate difference between the first relative position and the second relative position. When the coordinate difference is greater than a preset threshold, the product quality inspection system determines that the first relative position and the second relative position do not match.

[0015] The coordinate difference can include a difference in the horizontal coordinate and a difference in the vertical coordinate. The product quality inspection system can subtract the horizontal coordinates of the first relative position and the second relative position, and then take the absolute value to obtain the difference in the horizontal coordinate. Similarly, the product quality inspection system can subtract the vertical coordinates of the first relative position and the second relative position, and then take the absolute value to obtain the difference in the vertical coordinate. In some embodiments, the preset threshold can include a first preset threshold and a second preset threshold. When the difference in the horizontal coordinate is greater than the first preset threshold, or the difference in the vertical coordinate is greater than the second preset threshold, the product quality inspection system can determine that the product to be inspected is a defective product.

[0016] Compared to the approach of first registering the image to be detected and the template image and then performing target matching, this method is less affected by product offset or rotation, thus resulting in a lower false detection rate and better robustness.

[0017] In some possible implementations, the product quality inspection system can use deep learning algorithms to identify the image to be inspected, generate candidate boxes, classify and regress the candidate boxes, and obtain the positions of at least one first detected target in the image to be inspected and the positions of a first reference target in the image to be inspected within the reference coordinate system of the image to be inspected. Accordingly, the product quality inspection system can obtain the first relative position based on the positions of the at least one first detected target and the first reference target in the reference coordinate system of the image to be inspected.

[0018] In this product quality inspection system, when identifying an image to be inspected, features are extracted using deep learning algorithms to obtain a feature map. For each feature point in the feature map, several candidate boxes of different scales are generated. Then, the candidate boxes are classified based on whether they include a target (such as a first detected target or a first baseline target). Furthermore, the product quality inspection system can regress the candidate boxes that are classified as including the target to obtain the target's position in the baseline coordinate system of the image to be inspected.

[0019] This method first determines the positions of the first detected target and the first reference target in the reference coordinate system of the image to be detected through a deep learning algorithm. Then, it determines the first relative position based on the positions of the first detected target and the first reference target in the reference coordinate system of the image to be detected. Quality detection is then performed based on the first relative position and the second relative position, which has high usability.

[0020] In some possible implementations, the product quality inspection system can also present the quality inspection results to the user to provide information prompts, enabling the user to promptly handle non-conforming products and meet production needs.

[0021] In some possible implementations, when the quality inspection result indicates that the product to be inspected is a non-conforming product, the product quality inspection system can also receive processing instructions from the user regarding the product to be inspected through the result display interface, so as to process the product to be inspected according to the processing instructions. The processing instructions can be used to instruct the non-conforming product to be discarded or reworked. This can prevent non-conforming products from being mixed with qualified products and improve the pass rate of batch products.

[0022] In some possible implementations, the first reference target is the boundary of the product to be inspected, and the second reference target is the boundary of the qualified product. Using the product boundary as a reference target can provide a reference for determining the relative position of the target being inspected.

[0023] In some possible implementations, the second relative position of at least one second detected target in the template image relative to a second reference target in the template image can be pre-identified and calculated. Specifically, the product quality inspection system can pre-acquire a template image, then determine the second reference target in the template image, for example, by identifying a target specified by the user through a parameter configuration interface as the second reference target. Next, it can use a deep learning algorithm to identify the template image, obtain at least one second detected target in the template image, and determine the second relative position of at least one second detected target relative to the second reference target. The product quality inspection system can store the aforementioned second relative position in a storage device so that it can retrieve the second relative position from the storage device during subsequent quality inspections and perform quality inspections based on the first and second relative positions.

[0024] In this method, the product quality inspection system determines the second relative position of at least one second target to be detected in the template image relative to a second reference target by pre-identifying the template image, which can reduce the amount of calculation required for subsequent quality inspection and improve the efficiency of quality inspection.

[0025] In some possible implementations, the product quality inspection system can also identify template images in real time to obtain at least one second target to be inspected and a second reference target (e.g., a target specified by the user among the targets identified by the product quality inspection system), and determine the second relative position of at least one second target to be inspected relative to the second reference target. This allows for product quality inspection based on the first and second relative positions, which can reduce the false detection rate and improve the efficiency of quality inspection.

[0026] Secondly, this application provides a product quality inspection system. The system includes:

[0027] The interaction module is used to acquire the image to be detected, which is an image of the product to be detected captured by photography;

[0028] The recognition module is used to recognize the image to be detected and obtain at least one first detected target and a first reference target in the image to be detected;

[0029] The identification module is further configured to determine the first relative position of the at least one first detected target relative to the first reference target;

[0030] The detection module is used to obtain the second relative position of at least one second target to be detected in a template image relative to a second reference target in the template image, wherein the template image is an image obtained by taking a picture of a qualified product, and the product to be detected is subjected to quality inspection based on the first relative position and the second relative position to obtain a quality inspection result.

[0031] In some possible implementations, the detection module is specifically used for:

[0032] When the image to be detected includes multiple first targets, the quality inspection of the product to be detected is performed based on the number of first targets in the image to be detected, the number of second targets in the template image, the first relative position, and the second relative position, to obtain the quality inspection result.

[0033] In some possible implementations, the detection module is specifically used for:

[0034] The first relative position is matched with the second relative position to obtain a matching result;

[0035] When the matching result indicates that the first relative position and the second relative position do not match, the product to be tested is determined to be a non-conforming product.

[0036] In some possible implementations, the detection module is specifically used for:

[0037] Determine the coordinate difference between the first relative position and the second relative position;

[0038] When the coordinate difference is greater than a preset threshold, it is determined that the first relative position and the second relative position do not match.

[0039] In some possible implementations, the identification module is specifically used for:

[0040] The image to be detected is identified by a deep learning algorithm to generate candidate boxes. The candidate boxes are then classified and regressed to obtain the position of at least one first detected target in the image to be detected in the reference coordinate system of the image to be detected and the position of the first reference target in the image to be detected in the reference coordinate system of the image to be detected.

[0041] The first relative position is obtained based on the position of the at least one first detected target in the reference coordinate system of the image to be detected and the position of the first reference target in the reference coordinate system of the image to be detected.

[0042] In some possible implementations, the interaction module is also used for:

[0043] Present the quality test results to the user.

[0044] In some possible implementations, the interaction module is also used for:

[0045] When the quality inspection result indicates that the product to be inspected is a non-conforming product, the system receives the user's processing instruction for the product to be inspected through the result display interface, so as to process the product to be inspected according to the processing instruction.

[0046] In some possible implementations, the first benchmark target is the boundary of the product to be tested, and the second benchmark target is the boundary of the qualified product.

[0047] In some possible implementations, the second relative position of at least one second detected target in the template image relative to a second reference target in the template image is pre-identified and calculated.

[0048] Thirdly, this application provides a computing device cluster. The computing device cluster includes at least one computing device, which includes at least one processor and at least one memory. The at least one processor and the at least one memory communicate with each other. The at least one processor is configured to execute instructions stored in the at least one memory to cause the computing device or the computing device cluster to perform a product quality inspection method as described in the first aspect or any implementation thereof.

[0049] Fourthly, this application provides a computer-readable storage medium storing instructions that instruct a computing device or a cluster of computing devices to execute the product quality testing method described in the first aspect or any implementation thereof.

[0050] Fifthly, this application provides a computer program product containing instructions that, when run on a computing device or a cluster of computing devices, causes the computing device or cluster of computing devices to execute the product quality testing method described in the first aspect or any implementation thereof.

[0051] Based on the implementation methods provided in the above aspects, this application can be further combined to provide more implementation methods. Attached Figure Description

[0052] To more clearly illustrate the technical methods of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly described below.

[0053] Figure 1 This application provides a schematic diagram of the architecture of a product quality inspection system.

[0054] Figure 2 A schematic diagram of a parameter configuration interface provided in an embodiment of this application;

[0055] Figure 3 A schematic diagram of a result display interface provided in an embodiment of this application;

[0056] Figure 4 A flowchart of a product quality testing method provided in this application embodiment;

[0057] Figure 5 This is a schematic diagram illustrating a method for identifying a template image to determine a second relative position, as provided in an embodiment of this application.

[0058] Figure 6 A schematic diagram of a target matching process provided for an embodiment of this application;

[0059] Figure 7 A schematic diagram of a target matching process provided for an embodiment of this application;

[0060] Figure 8 This is a schematic diagram of the structure of a computing device cluster provided in an embodiment of this application. Detailed Implementation

[0061] The terms "first" and "second" used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first" and "second" may explicitly or implicitly include one or more of that feature.

[0062] First, some technical terms involved in the embodiments of this application will be introduced.

[0063] Quality inspection, also known as quality testing or quality control, refers to the testing of whether the inherent characteristics of a product meet the quality requirements stipulated in relevant regulations (such as suitability requirements and safety requirements). These relevant regulations can be one or more of the following: legal regulations, industry regulations, or customer regulations.

[0064] Machine learning (ML) is a branch of artificial intelligence (AI) that primarily involves automatically analyzing data to identify patterns and using these patterns to predict unknown data. Machine learning can be used for quality inspection of products in fields such as electronics, semiconductors, home appliances, printing, automobiles, and food packaging; this method of quality inspection is also known as machine vision-based quality inspection.

[0065] Depending on the machine learning algorithms employed, product quality inspection based on machine vision can be implemented in various ways. A typical example is product quality inspection based on deep learning. Deep learning-based product quality inspection usually involves regressing the position of the target object in the reference coordinate system of the image to be inspected using deep learning, and then registering the image to be inspected with a template image through affine transformation. However, affine transformation is sensitive to product offset and rotation. When the product is offset or rotated significantly within the camera's field of view, the affine transformation becomes inaccurate, leading to a higher false detection rate and affecting the robustness of deep learning-based quality inspection methods. Furthermore, affine transformation involves matrix calculations, which are time-consuming and difficult to meet production requirements.

[0066] In view of this, embodiments of this application provide a product quality inspection method. This method can be executed by a product quality inspection system (for brevity, it is sometimes simply referred to as an inspection system in this application). In some embodiments, the product quality inspection system can be a software system, and a computing device or cluster of computing devices executes the program code of the software system to perform the product quality inspection method. In other embodiments, the product quality inspection system can also be a hardware system for performing quality inspection on products, such as an industrial personal computer (IPC), a server, etc. Embodiments of this application illustrate this using a software system as an example.

[0067] Specifically, the product quality inspection system acquires an image to be inspected, which is an image obtained by taking pictures of the product to be inspected. Then, the product quality inspection system identifies the image to be inspected and obtains at least one target to be inspected in the image (in this application, the target to be inspected in the image is also referred to as the first target to be inspected) and a reference target in the image to be inspected (in some cases, also referred to as the first reference target). The system determines the relative position of at least one first target to be inspected relative to the first reference target (for convenience, it can also be referred to as the first relative position). The product quality inspection system also acquires the relative position of at least one target to be inspected in a template image (also referred to as the second target to be inspected) relative to the reference target (i.e., the second reference target) in the template image (also referred to as the second relative position). Based on the first relative position and the second relative position, the system performs quality inspection on the product to be inspected and obtains the quality inspection result.

[0068] This method introduces reference targets into both the image to be detected and the template image. Quality detection is achieved by comparing the relative positions of the detected target in the two images with respect to the reference targets in their respective images. This eliminates the need for template registration and affine transformation, thus avoiding inaccuracies caused by slight product offsets or rotations within the camera's field of view. This reduces the false detection rate and demonstrates good robustness. Furthermore, the absence of affine transformation significantly reduces computational load and detection time, meeting production requirements.

[0069] The product quality inspection method of this application can be applied to different industries and used to inspect the quality of different products. For example, the product quality inspection method of this application can be applied to the semiconductor field to inspect the quality of products such as display panels. As another example, the product quality inspection method of this application can be applied to the food packaging field to inspect the quality of food packaging bags or boxes.

[0070] To make the technical solution of this application clearer and easier to understand, the product quality inspection system of this application embodiment will be described in detail below with reference to the accompanying drawings.

[0071] See Figure 1 The diagram shows the architecture of a product quality inspection system. The product quality inspection system 100 is deployed in cloud environment 10. Here, cloud environment refers to a central computing device cluster owned by a cloud service provider, used to provide computing, storage, and communication resources. Specifically, the product quality inspection system 100 can be deployed in one or more computing devices (e.g., a central server) within the central computing device cluster. This product quality inspection system 100 is used to perform quality inspection on the products 200 (or simply product 200) produced on production line 20.

[0072] A camera 30 is installed on production line 20 to capture images of the products 200 produced on production line 20, obtaining images to be inspected. A product quality inspection system 100 deployed in cloud environment 10 acquires these images and processes them to perform quality inspection on the products 200. Furthermore, the product quality inspection system 100 can also return the quality inspection results to terminal 40, so as to present the results to users (e.g., operators).

[0073] In this embodiment, the product quality inspection system 100 includes the following functional modules: an interaction module 102, an identification module 104, and a detection module 106. The interaction module 102 is used to acquire an image to be inspected, which is an image captured by photographing the product 200. The interaction module 102 can provide a data upload interface, through which the camera 30 can report the image to be inspected to the product quality inspection system 100. The identification module 104 is used to identify the image to be inspected, obtain at least one first detected target and a first reference target in the image, and then determine the first relative position of at least one first detected target relative to the first reference target. The detection module 106 is used to acquire the second relative position of at least one second detected target in a template image relative to a second reference target, and perform quality inspection on the product 200 based on the first and second relative positions, for example, by matching the first and second relative positions to obtain a quality inspection result.

[0074] The template image is an image obtained by photographing a qualified product, which refers to a product that meets the quality requirements stipulated by relevant regulations such as laws and industry regulations. In some possible implementations, the second relative position of at least one second detected target in the template image relative to a second reference target can be pre-identified and calculated. Specifically, the interaction module 102 can also provide code for an interactive interface, which the terminal 40 can load to present to the user. This interactive interface may include an interface for configuring parameters of the quality detection algorithm, i.e., a parameter configuration interface. The parameters of the quality detection algorithm may include one or more of the template image, a reference target (e.g., a second reference target in the template image), etc.

[0075] See Figure 2The diagram shows a parameter configuration interface 201. This interface includes a product model configuration control 202, a template image configuration control 203, and a benchmark target configuration control 204. Users can configure the model of the product to be tested, such as product 200, using the product model configuration control 202. The product model configuration control 202 has a drop-down list, allowing users to select the model of product 200. In some embodiments, users can also directly input the model of product 200. Next, users can configure a template image corresponding to the model using the template image configuration control 203. Specifically, the template image configuration control 203 allows users to select one or more images from a file system as template images by browsing the file system. The parameter configuration interface 201 can also display template images. Correspondingly, when the benchmark target configuration control 204 is triggered, users can select a target from the template image as a second benchmark target by clicking. For example, if product 200 is a mobile phone, the user can select the bottom edge of the phone as the second benchmark target.

[0076] The parameter configuration interface 201 also includes a confirm control 205 and a cancel control 206. When the confirm control 205 is triggered, the user-configured parameters are submitted to the product quality inspection system 100. Correspondingly, in the offline phase, the recognition module 104 in the product quality inspection system 100 can pre-identify the template image to obtain at least one second target to be detected and a second reference target in the template image. The second reference target can be the target selected by the user from multiple targets identified by the recognition module 104 from the template image. Furthermore, the recognition module 204 can determine the second relative position of at least one second target to be detected relative to the second reference target, thereby assisting in product quality inspection. Figure 1 As shown, Figure 1 The dashed boxes represent the method steps executed in the offline phase, and the implementation boxes represent the method steps executed in the online phase. When the cancellation control 206 is triggered, the above parameter configuration can be cancelled.

[0077] In some possible implementations, the user interface may also include an interface for displaying the quality inspection results, also known as a results display interface. See also Figure 3The diagram shows a result display interface 301, which includes the quality inspection result 302 of product 200. The quality inspection result 302 can be displayed using text and / or images. For example, if the quality inspection result 302 indicates that product 200 is defective, the result display interface 301 can directly display the text "Defective" and show the reason for the defect through an image, such as the positional offset of a certain inspected target exceeding a specified value. Furthermore, the result display interface 301 also includes an exception handling control 303, which is used to handle the defective product 200. For example, the exception handling control 303 can be used to instruct the defective product 200 to be discarded or reworked.

[0078] It should be noted that, Figure 1 The example of deploying the product quality inspection system 100 in the cloud environment 10 is only used for illustration. In other possible implementations of the embodiments of this application, the product quality inspection system 100 may also be deployed in the edge environment or on the end device.

[0079] The term "edge environment" refers to a cluster of edge computing devices that are geographically close to the edge devices and are used to provide computing, storage, and communication resources. An edge computing device cluster includes one or more edge computing devices, which can be servers, computing boxes, etc. Edge devices can be terminals, including but not limited to industrial control computers, desktop computers, laptops, or smartphones.

[0080] Figure 1 An exemplary embodiment provides one way to divide the product quality inspection system 100. In other possible implementations of this application, the product quality inspection system 100 may also be divided into different modules in other ways. Multiple modules of the product quality inspection system 100 may also be deployed in a distributed manner in different environments, such as cloud environments and edge environments.

[0081] Next, from the perspective of the product quality inspection system 100, the product quality inspection method provided in the embodiments of this application will be described in detail.

[0082] See Figure 4The flowchart shown illustrates a product quality inspection method, which can be divided into offline and online phases. In the offline phase, the product quality inspection system 100 receives a user-configured template image, identifies the template image, obtains at least one second target to be detected and a second reference target within the template image, and determines the second relative position of the at least one second target to be detected relative to the second reference target. In the online phase, the product quality inspection system 100 acquires an image to be inspected, identifies the image using a deep learning algorithm, obtains at least one first target to be detected and a first reference target within the image, determines the first relative position of the at least one first target to be detected relative to the first reference target, and performs target matching based on the first and second relative positions, thereby achieving product quality inspection. The steps included in this method are described in detail below:

[0083] S402: Product quality inspection system 100 acquires template image.

[0084] Specifically, the product quality inspection system 100 can receive template images configured by the user. For example, the product quality inspection system 100 can present a parameter configuration interface to the user, through which it receives the template image configured by the user. The template image can be an image obtained by photographing a qualified product. In some embodiments, the user can filter images of qualified products from historical images to specify a template image. In other embodiments, the user can also select a qualified product, photograph the qualified product, and obtain a template image.

[0085] It should be noted that the product quality inspection system 100 can be used to perform quality inspections at different stages of product manufacturing. Therefore, a qualified product can be a finished product or a semi-finished product. For example, when the product is a mobile phone, the product quality inspection system 100 can perform product quality inspections during the production of the mobile phone's display screen and during the assembly of the display screen and circuit board into the mobile phone. Accordingly, a qualified product can be a qualified display screen or a qualified complete device.

[0086] S404: Product quality inspection system 100 determines a second reference target in a template image.

[0087] The second reference target is a benchmark in the template image, typically used as a reference point for the second target to be detected in the template image to determine its relative position. This benchmark can be a product boundary or a product component. Therefore, the second reference target can be the boundary or a component of a qualified product.

[0088] Specifically, the product quality inspection system 100 can present a parameter configuration interface to the user, which can display a template image. The user can specify a target in the template image, and the product quality inspection system 100 will determine the target specified by the user as the second benchmark target.

[0089] S406: Product quality inspection system 100 identifies template images and obtains at least one second target to be inspected.

[0090] The product quality inspection system 100 can identify template images using deep learning algorithms to obtain multiple targets. Among these targets, the targets other than the second baseline target are the second detected targets. In this way, the product quality inspection system 100 can obtain at least one second detected target.

[0091] Specifically, the product quality inspection system 100 can extract feature maps from the template image using deep learning algorithms, generate several candidate boxes for each feature point of the feature map, and then classify the candidate boxes, for example, classifying whether the candidate boxes include the target. For candidate boxes that are classified as including the target, the product quality inspection system 100 performs regression on them to obtain the position of the target in the reference coordinate system of the template image.

[0092] The position of the target in the reference coordinate system of the template image includes the position of the second reference target in the reference coordinate system of the template image and the position of at least one second detected target in the reference coordinate system of the template image. This position can be represented by the coordinates of the second reference target and at least one second detected target in the reference coordinate system.

[0093] S408: Product quality inspection system 100 determines a second relative position of at least one second inspected target relative to a second reference target.

[0094] Specifically, the position of the second detected target can be represented by the coordinates of its feature points, and the position of the second reference target can be represented by the coordinates of its feature points. These feature points can be vertices, center points, etc. For each second detected target, the product quality inspection system 100 can calculate the difference between the coordinates of the feature points of the second detected target and the coordinates of the feature points of the second reference target, thereby obtaining the second relative position of the second detected target relative to the second reference target.

[0095] To facilitate understanding, a specific example will be used as an illustration below.

[0096] See Figure 5The diagram shown illustrates the identification of a template image to determine a second relative position. The template image is an image taken of a qualified product, which can be a multi-functional ruler with hollow grooves for drawing triangles and ellipses. The user can select the upper boundary of the multi-functional ruler as the second reference target. The product quality inspection system 100 uses a deep learning algorithm to identify the template image, obtaining the coordinates of feature points of the second reference target and the second detected target (specifically including the triangular hollow groove and the elliptical hollow groove in the diagram). Then, based on the difference in the coordinates of the feature points, the second relative position of the second detected target relative to the second reference target is obtained. The feature point of the second reference target can be the left endpoint of the upper boundary of the product, i.e., the upper left corner vertex of the product. The feature points of the two second detected targets can be the center points of these two second detected targets. The second relative positions of the two second detected targets relative to the second reference target can be represented as (x1, y1) and (x2, y2).

[0097] In some possible implementations, the product quality inspection system 100 can also directly obtain the relative position of the second detected target with respect to the second reference target through deep learning algorithms, without first regressing the coordinates of the second detected target and the second reference target in the reference coordinate system, and then calculating the relative position based on those coordinates. For example, the product quality inspection system 100 can acquire labeled data with relative positions, train an image recognition model using supervised learning, and use this image recognition model to recognize a template image, thereby directly outputting the relative position of the second detected target with respect to the second reference target in the template image.

[0098] It should be noted that the product quality inspection system 100 can pre-execute the above-described S402 to S408 steps, storing the second relative position obtained from S402 to S408 in a storage device. When quality inspection of the product to be inspected is required, the second relative position can be retrieved from the storage device for quality inspection. In other embodiments, the product quality inspection system 100 can also identify a template image in real time, obtain at least one second target to be inspected and a second reference target (targets identified by the product quality inspection system 100 and specified by the user), and determine the second relative position of at least one second target to be inspected relative to the second reference target. The product quality inspection system 100 pre-determining the second relative position can reduce the computational load of quality inspection during the online stage, improve the efficiency of quality inspection during the online stage, and meet production needs.

[0099] It should also be noted that the product quality inspection method of this application embodiment may not necessarily execute S402 to S408 as described above. For example, when the product quality inspection system 100 is pre-configured with relevant parameters (such as reference values ​​or ranges of values ​​for the relative position of the inspected target relative to a reference target), the product quality inspection system 100 may not necessarily execute S402 to S408 as described above.

[0100] S410: Product quality inspection system 100 acquires the image to be inspected.

[0101] The product quality inspection system 100 provides a data upload interface, which can receive images to be inspected reported by a camera through the data upload interface. The image to be inspected refers to an image captured by photographing the product to be inspected. In some embodiments, the product to be inspected can run on a production line. When the product to be inspected reaches the shooting position, the camera can be triggered to take a picture. After taking a picture of the product to be inspected, the camera uploads the resulting image to the product quality inspection system 100 as the image to be inspected.

[0102] Since the product quality inspection system 100 can be used to perform quality inspection at different stages of product production, the product to be inspected can be a finished product or a semi-finished product. For example, when the product quality inspection system 100 performs quality inspection in the intermediate process of product production, the product to be inspected can be a semi-finished product. Alternatively, when the product quality inspection system 100 performs quality inspection in the final process of product production, the product to be inspected can be a finished product.

[0103] S412: Product quality inspection system 100 identifies the image to be inspected and obtains a first reference target and at least one first inspected target.

[0104] Similar to the second reference target and the second detected target, the first reference target refers to a reference target in the image to be inspected. It typically serves as a reference point for the first detected target in the image to determine its relative position. Therefore, the first reference target can be the boundary of the product to be inspected, such as its upper, lower, left, or right boundary. It should be noted that the first reference target is usually consistent with the second reference target. For example, if the second reference target is the upper boundary of a qualified product, the first reference target is the upper boundary of the product to be inspected. The first detected target can be a component or boundary of the product to be inspected.

[0105] In its implementation, the product quality inspection system 100 can use deep learning algorithms to identify the image to be inspected, generate candidate boxes, classify and regress these candidate boxes, and obtain the positions of at least one first detected target and the first reference target in the reference coordinate system of the image to be inspected. For example, the product quality inspection system 100 can first extract features from the image to be inspected using deep learning algorithms to obtain a feature map, then generate candidate boxes based on the feature map, and classify and regress the candidate boxes through inference to obtain the positions of the first reference target and at least one first detected target in the image to be inspected. Specifically, the positions of the first reference target and at least one first detected target can be their positions in the reference coordinate system of the image to be inspected.

[0106] S414: Product quality inspection system 100 determines at least one first inspected target with respect to a first reference target.

[0107] Specifically, the position of the first detected target can be represented by the coordinates of its feature points, and the position of the first reference target can be represented by the coordinates of its feature points. These feature points can be vertices, center points, etc. The feature points of the first detected target and the second detected target are of the same type, and the feature points of the first reference target and the second reference target are of the same type. Figure 6 To illustrate, since Figure 5 The feature point of the second reference target is the left endpoint of the upper boundary of the qualified product, and the feature point of the second target to be tested is the center point of the target to be tested in the qualified product. Therefore, the feature point of the first reference target can be the left endpoint of the upper boundary of the product to be tested, and the feature point of the first target to be tested is the center point of the target to be tested in the product to be tested.

[0108] For each first target to be inspected, the product quality inspection system 100 can calculate the difference between the coordinates of the feature points of the first target to be inspected and the coordinates of the feature points of the first reference target, thereby obtaining the first relative position of the first target to be inspected relative to the first reference target.

[0109] Figure 6To illustrate with an example of quality inspection of a product, a camera captures an image of the product to be inspected. The product quality inspection system 100 identifies the image and obtains at least one first target to be inspected and a first reference target. Then, it determines the first relative position of at least one first target to be inspected relative to the first reference target. Specifically, this is the relative position (x'1, y'1) of the center point of the triangular hollow groove of the product to be inspected relative to the left endpoint of the upper boundary of the product to be inspected, and the relative position (x'2, y'2) of the center point of the elliptical hollow groove of the product to be inspected relative to the left endpoint of the upper boundary of the product to be inspected.

[0110] It should be noted that the product quality inspection system 100 can also directly obtain the relative position of the first detected target relative to the first reference target through deep learning algorithms, without first regressing the coordinates of the first detected target and the first reference target in the reference coordinate system, and then calculating the relative position based on those coordinates. For example, the product quality inspection system 100 can acquire labeled data with relative positions, train an image recognition model using supervised learning, and use this image recognition model to recognize the image to be inspected, thereby directly outputting the relative position of the first detected target relative to the first reference target in the image to be inspected.

[0111] S416: The product quality inspection system 100 determines whether the quantity of the first inspected target is consistent with the quantity of the second inspected target. If yes, then execute S418; otherwise, execute S422.

[0112] Specifically, if the product to be inspected is qualified, the number of the first inspected targets in the image to be inspected is consistent with the number of the second inspected targets in the template image. Based on this, the product quality inspection system 100 can first count the first inspected targets in the template image and count the second inspected targets in the image to be inspected, and determine whether the number of the first inspected targets is consistent with the number of the second inspected targets. If they are inconsistent, it indicates that the product to be inspected is unqualified, and S422 can be executed. If they are consistent, the product quality inspection system 100 can execute S418 to perform further inspection.

[0113] It should be noted that the product quality inspection method of this application embodiment may also omit S416. For example, the product quality inspection system can directly execute S418 to perform quality inspection on the product. Specifically, by executing S416, the product quality inspection system 100 can pre-screen out some unqualified products, improving the efficiency of quality inspection. Especially when the image to be inspected includes multiple first targets, performing preliminary inspection on the product to be inspected based on the number of first targets and the number of second targets can effectively improve the efficiency of product quality inspection.

[0114] S418: The product quality inspection system 100 matches the first relative position with the second relative position. If they match, then execute S420; if they do not match, then execute S422.

[0115] Specifically, the product quality inspection system 100 can determine the coordinate difference between the first relative position and the second relative position. When the coordinate difference is greater than a preset threshold, it is determined that the first relative position and the second relative position do not match. The coordinate difference may include a difference in the horizontal coordinate and a difference in the vertical coordinate. When the coordinate difference is not greater than the preset threshold, it is determined that the first relative position and the second relative position match.

[0116] In this embodiment, the product quality inspection system 100 can subtract the abscissas of the first relative position and the second relative position, and then take the absolute value to obtain the abscissa difference. Similarly, the product quality inspection system 100 can subtract the ordinates of the first relative position and the second relative position, and then take the absolute value to obtain the ordinate difference. When the abscissa difference is greater than a first preset threshold dx, or the ordinate difference is greater than a second preset threshold dy, it is determined that the first relative position and the second relative position do not match. The first preset threshold and the second preset threshold can be pre-configured by the user or preset by the system.

[0117] When the product quality inspection system 100 determines that the first relative position of the first inspected target relative to the first reference target and the second relative position of the second inspected target relative to the second reference target do not match, the product to be inspected can be determined to be a non-conforming product. That is, the quality inspection result is non-conforming.

[0118] When the product quality inspection system 100 determines that the first relative position of each first inspected target relative to the first reference target and the second relative position of each inspected target relative to the second reference target match, the product to be inspected can be determined to be a qualified product. That is, the quality inspection result is qualified.

[0119] by Figure 6 , Figure 7 Examples will be provided for each. Figure 6In the example, the number of the first detected targets (referred to as target 1 and target 2) in the image to be detected is the same as the number of the second detected targets in the template image (both are 2). The relative position of target 1 with respect to the first reference target in the image to be detected is (x'1, y'1), and the relative position of target 2 with respect to the first reference target is (x'2, y'2). In the template image, the relative position of target 1 with respect to the second reference target is (x1, y1), and the relative position of target 2 with respect to the second reference target is (x2, y2). The coordinate difference between the first and second relative positions of target 1 is less than a preset threshold, i.e., |x1-x'1| < dx, |y1-y'1| < dy. Similarly, the coordinate difference between the first and second relative positions of target 2 is less than a preset threshold, i.e., |x2-x'2| < dx, |y2-y'2| < dy. Thus, the product quality inspection system 100 can determine that the product to be inspected is a qualified product.

[0120] exist Figure 7 In the example, the number of the first detected targets (referred to as target 1 and target 2) in the image to be detected is the same as the number of the second detected targets in the template image (both are 2). The relative position of target 1 with respect to the first reference target in the image to be detected is (x”1, y”1), and the relative position of target 2 with respect to the first reference target is (x”2, y”2). In the template image, the relative position of target 1 with respect to the second reference target is (x1, y1), and the relative position of target 2 with respect to the second reference target is (x2, y2). The coordinate difference between the first and second relative positions of target 1 is less than a preset threshold, i.e., |x1-x”1| < dx, |y1-y”1| < dy. However, the coordinate difference between the first and second relative positions of target 2 is greater than a preset threshold, i.e., |x2-x”2| > dx, |y2-y”2| > dy. Thus, the product quality inspection system 100 can determine that the product to be inspected is a non-conforming product.

[0121] It should be noted that S416 to 418 above are only one implementation of the product quality inspection system 100 in this application embodiment to perform quality inspection on the product to be inspected based on the first relative position and the second relative position. In other possible implementations in this application embodiment, the product quality inspection system 100 may also directly perform target matching based on the first relative position and the second relative position to perform quality inspection on the product to be inspected.

[0122] S420: The product quality inspection system 100 determines that the quality inspection result is qualified.

[0123] S422: The product quality inspection system 100 determines that the quality inspection result is unqualified.

[0124] In some possible implementations, the product quality inspection system 100 can also present the quality inspection results to the user. For example, the product quality inspection system 100 can present the quality inspection results to the user through a result display interface. Furthermore, when the quality inspection result indicates that the product to be inspected is a non-conforming product, the product quality inspection system 100 can also receive the user's processing instructions for the product to be inspected through the result display interface, so as to process the product to be inspected according to the processing instructions. The processing instructions may be instructions to discard the non-conforming product or to rework the non-conforming product; this embodiment does not limit this.

[0125] In some possible implementations, the product quality inspection system 100 can be connected to the control equipment of the production line. When the product quality inspection system 100 receives a processing instruction from the user on the result display interface for the product to be inspected, it can send the processing instruction to the control equipment so that the control equipment can process the product to be inspected accordingly.

[0126] Based on the above description, this application provides a product quality inspection method. This method introduces reference targets into both the image to be inspected and the template image. Quality inspection is achieved by comparing the relative positions of the detected target in the image to be inspected and the reference targets in their respective images. This eliminates the need for template registration and affine transformation, thus avoiding inaccuracies caused by slight product offsets or rotations in the camera's field of view. This reduces the false detection rate and demonstrates good robustness. Furthermore, the absence of affine transformation significantly reduces computational load and shortens inspection time, meeting production requirements.

[0127] Based on the product quality inspection method provided in the embodiments of this application, the embodiments of this application also provide a product quality inspection system 100 as described above. The product quality inspection system 100 provided in the embodiments of this application will now be described in conjunction with the accompanying drawings.

[0128] See Figure 1 The diagram shows the structure of a product quality inspection system 100, which includes an interaction module 102, an identification module 104, and a detection module 106. The interaction module 102 is specifically used to perform the aforementioned... Figure 4 In the method flow, S410, the identification module 104 is used to execute the aforementioned... Figure 4 In the method flow, S412 and S414, the detection module 106 is used to obtain the second relative position of at least one second detected target in the template image relative to the second reference target, and then obtain the quality detection result based on the first relative position and the second relative position obtained by executing S414.

[0129] In some possible implementations, the image to be detected includes multiple first targets to be detected. Accordingly, the detection module 106 is specifically used to perform quality inspection on the product to be detected based on the number of first targets to be detected in the image to be detected, the number of second targets to be detected in the template image, and the first relative position and the second relative position, so as to obtain a quality inspection result.

[0130] In some possible implementations, the detection module 104 is specifically used for:

[0131] The first relative position is matched with the second relative position to obtain a matching result;

[0132] When the matching result indicates that the first relative position and the second relative position do not match, the product to be tested is determined to be a non-conforming product.

[0133] In specific implementation, the detection module 104 can perform the aforementioned... Figure 4 S416 and S418 in the method flow are used to perform quality inspection on the products to be tested. In this method, by executing S416 first, the products to be tested can be preliminarily screened, thereby identifying some unqualified products in advance and improving the efficiency of quality inspection.

[0134] In some possible implementations, when the detection module 106 executes S418, it can first determine the coordinate difference between the first relative position and the second relative position. When the coordinate difference is greater than a preset threshold, it is determined that the first relative position and the second relative position do not match.

[0135] In some possible implementations, when the recognition module 104 executes S412, it can use a deep learning algorithm to recognize the image to be detected, generate candidate boxes, classify and regress the candidate boxes, and obtain the position of at least one first detected target in the image to be detected in the reference coordinate system of the image to be detected and the position of the first reference target in the image to be detected in the reference coordinate system of the image to be detected.

[0136] Accordingly, when the recognition module 104 executes S414, it can obtain the first relative position based on the position of the at least one first detected target in the reference coordinate system of the image to be detected and the position of the first reference target in the reference coordinate system of the image to be detected.

[0137] In some possible implementations, the interaction module 102 is also used to present the detection results to the user.

[0138] In some possible implementations, the interaction module 102 is further configured to: when the quality inspection result indicates that the product to be inspected is a non-conforming product, receive the user's processing instruction for the product to be inspected through the result display interface, so as to process the product to be inspected according to the processing instruction.

[0139] In some possible implementations, the first benchmark target is the boundary of the product to be tested, and the second benchmark target is the boundary of the qualified product.

[0140] In some possible implementations, the second relative position of at least one second detected target in the template image relative to a second reference target in the template image is pre-identified and calculated.

[0141] The product quality inspection system 100 according to the embodiments of this application can correspond to the execution of the methods described in the embodiments of this application, and the above and other operations and / or functions of each module / unit of the product quality inspection system 100 are respectively for implementing Figure 4 For the sake of brevity, the corresponding processes of each method in the illustrated embodiments will not be described in detail here.

[0142] This application also provides a computing device cluster. The computing device cluster includes at least one computing device, any one of which can originate from a cloud environment or an edge environment, or it can be a terminal. Specifically, the computing device cluster is used to implement, for example... Figure 1 The product quality inspection system 100 in the illustrated embodiment has the following functions.

[0143] Figure 8 A schematic diagram of the structure of a computing device cluster is provided, such as... Figure 8 As shown, the computing device cluster 80 includes multiple computing devices 800, each of which includes a bus 801, a processor 802, a communication interface 803, and a memory 804. The processor 802, the memory 804, and the communication interface 803 communicate with each other via the bus 801.

[0144] The 801 bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0145] The processor 802 can be any one or more of the following processors: central processing unit (CPU), graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).

[0146] The communication interface 803 is used for external communication. For example, the communication interface 803 is used to acquire images to be inspected, such as receiving reported images to be inspected through a data upload interface, presenting quality inspection results to the user, and receiving processing instructions from the user on the product to be inspected through the result display interface, etc.

[0147] Memory 804 may include volatile memory, such as random access memory (RAM). Memory 804 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0148] The memory 804 stores computer-readable instructions, which the processor 802 executes to cause the computing device cluster 80 to perform the aforementioned product quality inspection method (or implement the functions of the aforementioned product quality inspection system 100).

[0149] Specifically, in achieving Figure 1 In the case of the embodiment of the system shown, and Figure 1 The functions of the modules of the product quality inspection system 100 described herein, such as the interaction module 102, the identification module 104, and the detection module 106, are implemented through software and perform the following: Figure 1 The software or program code required for the functions of each module can be stored in at least one memory 804 in the computing device cluster 80. At least one processor 802 executes the program code stored in the memory 804 to cause the computing device cluster 80 to perform the aforementioned product quality inspection method.

[0150] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium capable of being stored by a computing device, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive). The computer-readable storage medium includes instructions that instruct a computing device or cluster of computing devices to execute the aforementioned product quality testing method.

[0151] This application also provides a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in this application are generated. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computing device, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer program product may be a software installation package. When any of the aforementioned product quality testing methods is required, the computer program product may be downloaded and executed on a computing device or cluster of computing devices.

[0152] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

Claims

1. A product quality testing method, characterized in that, The method includes: Acquire an image to be inspected, wherein the image to be inspected is an image obtained by taking a picture of the product to be inspected; Identify the image to be detected, and obtain at least one first detected target and a first reference target in the image to be detected; Determine the first relative position of the at least one first detected target relative to the first reference target; Obtain the second relative position of at least one second detected target in a template image relative to a second reference target in the template image, wherein the template image is an image obtained by photographing a qualified product; Based on the first relative position and the second relative position, the product to be tested is subjected to quality testing to obtain the quality testing results.

2. The method according to claim 1, characterized in that, When the image to be detected includes multiple first targets to be detected, the step of performing quality inspection on the product to be detected based on the first relative position and the second relative position to obtain a quality inspection result includes: Based on the number of the first detected targets in the image to be detected, the number of the second detected targets in the template image, and the first relative position and the second relative position, the product to be detected is subjected to quality inspection to obtain quality inspection results.

3. The method according to claim 1, characterized in that, The step of performing quality testing on the product to be tested based on the first relative position and the second relative position to obtain quality testing results includes: The first relative position is matched with the second relative position to obtain a matching result; When the matching result indicates that the first relative position and the second relative position do not match, the product to be tested is determined to be a non-conforming product.

4. The method according to claim 3, characterized in that, The step of matching the first relative position with the second relative position includes: Determine the coordinate difference between the first relative position and the second relative position; When the coordinate difference is greater than a preset threshold, it is determined that the first relative position and the second relative position do not match.

5. The method according to any one of claims 1 to 4, characterized in that, The step of identifying the image to be detected and obtaining at least one first detected target and a first reference target in the image to be detected includes: The image to be detected is identified by a deep learning algorithm to generate candidate boxes. The candidate boxes are then classified and regressed to obtain the position of at least one first detected target in the image to be detected in the reference coordinate system of the image to be detected and the position of the first reference target in the image to be detected in the reference coordinate system of the image to be detected. Determining the first relative position of the at least one first detected target relative to the first reference target includes: The first relative position is obtained based on the position of the at least one first detected target in the reference coordinate system of the image to be detected and the position of the first reference target in the reference coordinate system of the image to be detected.

6. The method according to any one of claims 1 to 4, characterized in that, The method further includes: Present the quality test results to the user.

7. The method according to claim 6, characterized in that, When the quality inspection result indicates that the product to be inspected is a non-conforming product, the method further includes: The system receives processing instructions from the user regarding the product to be tested through the results display interface, so as to process the product to be tested according to the processing instructions.

8. The method according to any one of claims 1 to 4, characterized in that, The first benchmark target is the boundary of the product to be tested, and the second benchmark target is the boundary of the qualified product.

9. The method according to any one of claims 1 to 4, characterized in that, The second relative position of at least one second detected target in the template image with respect to a second reference target in the template image is pre-identified and calculated.

10. A product quality inspection system, characterized in that, The system includes: The interaction module is used to acquire the image to be detected, which is an image of the product to be detected captured by photography; The recognition module is used to recognize the image to be detected and obtain at least one first detected target and a first reference target in the image to be detected; The identification module is further configured to determine the first relative position of the at least one first detected target relative to the first reference target; The detection module is used to obtain the second relative position of at least one second target to be detected in a template image relative to a second reference target in the template image, wherein the template image is an image obtained by taking a picture of a qualified product, and the product to be detected is subjected to quality inspection based on the first relative position and the second relative position to obtain a quality inspection result.

11. The system according to claim 10, characterized in that, The detection module is specifically used for: When the image to be detected includes multiple first targets to be detected, the product to be detected is subjected to quality inspection based on the number of first targets to be detected in the image to be detected, the number of second targets to be detected in the template image, and the first relative position and the second relative position, so as to obtain the quality inspection result.

12. The system according to claim 10, characterized in that, The detection module is specifically used for: The first relative position is matched with the second relative position to obtain a matching result; When the matching result indicates that the first relative position and the second relative position do not match, the product to be tested is determined to be a non-conforming product.

13. The system according to claim 12, characterized in that, The detection module is specifically used for: Determine the coordinate difference between the first relative position and the second relative position; When the coordinate difference is greater than a preset threshold, it is determined that the first relative position and the second relative position do not match.

14. The system according to any one of claims 10 to 13, characterized in that, The identification module is specifically used for: The image to be detected is identified by a deep learning algorithm to generate candidate boxes. The candidate boxes are then classified and regressed to obtain the position of at least one first detected target in the image to be detected in the reference coordinate system of the image to be detected and the position of the first reference target in the image to be detected in the reference coordinate system of the image to be detected. The first relative position is obtained based on the position of the at least one first detected target in the reference coordinate system of the image to be detected and the position of the first reference target in the reference coordinate system of the image to be detected.

15. The system according to any one of claims 10 to 13, characterized in that, The interaction module is also used for: Present the quality test results to the user.

16. The system according to claim 15, characterized in that, The interaction module is also used for: When the quality inspection result indicates that the product to be inspected is a non-conforming product, the system receives the user's processing instruction for the product to be inspected through the result display interface, so as to process the product to be inspected according to the processing instruction.

17. The system according to any one of claims 10 to 13, characterized in that, The first benchmark target is the boundary of the product to be tested, and the second benchmark target is the boundary of the qualified product.

18. The system according to any one of claims 10 to 13, characterized in that, The second relative position of at least one second detected target in the template image with respect to a second reference target in the template image is pre-identified and calculated.

19. A computing device cluster, characterized in that, The computing device cluster includes at least one computing device, the at least one computing device including at least one processor and at least one memory, the at least one memory storing computer-readable instructions, the at least one processor executing the computer-readable instructions, causing the computing device cluster to perform the method as described in any one of claims 1 to 9.

20. A computer-readable storage medium, characterized in that, Includes computer-readable instructions that, when executed on a computing device or cluster of computing devices, cause the computing device or cluster of computing devices to perform the method as described in any one of claims 1 to 9.

21. A computer program product, characterized in that, Includes computer-readable instructions that, when executed on a computing device or cluster of computing devices, cause the computing device or cluster of computing devices to perform the method as described in any one of claims 1 to 9.

Images