Motor finished product automatic inspection system and method

Abstract

The invention discloses a motor finished product automatic inspection system and method, and the method comprises the steps: collecting a motor appearance detection image through deploying an industrial camera at the tail end of a production line, carrying out the image analysis of the motor appearance detection image through employing an image processing technology based on deep learning, and carrying out the gray processing of the motor appearance detection image, the method comprises the following steps of: extracting a stable motor appearance feature by using a detection image of the motor appearance, and simultaneously combining an appearance reference image of a qualified motor to realize intelligent identification of the abnormal motor appearance based on the appearance difference between the motor appearance feature in the reference image and the motor appearance feature in the detection image. Therefore, the detection efficiency and accuracy of the production line can be effectively improved, and the problems of missing detection and false detection caused by subjectivity and fatigue of manual detection are reduced.

Description

Technical Field

[0001] The present application relates to the field of intelligent inspection, and more specifically, to a system and method for automatic inspection of finished motor products. Background Art

[0002] In the field of industrial automation, especially in the core sector of motor manufacturing, the quality and performance of motor products are not only important indicators to measure the level of production technology, but also directly related to the stability and sustainable development of the entire industrial chain. With the advancement of science and technology and the growing market demand, the quality requirements for motor products are getting higher and higher.

[0003] Traditionally, the quality inspection of motors is highly dependent on manual operation. Workers need to use visual observation, touch or simple tools to detect surface scratches, dents, dimensional deviations and other problems of the motors. However, with the continuous expansion of production scale, the number of finished motors on the production line has increased dramatically, and this manual inspection method has gradually revealed its limitations. First, manual inspection is inefficient and difficult to cope with the huge inspection tasks brought about by large-scale production; second, due to the existence of human factors, such as differences in subjective judgment and visual fatigue, the inspection results are often unstable and inconsistent, increasing the risk of missed inspections and false inspections; third, manual inspection is costly, involving not only a large investment in human resources, but also additional costs such as rework and waste disposal due to inspection errors.

[0004] In today's highly automated and intelligent modern industrial system, manual inspection has become a bottleneck restricting the improvement of production efficiency and product quality. Therefore, in order to achieve an efficient, accurate and sustainable production model, an automatic inspection system and method for motor finished products is expected. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a motor finished product automatic inspection system and method, which collects motor appearance inspection images by deploying industrial cameras at the end of the production line, and uses deep learning-based image processing technology to perform image analysis on the motor appearance inspection images, and grayscales the motor appearance inspection images to extract stable motor appearance features. At the same time, combined with the appearance reference image of a qualified motor, based on the appearance difference between the motor appearance features in the reference image and the motor appearance features in the inspection image, intelligent recognition of motor appearance abnormalities is achieved. In this way, the inspection efficiency and accuracy of the production line can be effectively improved, and the problems of missed detection and false detection caused by the subjectivity and fatigue of manual inspection can be reduced.

[0006] According to one aspect of the present application, a method for automatically inspecting a finished motor product is provided, comprising: Obtaining motor appearance inspection images collected by an industrial camera installed at the end of the production line, and extracting qualified motor appearance reference images from the database; Performing motor appearance feature extraction on the motor appearance detection image and the qualified motor appearance reference image respectively to obtain a motor appearance detection feature map and a qualified motor appearance reference feature map; Performing feature saliency processing on the motor appearance detection feature map to obtain a motor appearance salient detection feature map; Based on the difference in appearance features between the motor appearance significant detection feature map and the qualified motor appearance reference feature map, determine whether there is an abnormality in the motor appearance.

[0007] According to another aspect of the present application, a motor finished product automatic inspection system is provided, comprising: An image acquisition module is used to acquire a motor appearance inspection image collected by an industrial camera installed at the end of the production line, and to extract a qualified motor appearance reference image from a database; a motor appearance feature extraction module, used to extract motor appearance features from the motor appearance detection image and the qualified motor appearance reference image respectively to obtain a motor appearance detection feature map and a qualified motor appearance reference feature map; A feature saliency module, used for performing feature saliency processing on the motor appearance detection feature map to obtain a motor appearance salient detection feature map; The abnormality inspection result generating module is used to determine whether there is an abnormality in the motor appearance based on the difference in appearance features between the motor appearance significant detection feature map and the qualified motor appearance reference feature map.

[0008] Compared with the prior art, the present application provides a motor finished product automatic inspection system and method, which collects motor appearance inspection images by deploying industrial cameras at the end of the production line, and uses deep learning-based image processing technology to analyze the motor appearance inspection images, and grayscales the motor appearance inspection images to extract stable motor appearance features. At the same time, combined with the appearance reference image of a qualified motor, based on the appearance difference between the motor appearance features in the reference image and the motor appearance features in the inspection image, intelligent recognition of motor appearance anomalies is achieved. In this way, the inspection efficiency and accuracy of the production line can be effectively improved, and the problems of missed detection and false detection caused by the subjectivity and fatigue of manual inspection can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 is a flow chart of a method for automatically inspecting finished motor products according to an embodiment of the present application; Figure 2 A data flow diagram of a method for automatically inspecting finished motor products according to an embodiment of the present application; Figure 3 is a flowchart of sub-step S2 of the method for automatically inspecting finished motor products according to an embodiment of the present application; Figure 4 is a flowchart of sub-step S4 of the method for automatically inspecting finished motor products according to an embodiment of the present application; Figure 5 4 is a block diagram of a motor finished product automatic inspection system according to an embodiment of the present application. DETAILED DESCRIPTION

[0011] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0012] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0013] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0014] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0015] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0016] Traditionally, the quality inspection of motors is highly dependent on manual operation. Workers need to observe with the naked eye, touch with hands or use simple tools to detect surface scratches, dents, dimensional deviations and other problems of motors. However, with the continuous expansion of production scale, the number of motor products on the production line has increased dramatically, and this manual inspection method has gradually revealed its limitations. First, manual inspection is inefficient and difficult to cope with the huge inspection tasks brought by large-scale production; second, due to the existence of human factors, such as differences in subjective judgment and visual fatigue, the inspection results are often unstable and inconsistent, increasing the risk of missed inspections and false inspections; third, manual inspection is costly, involving not only a large investment in human resources, but also additional costs such as rework and waste disposal caused by inspection errors. In today's highly automated and intelligent modern industrial system, manual inspection has become a bottleneck restricting the improvement of production efficiency and the upgrading of product quality. Therefore, in order to achieve an efficient, accurate and sustainable production model, an automatic inspection system and method for motor finished products is expected.

[0017] In the technical solution of the present application, a method for automatically inspecting finished motor products is proposed. Figure 1 Flow chart of a method for automatically inspecting finished motor products according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the method for automatic inspection of finished motor products according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the method for automatic inspection of motor finished products according to an embodiment of the present application includes the steps of: S1, acquiring a motor appearance inspection image collected by an industrial camera installed at the end of the production line, and extracting a qualified motor appearance reference image from a database; S2, performing motor appearance feature extraction on the motor appearance inspection image and the qualified motor appearance reference image to obtain a motor appearance inspection feature map and a qualified motor appearance reference feature map; S3, performing feature significant processing on the motor appearance inspection feature map to obtain a motor appearance significant detection feature map; S4, determining whether there is any abnormality in the motor appearance based on the difference in appearance features between the motor appearance significant detection feature map and the qualified motor appearance reference feature map.

[0018] In particular, the S1 obtains the motor appearance inspection image collected by the industrial camera installed at the end of the production line, and extracts the qualified motor appearance reference image from the database. It should be understood that the motor appearance inspection image provides the actual appearance information of the motor to be inspected, and the qualified motor appearance reference image represents the ideal state or acceptable range that the motor appearance should reach. Using the reference image as a benchmark, and comparing and analyzing it with the motor appearance inspection image, any abnormality or defect in the appearance of the motor to be inspected can be more accurately identified.

[0019] In particular, the step S2 extracts motor appearance features from the motor appearance detection image and the qualified motor appearance reference image to obtain a motor appearance detection feature map and a qualified motor appearance reference feature map. In particular, in a specific example of the present application, Figure 3 As shown, the S2 includes: S21, performing grayscale processing on the motor appearance detection image to obtain a motor appearance detection grayscale image; S22, inputting the motor appearance detection grayscale image and the qualified motor appearance reference image into a motor appearance feature extractor based on the MBCNet model to obtain the motor appearance detection feature map and the qualified motor appearance reference feature map.

[0020] Specifically, in step S21, grayscale processing is performed on the motor appearance detection image to obtain a motor appearance detection grayscale image. Here, by grayscale processing the motor appearance detection image, a clearer and easier to analyze image can be obtained, thereby improving the accuracy and efficiency of motor appearance detection.

[0021] Specifically, in S22, the motor appearance detection grayscale image and the qualified motor appearance reference image are input into the motor appearance feature extractor based on the MBCNet model to obtain the motor appearance detection feature map and the qualified motor appearance reference feature map. In order to accurately and comprehensively extract the global structural contour features and texture detail features of the motor appearance to achieve high-precision motor defect detection, the present application adopts the MBCNet (Multi-Branch Convolutional Network) model as a motor appearance feature extractor, and processes the motor appearance detection grayscale image and the qualified motor appearance reference image separately, so as to utilize its multi-branch structure, focus on the coarse-grained global features and fine-grained local features of the image respectively, and achieve comprehensive coverage and accurate capture of motor appearance features by extracting and fusing features of different scales of the image, thereby generating a motor appearance detection feature map and a qualified motor appearance reference feature map.

[0022] It is worth mentioning that in other specific examples of the present application, the motor appearance detection image and the qualified motor appearance reference image can be respectively subjected to motor appearance feature extraction in other ways to obtain a motor appearance detection feature map and a qualified motor appearance reference feature map, for example: inputting the motor appearance detection image and the qualified motor appearance reference image; preprocessing the motor appearance detection image and the qualified motor appearance reference image; using feature extraction methods such as local binary patterns, oriented gradient histograms, scale-invariant feature transforms, and deep learning convolutional neural networks to extract features from the motor appearance detection image and the qualified motor appearance reference image to obtain the motor appearance detection feature map and the qualified motor appearance reference feature map.

[0023] In particular, the S3 performs feature saliency processing on the motor appearance detection feature map to obtain a motor appearance salient detection feature map. In a specific example of the present application, the motor appearance detection feature map is input into a feature distribution gradient mask salient device to obtain the motor appearance salient detection feature map. Here, in order to further enhance the defect features such as scratches, dents, cracks, etc. that may exist in the motor appearance detection feature map, in the technical solution of the present application, the motor appearance detection feature map is input into a feature distribution gradient mask salient device for feature enhancement processing. Specifically, the feature distribution gradient mask salient device first generates a motor appearance detection feature gradient amplitude distribution map by calculating the gradient amplitude value of each position in the motor appearance detection feature map to reveal the local change intensity of different positions in the motor appearance detection feature map. Then, the gradient amplitude local description operator is calculated by measuring the significance of the gradient amplitude value of each position relative to the gradient amplitude value of other positions in the local neighborhood to generate a motor appearance detection feature gradient amplitude local salient distribution map, highlighting the significant change area in the motor appearance detection feature map. Then, based on the gated mask mechanism, the motor appearance detection feature gradient amplitude local significant distribution map is nonlinearly activated to generate a gated mask, and the original motor appearance detection feature map is feature weighted and masked to obtain the motor appearance significant detection feature map. In this way, by combining the gradient significance information of the motor appearance detection feature map and strengthening its significant feature area, it is helpful to more effectively perform subsequent abnormality recognition and judgment, thereby effectively improving the accuracy and reliability of defect detection.

[0024] In an embodiment of the present application, the motor appearance detection feature map is subjected to feature saliency processing to obtain a motor appearance salient detection feature map, including: calculating the multi-directional gradient value distribution of each position in the motor appearance detection feature map, and determining the gradient amplitude value of each position in the motor appearance detection feature map based on the multi-directional gradient value distribution of each position to obtain a motor appearance detection feature gradient amplitude distribution map; calculating the gradient amplitude local description operator of each position in the motor appearance detection feature gradient amplitude distribution map to obtain a motor appearance detection feature gradient amplitude local salient distribution map; inputting the motor appearance detection feature gradient amplitude local salient distribution map into a gated masker based on a GELU function to obtain a gradient amplitude local salient gated mask map; calculating the position point multiplication between the gradient amplitude local salient gated mask map and the motor appearance detection feature map to obtain the motor appearance salient detection feature map.

[0025] Among them, calculating the gradient amplitude local description operator of each position in the motor appearance detection feature gradient amplitude distribution map to obtain the motor appearance detection feature gradient amplitude local significant distribution map, including: determining the scale of the local neighborhood, calculating the mean of the difference between the gradient amplitude value at a predetermined position in the motor appearance detection feature gradient amplitude distribution map and the gradient amplitude values ​​at other positions in the local neighborhood to obtain the gradient amplitude local description operator corresponding to the predetermined position.

[0026] In summary, in the above embodiment, the local description operator of the gradient amplitude at each position in the motor appearance detection feature gradient amplitude distribution map is calculated to obtain the motor appearance detection feature gradient amplitude local significant distribution map, including: processing the motor appearance detection feature map with the following feature distribution gradient mask formula to obtain the motor appearance significant detection feature map, wherein the feature distribution gradient mask formula is: ;in, The first figure represents the motor appearance detection characteristic diagram The gray value of the position, The motor appearance detection feature diagram is shown in The gradient value in the horizontal direction of the pixel point. The motor appearance detection feature diagram is shown in The gradient value in the ordinate direction of the pixel point. The motor appearance detection characteristic diagram is shown in FIG. The channel direction gradient value of the position pixel point, The first figure represents the motor appearance detection characteristic diagram The gradient magnitude value at the position, The motor appearance detection characteristic gradient amplitude distribution diagram is shown The set of gradient magnitude values ​​in the local neighborhood centered at position, represents the number of gradient magnitude values ​​in the set of gradient magnitude values ​​in the local neighborhood, and Respectively represent the offset in the horizontal axis direction, the vertical axis direction and the channel direction, The first figure represents the motor appearance detection characteristic diagram The local description operator of the gradient magnitude of the position, represents the GELU function, The first part of the gradient magnitude local significant gated mask map is represented by The local description operator of the gradient magnitude of the position, represents the local significant gated mask map of the gradient magnitude, It means point multiplication by position. A graph showing significant detection features of the motor appearance.

[0027] In particular, the step S4 determines whether the motor appearance is abnormal based on the difference in appearance features between the motor appearance significant detection feature map and the qualified motor appearance reference feature map. In particular, in a specific example of the present application, Figure 4 As shown, the S4 includes: S41, calculating the appearance difference semantic measurement coefficient between the motor appearance significant detection feature map and the qualified motor appearance reference feature map; S42, determining whether there is any abnormality in the motor appearance based on the comparison between the appearance difference semantic measurement coefficient and a preset threshold.

[0028] Specifically, the S41 calculates the appearance difference semantic measurement coefficient between the motor appearance significant detection feature map and the qualified motor appearance reference feature map. The motor appearance significant detection feature map is compared and analyzed with the qualified motor appearance reference feature map to determine whether there is an abnormality or defect in the appearance of the motor to be detected. In the technical solution of the present application, a comparative analysis method of calculating the appearance difference semantic measurement coefficient between the motor appearance significant detection feature map and the qualified motor appearance reference feature map is adopted to quantitatively evaluate the degree of difference between the motor to be detected and the qualified standard. In a specific example of the present application, the Mahalanobis distance between the motor appearance significant detection feature map and the qualified motor appearance reference feature map is used as the appearance difference semantic measurement coefficient. It should be understood that the Mahalanobis distance is a distance measurement method used to measure the similarity between two probability distributions. Compared with the Euclidean distance, it can more accurately reflect the similarity between samples by considering the covariance structure of the data, thereby providing a more accurate appearance difference semantic measurement result of the motor to be detected.

[0029] In a preferred example, calculating the appearance difference semantic measurement coefficient between the motor appearance salient detection feature map and the qualified motor appearance reference feature map comprises the steps of: Cascading the motor appearance significant detection feature map and the qualified motor appearance reference feature map along a channel to obtain a motor appearance detection-qualified motor appearance reference cascade feature map; Determine the eigenvalue mean and eigenvalue standard deviation corresponding to the motor appearance detection-qualified motor appearance reference cascade feature map; Multiply the motor appearance detection-qualified motor appearance reference cascade feature map with the point-subtraction feature map of the feature value mean and the feature value standard deviation to obtain a first motor appearance detection-qualified motor appearance reference cascade intermediate feature map, and multiply the motor appearance detection-qualified motor appearance reference cascade feature map with the point-subtraction feature map of the feature value standard deviation and the feature value mean to obtain a second motor appearance detection-qualified motor appearance reference cascade intermediate feature map; After multiplying the bit-by-bit reciprocal of the second motor appearance detection-qualified motor appearance reference cascade intermediate feature map by the first motor appearance detection-qualified motor appearance reference cascade intermediate feature map, calculate the bit-by-bit logarithm with base 2 to obtain the motor appearance detection-qualified motor appearance reference cascade correction feature map; The square root of the quotient of the eigenvalue mean divided by the eigenvalue standard deviation is multiplied by the weight hyperparameter, and then the result is added to the motor appearance detection-qualified motor appearance reference cascade correction feature map to obtain an optimized motor appearance detection-qualified motor appearance reference cascade feature map; Splitting the optimized motor appearance detection-qualified motor appearance reference cascade feature map into an optimized motor appearance significant detection feature map and an optimized qualified motor appearance reference feature map; An appearance difference semantic metric coefficient between the optimized motor appearance salient detection feature map and the optimized qualified motor appearance reference feature map is calculated.

[0030] That is, considering that the motor appearance significant detection feature map and the qualified motor appearance reference feature map respectively represent the feature distribution gradient significantly enhanced image semantic features of the motor appearance detection image and the image semantic features of the qualified motor appearance reference image, when calculating the appearance difference semantic measurement coefficients therebetween to perform aggregate mapping to the common image difference semantic measurement domain, differences in the attributes of the image semantic feature populations will cause differences in the fairness of the aggregate mapping alignment, thereby affecting the calculation accuracy of the appearance difference semantic measurement coefficients between the motor appearance significant detection feature map and the qualified motor appearance reference feature map.

[0031] Therefore, considering that the motor appearance salient detection feature map and the qualified motor appearance reference feature map are based on the aggregate mapping toward the common image difference semantic measurement domain, and the aggregation alignment fairness difference caused by the attribute level difference of the data population, in order to improve the aggregation inclusiveness of the motor appearance salient detection feature map and the qualified motor appearance reference feature map under the diversity of feature distribution, the motor appearance detection-qualified motor appearance reference cascade feature map obtained by cascading the motor appearance salient detection feature map and the qualified motor appearance reference feature map is constrained by taking the cross probability value constraint of the motor appearance detection-qualified motor appearance reference cascade feature map obtained by cascading the motor appearance salient detection feature map and the qualified motor appearance reference feature map as the interactive fairness objective representation, to improve the aggregation inclusiveness of the motor appearance detection-qualified motor appearance reference feature map under the diversity of feature distribution. The interactive propagation of group feature information of the qualified motor appearance reference cascade feature map is corrected, and the unified statistical feature response interaction based on the motor appearance detection-qualified motor appearance reference cascade feature map is used as the multi-level fairness target bias of the feature distribution, so as to achieve a robust unified representation of the distribution fairness of the motor appearance detection-qualified motor appearance reference cascade feature map, form a fair collaboration paradigm under the feature distribution framework of the motor appearance detection-qualified motor appearance reference cascade feature map, and improve the calculation accuracy of the appearance difference semantic measurement coefficient between the optimized motor appearance saliency detection feature map and the optimized qualified motor appearance reference feature map.

[0032] Specifically, the step S42 determines whether the motor appearance is abnormal based on a comparison between the appearance difference semantic metric coefficient and a preset threshold. In a specific example of the present application, in response to the appearance difference semantic metric coefficient being greater than the preset threshold, it indicates that the motor appearance is abnormal.

[0033] It is worth mentioning that in other specific examples of the present application, it is also possible to determine whether the motor appearance is abnormal based on the appearance feature difference between the motor appearance significant detection feature map and the qualified motor appearance reference feature map by other methods, for example: input the motor appearance significant detection feature map and the qualified motor appearance reference feature map; use a difference measurement method such as Euclidean distance, Manhattan distance or cosine similarity to calculate the difference between the motor appearance significant detection feature map and the qualified motor appearance reference feature map; set a threshold value for distinguishing between normal and abnormal appearance feature differences; and compare the calculated feature difference with the set threshold value. If the difference is greater than the threshold value, it is considered that the motor appearance is abnormal.

[0034] In summary, the method for automatic inspection of finished motor products according to the embodiment of the present application is explained, which collects motor appearance inspection images by deploying industrial cameras at the end of the production line, and uses deep learning-based image processing technology to analyze the motor appearance inspection images, and grayscales the motor appearance inspection images to extract stable motor appearance features. At the same time, combined with the appearance reference image of a qualified motor, based on the appearance difference between the motor appearance features in the reference image and the motor appearance features in the inspection image, intelligent recognition of motor appearance anomalies is achieved. In this way, the inspection efficiency and accuracy of the production line can be effectively improved, and the problems of missed detection and false detection caused by the subjectivity and fatigue of manual inspection can be reduced.

[0035] Furthermore, a motor finished product automatic inspection system is also provided.

[0036] Figure 5 FIG. 1 is a block diagram of a motor finished product automatic inspection system according to an embodiment of the present application. Figure 5 As shown, the motor finished product automatic inspection system 300 according to the embodiment of the present application includes: an image acquisition module 310, which is used to acquire a motor appearance inspection image collected by an industrial camera installed at the end of the production line, and extract a qualified motor appearance reference image from a database; a motor appearance feature extraction module 320, which is used to perform motor appearance feature extraction on the motor appearance inspection image and the qualified motor appearance reference image respectively to obtain a motor appearance inspection feature map and a qualified motor appearance reference feature map; a feature significant module 330, which is used to perform feature significant processing on the motor appearance inspection feature map to obtain a motor appearance significant detection feature map; an abnormality inspection result generation module 340, which is used to determine whether there is an abnormality in the motor appearance based on the difference in appearance features between the motor appearance significant detection feature map and the qualified motor appearance reference feature map.

[0037] As described above, the motor finished product automatic inspection system 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a motor finished product automatic inspection algorithm. In one possible implementation, the motor finished product automatic inspection system 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the motor finished product automatic inspection system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the motor finished product automatic inspection system 300 can also be one of the many hardware modules of the wireless terminal.

[0038] Alternatively, in another example, the motor finished product automatic inspection system 300 and the wireless terminal may also be separate devices, and the motor finished product automatic inspection system 300 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0039] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for automatically inspecting a finished motor product, characterized in that: include: Obtaining motor appearance inspection images collected by an industrial camera installed at the end of the production line, and extracting qualified motor appearance reference images from the database; Performing motor appearance feature extraction on the motor appearance detection image and the qualified motor appearance reference image respectively to obtain a motor appearance detection feature map and a qualified motor appearance reference feature map; Performing feature saliency processing on the motor appearance detection feature map to obtain a motor appearance salient detection feature map; Based on the difference in appearance features between the motor appearance significant detection feature map and the qualified motor appearance reference feature map, determine whether there is an abnormality in the motor appearance.

2. The method for automatically inspecting finished motor products according to claim 1, characterized in that: The motor appearance detection image and the qualified motor appearance reference image are respectively subjected to motor appearance feature extraction to obtain a motor appearance detection feature map and a qualified motor appearance reference feature map, including: Performing grayscale processing on the motor appearance detection image to obtain a motor appearance detection grayscale image; The motor appearance detection grayscale image and the qualified motor appearance reference image are input into the motor appearance feature extractor based on the MBCNet model to obtain the motor appearance detection feature map and the qualified motor appearance reference feature map.

3. The method for automatically inspecting finished motor products according to claim 2, characterized in that: Performing feature saliency processing on the motor appearance detection feature map to obtain a motor appearance salient detection feature map, including: The motor appearance detection feature map is input into a feature distribution gradient mask salient device to obtain the motor appearance salient detection feature map.

4. The method for automatically inspecting finished motor products according to claim 3, characterized in that: Inputting the motor appearance detection feature map into a feature distribution gradient mask salient device to obtain the motor appearance salient detection feature map, including: Calculating the multi-directional gradient value distribution of each position in the motor appearance detection feature map, and determining the gradient amplitude value of each position in the motor appearance detection feature map based on the multi-directional gradient value distribution of each position to obtain a motor appearance detection feature gradient amplitude distribution map; Calculating the local description operator of the gradient amplitude at each position in the motor appearance detection feature gradient amplitude distribution map to obtain a motor appearance detection feature gradient amplitude local significant distribution map; Inputting the motor appearance detection feature gradient amplitude local significant distribution map into a gated masker based on a GELU function to obtain a gradient amplitude local significant gated mask map; The motor appearance salient detection feature map is obtained by calculating the point-by-point multiplication between the gradient amplitude local salient gated mask map and the motor appearance detection feature map.

5. The method for automatically inspecting finished motor products according to claim 4, characterized in that: Calculating the local description operator of the gradient amplitude at each position in the motor appearance detection feature gradient amplitude distribution map to obtain the motor appearance detection feature gradient amplitude local significant distribution map, including: The scale of the local neighborhood is determined, and the average of the difference between the gradient amplitude value at a predetermined position in the motor appearance detection feature gradient amplitude distribution map and the gradient amplitude values ​​at other positions in the local neighborhood is calculated to obtain a gradient amplitude local description operator corresponding to the predetermined position.

6. The method for automatically inspecting finished motor products according to claim 5, characterized in that: Determining whether the motor appearance is abnormal based on the difference in appearance features between the motor appearance significant detection feature map and the qualified motor appearance reference feature map includes: Calculating a semantic metric coefficient of appearance difference between the motor appearance salient detection feature map and the qualified motor appearance reference feature map; Based on the comparison between the appearance difference semantic measurement coefficient and a preset threshold, it is determined whether there is any abnormality in the appearance of the motor.

7. The method for automatically inspecting finished motor products according to claim 6, characterized in that: Calculating the appearance difference semantic metric coefficient between the motor appearance salient detection feature map and the qualified motor appearance reference feature map, including: The Mahalanobis distance between the motor appearance salient detection feature map and the qualified motor appearance reference feature map is calculated as the appearance difference semantic measurement coefficient.

8. The method for automatically inspecting finished motor products according to claim 7, characterized in that: Calculating the Mahalanobis distance between the motor appearance salient detection feature map and the qualified motor appearance reference feature map as the appearance difference semantic measurement coefficient includes: Performing global mean pooling along the channel dimension on the motor appearance salient detection feature map and the qualified motor appearance reference feature map to obtain a motor appearance salient detection feature vector and a qualified motor appearance reference feature vector; Calculating the position difference between the motor appearance significant detection feature vector and the qualified motor appearance reference feature vector to obtain an appearance difference semantic feature vector; Calculating a covariance matrix between the motor appearance salient detection feature vector and the qualified motor appearance reference feature vector; The product result between the transposed vector of the appearance difference semantic feature vector, the inverse matrix of the covariance matrix and the appearance difference semantic feature vector is calculated, and then the square root of the product result is calculated to obtain the appearance difference semantic measurement coefficient.

9. A motor finished product automatic inspection system, characterized in that: include: An image acquisition module is used to acquire a motor appearance inspection image collected by an industrial camera installed at the end of the production line, and to extract a qualified motor appearance reference image from a database; a motor appearance feature extraction module, used to extract motor appearance features from the motor appearance detection image and the qualified motor appearance reference image respectively to obtain a motor appearance detection feature map and a qualified motor appearance reference feature map; A feature saliency module, used for performing feature saliency processing on the motor appearance detection feature map to obtain a motor appearance salient detection feature map; The abnormality inspection result generating module is used to determine whether there is an abnormality in the motor appearance based on the difference in appearance features between the motor appearance significant detection feature map and the qualified motor appearance reference feature map.

Images