Automatic essence production line and method thereof

Abstract

The invention relates to the technical field of quality detection, and particularly discloses an automatic essence production line and a method thereof.A high-resolution camera is installed on the production line, a sealing image of an essence bottle is shot in real time, and the sealing image is analyzed and processed through the deep learning technology; and fine-grained seal image space structure features are extracted from the seal image space structure features so as to intelligently identify seal quality defects. By means of the mode, automatic and non-contact detection of essence bottle sealing can be achieved, outflow of defective products is reduced, and the stability of product quality is ensured.

Description

Technical Field

[0001] The present application relates to the technical field of quality inspection, and more specifically, to an automated production line for essence and a method thereof. Background Art

[0002] In the modern cosmetics industry, essences are deeply loved by consumers because of their rich and highly effective active ingredients and remarkable beauty effects. In the production process of essences, the sealing quality of the essence bottles is the key to ensuring the sterility and effectiveness of the product. Any product with a loose seal that enters the market will cause a major blow to the brand reputation and quickly erode the trust of consumers. Therefore, manufacturers usually adopt strict quality control strategies to ensure the safety and effectiveness of essence products.

[0003] However, in the existing essence production process, the detection of sealing quality mainly relies on manual visual inspection. However, this method is generally inefficient and easily affected by personal subjective judgment, making it difficult to maintain a unified inspection standard in large-scale continuous production. In addition, the high-speed operation of the production line often exceeds the response capacity of manual inspection, which may lead to the omission of sealing defects, thereby affecting the quality of the product and even the reputation and credibility of the company. Therefore, an intelligent sealing quality detection solution applied to the essence automated production line is expected. Summary of the invention

[0004] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides an automated production line for essence liquid and its method, which installs a high-resolution camera on the production line to capture the sealing image of the essence liquid bottle in real time, and uses deep learning technology to analyze and process the sealing image, extracting fine-grained spatial structural features of the sealing image from it, so as to intelligently identify the quality defects of the sealing. In this way, automated and non-contact detection of the sealing of the essence liquid bottle can be achieved, reducing the outflow of defective products and ensuring the stability of product quality.

[0005] Accordingly, according to one aspect of the present application, there is provided an automated production line for serum, comprising:

[0006] A sealing status monitoring module is used to obtain the sealing image of the essence bottle collected by a high-definition camera;

[0007] A sealing quality feature extraction module, used for performing multi-scale sealing quality feature extraction on the essence bottle sealing image to obtain a multi-scale essence bottle sealing quality feature map;

[0008] A fine-grained spatial perception module, used for performing pixel-level global spatial context perception on the multi-scale essence bottle sealing quality feature map to obtain a global semantic association feature vector of the sealing quality;

[0009] The sealing quality detection module is used to determine whether the sealing quality is qualified based on the sealing quality global semantic association feature vector.

[0010] In the above-mentioned automated production line for essence, the sealing quality feature extraction module includes: an image preprocessing unit, which is used to perform image preprocessing on the sealing image of the essence bottle to obtain a detail-enhanced essence bottle sealing area image; and a multi-scale feature extraction unit, which is used to pass the detail-enhanced essence bottle sealing area image through a sealing quality feature extractor based on a bidirectional pyramid network to obtain the multi-scale essence bottle sealing quality feature map.

[0011] In the above-mentioned automated production line for essence, the image preprocessing unit includes: a sealing area target detection subunit, which is used to pass the sealing area image of the essence bottle through a sealing area detector based on the YOLO model to obtain the sealing area image of the essence bottle; and an image detail enhancement subunit, which is used to perform histogram equalization preprocessing and CLAHE color correction on the sealing area image of the essence bottle to obtain the detail-enhanced sealing area image of the essence bottle.

[0012] In the above-mentioned automated production line for essence, the fine-grained spatial perception module includes: a pixel-level feature discretization unit, which is used to perform pixel-level feature discretization processing on the multi-scale essence bottle sealing quality feature map to obtain a sequence of essence bottle sealing quality pixel-level feature vectors; a global semantic association encoding unit, which is used to perform global context semantic association encoding on the sequence of essence bottle sealing quality pixel-level feature vectors to obtain the sealing quality global semantic association feature vector.

[0013] In the above-mentioned automated production line for essence, the global semantic association encoding unit is used to: pass the sequence of pixel-level feature vectors of the sealing quality of the essence bottle through a global spatial context sensor based on a Transformer model to obtain the global semantic association feature vector of the sealing quality.

[0014] In the above-mentioned automated production line for essence, the sealing quality detection module includes: a feature optimization unit, used to perform backward scaling optimization on the global semantic association feature vector of sealing quality based on fine-grained error measurement to obtain an optimized global semantic association feature vector of sealing quality; and a detection result generation unit, used to pass the optimized global semantic association feature vector of sealing quality through a quality detector based on a classifier to obtain a detection result, and the detection result is used to indicate whether the sealing quality is qualified.

[0015] In the above-mentioned automated production line for essence, the feature optimization unit includes: a probabilistic activation subunit, which is used to perform probabilistic activation on the global semantic association feature vector of sealing quality based on the Sigmoid function to obtain a probabilistic global semantic association feature vector of sealing quality; a pre-classification subunit, which is used to input the global semantic association feature vector of sealing quality into a classifier to obtain an initial class label probability value; an error measurement subunit, which is used to calculate the absolute value of the difference between the feature value of each position in the probabilistic global semantic association feature vector of sealing quality and the initial class label probability value to obtain a backward machine learning fine-grained error measurement vector; an attention optimization subunit, which is used to use the backward machine learning fine-grained error measurement vector as a key The key vector, the query vector and the value vector are input into a converter structure to obtain a backward learning-like attention-optimized global semantic association feature vector for sealing quality; an optimization scaling factor calculation subunit is used to calculate the ratio between the sum of the backward machine learning fine-grained error metric vector and the length of the backward machine learning fine-grained error metric vector, and calculate the natural exponential function value with the ratio as the power as the optimization scaling factor; a feature optimization scaling subunit is used to calculate the product between the optimization scaling factor and the backward learning-like attention-optimized global semantic association feature vector for sealing quality to obtain the optimized global semantic association feature vector for sealing quality.

[0016] According to another aspect of the present application, there is provided a method for automated production of an essence, comprising:

[0017] Acquire the seal image of the essence bottle captured by a high-definition camera;

[0018] Performing multi-scale sealing quality feature extraction on the essence bottle sealing image to obtain a multi-scale essence bottle sealing quality feature map;

[0019] Performing pixel-level global spatial context perception on the multi-scale essence bottle sealing quality feature map to obtain a global semantic association feature vector of the sealing quality;

[0020] Based on the sealing quality global semantic association feature vector, it is determined whether the sealing quality is qualified.

[0021] Compared with the prior art, the automated production line and method for essence liquid provided by the present application installs a high-resolution camera on the production line to take the sealing image of the essence liquid bottle in real time, and uses deep learning technology to analyze and process the sealing image, thereby extracting fine-grained spatial structural features of the sealing image, so as to intelligently identify the quality defects of the sealing. In this way, automated and non-contact detection of the sealing of essence liquid bottles can be achieved, reducing the outflow of defective products and ensuring the stability of product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] 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.

[0023] Figure 1 4 is a block diagram of an automated production line for essence according to an embodiment of the present application.

[0024] Figure 2 Schematic diagram of the structure of the automated production line of essence according to an embodiment of the present application.

[0025] Figure 3 4 is a block diagram of a sealing quality feature extraction module in an automated production line for essence according to an embodiment of the present application.

[0026] Figure 4 4 is a block diagram of an image preprocessing unit in an automated production line for essence according to an embodiment of the present application.

[0027] Figure 5 It is a block diagram of a fine-grained spatial perception module in an automated production line of essence according to an embodiment of the present application.

[0028] Figure 6 4 is a block diagram of a sealing quality detection module in an automated production line for essence according to an embodiment of the present application.

[0029] Figure 7 Flow chart of an automated production method of essence according to an embodiment of the present application. DETAILED DESCRIPTION

[0030] Below, the embodiments of the present application will be described in more detail in conjunction with the accompanying drawings, and the above and other purposes, features and advantages of the present application will become more apparent. 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 example embodiments described herein.

[0031] Figure 1 4 is a block diagram of an automated production line for essence according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the structure of the automated production line of the essence according to the embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the automated production line 100 for essence liquid includes: a sealing state monitoring module 110, which is used to obtain an essence liquid bottle sealing image collected by a high-definition camera; a sealing quality feature extraction module 120, which is used to perform multi-scale sealing quality feature extraction on the essence liquid bottle sealing image to obtain a multi-scale essence liquid bottle sealing quality feature map; a fine-grained spatial perception module 130, which is used to perform pixel-level global spatial context perception on the multi-scale essence liquid bottle sealing quality feature map to obtain a global semantic association feature vector of sealing quality; and a sealing quality detection module 140, which is used to determine whether the sealing quality is qualified based on the global semantic association feature vector of sealing quality.

[0032] As mentioned in the above background technology, in the existing essence production process, the detection of sealing quality mainly relies on manual visual inspection. However, this method generally has the problem of low efficiency and is easily affected by personal subjective judgment. It is difficult to maintain a unified inspection standard in large-scale continuous production. In addition, the high-speed operation speed of the production line often exceeds the response capacity of manual inspection, which may lead to the omission of sealing defects, thereby affecting the quality of the product and even affecting the reputation and credibility of the company. In response to the above technical problems, the technical concept of this application is to install a high-resolution camera on the production line, take real-time sealing images of essence bottles, and use deep learning technology to analyze and process the sealing images, extract fine-grained sealing image spatial structure features from them, so as to intelligently identify the quality defects of the sealing. In this way, automated and contactless detection of the essence bottle seals can be achieved, reducing the outflow of defective products and ensuring the stability of product quality.

[0033] In the above-mentioned essence automated production line 100, the sealing state monitoring module 110 is used to obtain the essence bottle sealing image captured by the high-definition camera. It should be understood that the high-definition camera has a higher resolution and better image capture capability, and can clearly present the fine structure of the essence bottle sealing area, such as the flatness and uniformity of the seal, and the presence or absence of leakage marks, cracks or foreign matter. In the technical solution of the present application, the essence bottle sealing image captured by the camera is used as the input data for the sealing quality detection, which can provide accurate visual information for the evaluation of the sealing quality, so as to accurately identify the quality defects of the seal and realize the automated detection of the sealing quality.

[0034] In the above-mentioned essence liquid automated production line 100, the sealing quality feature extraction module 120 is used to extract multi-scale sealing quality features from the essence liquid bottle sealing image to obtain a multi-scale essence liquid bottle sealing quality feature map. Figure 3 FIG. 1 is a block diagram of a sealing quality feature extraction module in an automated production line of an essence according to an embodiment of the present application. Figure 3As shown, the sealing quality feature extraction module 120 includes: an image preprocessing unit 121, which is used to perform image preprocessing on the essence bottle sealing image to obtain a detail-enhanced essence bottle sealing area image; a multi-scale feature extraction unit 122, which is used to pass the detail-enhanced essence bottle sealing area image through a sealing quality feature extractor based on a bidirectional pyramid network to obtain the multi-scale essence bottle sealing quality feature map.

[0035] Specifically, the image preprocessing unit 121 is used to perform image preprocessing on the essence bottle seal image to obtain a detail-enhanced essence bottle seal area image. Figure 4 FIG. 1 is a block diagram of an image preprocessing unit in an automated production line of an essence according to an embodiment of the present application. Figure 4 As shown, the image preprocessing unit 121 includes: a sealing area target detection subunit 1211, which is used to pass the essence bottle sealing area image through a sealing area detector based on the YOLO model to obtain an essence bottle sealing area image; an image detail enhancement subunit 1212, which is used to perform histogram equalization preprocessing and CLAHE color correction on the essence bottle sealing area image to obtain the detail enhanced essence bottle sealing area image.

[0036] Specifically, the sealing area target detection subunit 1211 is used to pass the essence bottle sealing image through a sealing area detector based on the YOLO model to obtain the essence bottle sealing area image. It should be understood that, considering that in the essence bottle sealing image, the sealing area may occupy a smaller part of the image, directly extracting image features from the essence bottle sealing image may result in the loss of detail information and introduce a large amount of background noise interference. Therefore, in the technical solution of the present application, in order to improve the accuracy of sealing area feature extraction, a target detection model based on YOLO (You On ly LookOnce) is used to locate the sealing area in the essence bottle sealing image. Among them, the YOLO model is a real-time target detection algorithm that can quickly and accurately identify the sealing area in the image and effectively separate the sealing area from the background, thereby providing more accurate image input for subsequent sealing quality feature extraction.

[0037] Specifically, the image detail enhancement subunit 1212 is used to perform histogram equalization preprocessing and CLAHE color correction on the essence bottle sealing area image to obtain the detail enhanced essence bottle sealing area image. It should be understood that, considering that in the process of collecting the essence bottle sealing image, the essence bottle sealing area image may have low contrast and unclear local details due to uneven lighting, shadows or limitations of the camera sensor. Therefore, in the technical solution of the present application, a histogram equalization preprocessing method is first used to adjust the pixel distribution of the essence bottle sealing area image, thereby improving the overall contrast of the image, so that the fine structure in the image can be better displayed. Then, the image is locally enhanced by CLAHE (Contrast Limited Adaptive Histogram Equa li zat ion, i.e., adaptive histogram equalization with limited contrast) technology. On the basis of histogram equalization, CLAHE technology adds control of local contrast, which can effectively improve the visibility of local details of the image, suppress excessive local contrast, and prevent overexposure of the image due to excessive enhancement.

[0038] Specifically, the multi-scale feature extraction unit 122 is used to pass the detail-enhanced essence bottle sealing area image through a sealing quality feature extractor based on a bidirectional pyramid network to obtain the multi-scale essence bottle sealing quality feature map. It should be understood that since the detail features of the sealing area may exist in different scales, such as tiny cracks or foreign matter, uneven surfaces or overall deformation, etc., the feature extraction of a single scale may not be able to capture comprehensive sealing quality feature information. To this end, in the technical solution of the present application, a sealing quality feature extractor based on a bidirectional pyramid network is used to process the detail-enhanced essence bottle sealing area image. Specifically, the bidirectional pyramid network is composed of a plurality of feature layers of different resolutions, which can extract multi-scale sealing quality features from coarse to fine, from global to local, and ensure good recognition capabilities for sealing defects of various scales, thereby improving the comprehensiveness and accuracy of sealing quality detection. In the bidirectional pyramid network, the detail-enhanced essence bottle sealing area image is first subjected to rough feature extraction at a low-resolution layer to obtain global structural information of the seal. Then, through upsampling and downsampling, the image features are transferred between layers of different resolutions, gradually refining the feature extraction while maintaining the accuracy of the spatial position information, as well as the possible cracks and foreign body features.

[0039] In the above-mentioned automated production line 100 for essence, the fine-grained spatial perception module 130 is used to perform pixel-level global spatial context perception on the multi-scale essence bottle sealing quality feature map to obtain a global semantic association feature vector for sealing quality. Figure 5 FIG. 1 is a block diagram of a fine-grained spatial perception module in an automated production line of an essence according to an embodiment of the present application. Figure 5 As shown, the fine-grained spatial perception module 130 includes: a pixel-level feature discretization unit 131, which is used to perform pixel-level feature discretization processing on the multi-scale essence bottle sealing quality feature map to obtain a sequence of essence bottle sealing quality pixel-level feature vectors; a global semantic association encoding unit 132, which is used to perform global context semantic association encoding on the sequence of essence bottle sealing quality pixel-level feature vectors to obtain the sealing quality global semantic association feature vector.

[0040] Specifically, the pixel-level feature dispersion unit 131 is used to perform pixel-level feature dispersion processing on the multi-scale essence bottle sealing quality feature map to obtain a sequence of essence bottle sealing quality pixel-level feature vectors. It should be understood that in the essence bottle sealing area image, each pixel point contains rich sealing quality information. In order to be able to understand the slight changes in the sealing area more finely, in the technical solution of the present application, the multi-scale essence bottle sealing quality feature map is further decomposed in the spatial dimension through pixel-level feature dispersion processing, and the multi-channel information of each pixel point is integrated to form a pixel-level cross-channel representation vector, so that the feature information of the multi-scale essence bottle sealing quality feature map is refined to the pixel level, which is conducive to the subsequent identification of slight defects and provides an effective data basis for the refined analysis of sealing quality.

[0041] Specifically, the global semantic association encoding unit 132 is used to perform global context semantic association encoding on the sequence of the essence bottle sealing quality pixel-level feature vectors to obtain the sealing quality global semantic association feature vector. In a specific example of the present application, the processing method of performing global context semantic association encoding on the sequence of the essence bottle sealing quality pixel-level feature vectors is to pass the sequence of the essence bottle sealing quality pixel-level feature vectors through a global spatial context perceptor based on a Transformer model to obtain the sealing quality global semantic association feature vector. It should be understood that in the analysis process of the essence bottle sealing quality feature, the essence bottle sealing quality pixel-level feature vector reflects the pixel-level sealing quality feature information, and there is a spatial correlation between each essence bottle sealing quality pixel-level feature vector, for example, the features of adjacent pixels jointly describe a specific sealing defect. Therefore, in the technical solution of the present application, the sequence of the essence bottle sealing quality pixel-level feature vector is further globally context semantically associated encoded through a Transformer model to capture the spatial dependency between each pixel in the essence bottle sealing area image and enhance the context continuity of the feature representation. Specifically, the Transformer model uses the self-attention mechanism to effectively process long-distance dependencies, and has significant advantages in identifying cross-pixel sealing defects, such as continuous cracks, foreign body distribution, etc. Through such processing, local pixel-level information can be integrated into the global semantic framework, improving the accuracy and robustness of sealing quality anomaly detection.

[0042] In the above-mentioned automated production line 100 for essence, the sealing quality detection module 140 is used to determine whether the sealing quality is qualified based on the global semantic association feature vector of the sealing quality. Figure 6 FIG. 1 is a block diagram of a sealing quality detection module in an automated production line of an essence according to an embodiment of the present application. Figure 6 As shown, the sealing quality detection module 140 includes: a feature optimization unit 141, used to perform backward scaling optimization on the global semantic association feature vector of sealing quality based on fine-grained error measurement to obtain an optimized global semantic association feature vector of sealing quality; a detection result generation unit 142, used to pass the optimized global semantic association feature vector of sealing quality through a quality detector based on a classifier to obtain a detection result, and the detection result is used to indicate whether the sealing quality is qualified.

[0043] Specifically, the feature optimization unit 141 is used to perform backward scaling optimization based on a fine-grained error metric on the global semantic association feature vector of the sealing quality to obtain an optimized global semantic association feature vector of the sealing quality. In particular, in the technical solution of the present application, it is considered that the sealing area detector based on the YOLO model may not be able to detect the sealing area 100% accurately, especially when the sealing shape is irregular or there is occlusion. Moreover, the preprocessing steps of histogram equalization and CLAHE color correction may not be able to fully adapt to all types of sealing images, and may introduce some unexpected distortion or noise. At the same time, the sealing quality feature extractor based on the bidirectional pyramid network may not be able to capture all features related to the sealing quality, especially when the sealing defects are very small or hidden. Pixel-level feature dispersion processing may not be able to retain all important spatial information, resulting in some local features being ignored in the global feature vector. The global spatial context perceptron based on the Transformer model may not be able to fully and accurately capture the global semantic information of the sealing area. Due to the above reasons, the global semantic association feature vector of the sealing quality may have a class harmony offset relative to the predetermined class label of the classifier. Therefore, in order to reduce the category harmony deviation of the global semantic association feature vector of sealing quality, in the technical solution of the present application, the global semantic association feature vector of sealing quality is further subjected to backward scaling optimization based on fine-grained error measurement.

[0044] Specifically, the feature optimization unit 141 includes: a probabilistic activation subunit, which is used to perform probabilistic activation on the global semantic association feature vector of sealing quality based on a sigmoid function to obtain a probabilistic global semantic association feature vector of sealing quality; a pre-classification subunit, which is used to input the global semantic association feature vector of sealing quality into a classifier to obtain an initial class label probability value; an error measurement subunit, which is used to calculate the absolute value of the difference between the feature value of each position in the probabilistic global semantic association feature vector of sealing quality and the initial class label probability value to obtain a backward machine learning fine-grained error measurement vector; an attention optimization subunit, which is used to use the backward machine learning fine-grained error measurement vector as a key vector, the global semantic association feature vector of sealing quality as a query vector and a value vector, The key vector, the query vector and the value vector are input into the converter structure to obtain the backward learning type attention optimized sealing quality global semantic association feature vector; the optimization scaling factor calculation subunit is used to calculate the ratio between the sum of the backward machine learning fine-grained error metric vector and the length of the backward machine learning fine-grained error metric vector, and calculate the natural exponential function value with the ratio as the power as the optimization scaling factor; the feature optimization scaling subunit is used to calculate the product between the optimization scaling factor and the backward learning type attention optimized sealing quality global semantic association feature vector to obtain the optimized sealing quality global semantic association feature vector.

[0045] That is, the reverse interpretability analysis based on backward learning sensitive preference is performed on the global semantic association feature vector of sealing quality by using the following feature optimization formula to obtain the optimized global semantic association feature vector of sealing quality, wherein the feature optimization formula is:

[0046]

[0047]

[0048] Among them, V c represents the probabilistic global semantic association feature vector of sealing quality obtained after probabilistic activation of the global semantic association feature vector of sealing quality, P represents the probability value of the initial class label, Indicates subtraction by position, V e represents the fine-grained error metric vector for backward machine learning, v i represents the eigenvalue of the ith position of the backward machine learning fine-grained error metric vector, L represents the length of the backward machine learning fine-grained error metric vector, (·) T represents the transpose of a vector, represents vector multiplication, Softmax represents the normalized exponential function, ⊙ represents point multiplication by position, V' cThe global semantic association feature vector representing the optimized sealing quality.

[0049] Based on this, in the technical solution of the present application, the global semantic association feature vector of sealing quality is subjected to backward scaling optimization based on fine-grained error metrics, which first implements probabilistic activation based on Sigmoid function on the global semantic association feature vector of sealing quality to obtain a probabilistic global semantic association feature vector of sealing quality, and inputs the global semantic association feature vector of sealing quality into the classifier to obtain the initial class label probability value. Then, the error between the probabilistic global semantic association feature vector of sealing quality and the initial class label probability value is calculated by Backward Error Analysis to generate a fine-grained error metric vector, which quantifies the viewpoint metric contribution of each feature position to the classification error.

[0050] Then, the backward machine learning fine-grained error metric vector is used as the key vector, and the sealing quality global semantic association feature vector is used as the query vector and the value vector, and the key vector, the query vector and the value vector are input into the transformer structure to perform error metric-based attention-like backward machine learning on the sealing quality global semantic association feature vector through a Transformer-like architecture. In this way, a backward interpretability analysis (Backward Interpretability Analysis) is performed on the sealing quality global semantic association feature vector so that the obtained backward learning-like attention-optimized sealing quality global semantic association feature vector has a backward learning sensitive preference relative to the predetermined class label of the classifier. And further calculate the ratio of the sum of the error metric vector and the vector length, and use this ratio as the exponential to calculate the natural exponential function value to obtain the optimized scaling factor to perform reverse scaling optimization (Backward Optimization) on the backward learning class attention optimized sealing quality global semantic association feature vector. In this way, the sensitivity preference and robustness of the sealing quality global semantic association feature vector relative to the predetermined class label of the classifier can be effectively improved, thereby improving the accuracy of the classification result.

[0051] Specifically, the detection result generating unit 142 is used to pass the optimized sealing quality global semantic association feature vector through a quality detector based on a classifier to obtain a detection result, and the detection result is used to indicate whether the sealing quality is qualified. That is, the optimized sealing quality global semantic association feature vector is used as input, and a trained classifier model is used as a quality detector to perform feature analysis on it, and the optimized sealing quality global semantic association feature vector is classified and mapped in combination with the predetermined class label classification knowledge learned by the classifier during the training process to determine the corresponding sealing quality category. Under the action of the classifier, the optimized sealing quality global semantic association feature vector is converted into a specific detection result, for example, a binary label, indicating whether the sealing quality is qualified. If the detection result indicates that the sealing quality is qualified, it can be considered that the sealing process meets the preset standard; on the contrary, if the detection result indicates that the sealing quality is unqualified, it may be necessary to adjust the sealing equipment or rework the sealed product to ensure the quality and safety of the product.

[0052] In summary, the automated production line of essence according to the embodiment of the present application is explained, which installs a high-resolution camera on the production line to take real-time images of the seals of essence bottles, and uses deep learning technology to analyze and process the seal images, extracting fine-grained spatial structural features of the seal images, so as to intelligently identify quality defects of the seals. In this way, automated and non-contact detection of the seals of essence bottles can be achieved, reducing the outflow of defective products and ensuring the stability of product quality.

[0053] Figure 7 Flow chart of the automated production method of the essence according to the embodiment of the present application. Figure 7 As shown, the automated production method of essence according to the embodiment of the present application includes the following steps: S110, acquiring an essence bottle sealing image captured by a high-definition camera; S120, performing multi-scale sealing quality feature extraction on the essence bottle sealing image to obtain a multi-scale essence bottle sealing quality feature map; S130, performing pixel-level global spatial context perception on the multi-scale essence bottle sealing quality feature map to obtain a global semantic association feature vector of sealing quality; S140, determining whether the sealing quality is qualified based on the global semantic association feature vector of sealing quality.

[0054] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned automated production method of the essence have been described in detail above. Figures 1 to 6 The description of the essence automated production line has been described in detail, and therefore, its repeated description will be omitted.

[0055] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0056] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0057] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0058] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0059] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0060] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0061] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. An automated production line for essence, characterized in that: include: A sealing status monitoring module is used to obtain the sealing image of the essence bottle collected by a high-definition camera; A sealing quality feature extraction module, used for performing multi-scale sealing quality feature extraction on the essence bottle sealing image to obtain a multi-scale essence bottle sealing quality feature map; A fine-grained spatial perception module, used for performing pixel-level global spatial context perception on the multi-scale essence bottle sealing quality feature map to obtain a global semantic association feature vector of the sealing quality; The sealing quality detection module is used to determine whether the sealing quality is qualified based on the sealing quality global semantic association feature vector.

2. The automated production line for essence according to claim 1, characterized in that: The sealing quality feature extraction module comprises: An image preprocessing unit, used for performing image preprocessing on the essence bottle sealing image to obtain a detail-enhanced essence bottle sealing area image; The multi-scale feature extraction unit is used to pass the detail-enhanced essence bottle sealing area image through a sealing quality feature extractor based on a bidirectional pyramid network to obtain the multi-scale essence bottle sealing quality feature map.

3. The automated production line for essence according to claim 2, characterized in that: The image preprocessing unit comprises: A sealing area target detection subunit, used for passing the essence bottle sealing image through a sealing area detector based on a YOLO model to obtain an essence bottle sealing area image; The image detail enhancement subunit is used to perform histogram equalization preprocessing and CLAHE color correction on the image of the seal area of ​​the essence bottle to obtain the detail enhanced essence bottle seal area image.

4. The automated production line for essence according to claim 3, characterized in that: The fine-grained spatial perception module includes: A pixel-level feature dispersion unit, used for performing pixel-level feature dispersion processing on the multi-scale essence bottle sealing quality feature map to obtain a sequence of essence bottle sealing quality pixel-level feature vectors; The global semantic association coding unit is used to perform global context semantic association coding on the sequence of the essence bottle sealing quality pixel-level feature vectors to obtain the sealing quality global semantic association feature vector.

5. The automated production line for essence according to claim 4, characterized in that: The global semantic association encoding unit is used to: The sequence of pixel-level feature vectors of the sealing quality of the essence bottle is passed through a global spatial context sensor based on a Transformer model to obtain the global semantic association feature vector of the sealing quality.

6. The automated production line for essence according to claim 5, characterized in that: The sealing quality detection module comprises: A feature optimization unit, configured to perform backward scaling optimization on the sealing quality global semantic association feature vector based on a fine-grained error metric to obtain an optimized sealing quality global semantic association feature vector; The detection result generating unit is used for passing the optimized sealing quality global semantic association feature vector through a quality detector based on a classifier to obtain a detection result, wherein the detection result is used to indicate whether the sealing quality is qualified.

7. The automated production line for essence according to claim 6, characterized in that: The feature optimization unit comprises: A probabilistic activation subunit, used for performing probabilistic activation on the sealing quality global semantic association feature vector based on a Sigmoid function to obtain a probabilistic sealing quality global semantic association feature vector; A pre-classification subunit, used for inputting the sealing quality global semantic association feature vector into a classifier to obtain an initial class label probability value; An error metric subunit, used to calculate the absolute value of the difference between the feature value of each position in the probabilistic global semantic association feature vector of sealing quality and the initial class label probability value to obtain a backward machine learning fine-grained error metric vector; an attention optimization subunit, configured to use the backward machine learning fine-grained error metric vector as a key vector, the sealing quality global semantic association feature vector as a query vector and a value vector, and input the key vector, the query vector and the value vector into a converter structure to obtain a backward learning class attention optimized sealing quality global semantic association feature vector; an optimization scaling factor calculation subunit, configured to calculate a ratio between a sum of the backward machine learning fine-grained error metric vector and a length of the backward machine learning fine-grained error metric vector, and to calculate a natural exponential function value raised to a power of the ratio as an optimization scaling factor; The feature optimization scaling subunit is used to calculate the product between the optimization scaling factor and the backward learning type attention optimized sealing quality global semantic association feature vector to obtain the optimized sealing quality global semantic association feature vector.

8. An automated production method for an essence, characterized in that: include: Acquire the seal image of the essence bottle captured by a high-definition camera; Performing multi-scale sealing quality feature extraction on the essence bottle sealing image to obtain a multi-scale essence bottle sealing quality feature map; Performing pixel-level global spatial context perception on the multi-scale essence bottle sealing quality feature map to obtain a global semantic association feature vector of the sealing quality; Based on the sealing quality global semantic association feature vector, it is determined whether the sealing quality is qualified.

9. The automated production method of essence according to claim 8, characterized in that: The multi-scale sealing quality feature extraction is performed on the essence bottle sealing image to obtain a multi-scale essence bottle sealing quality feature map, including: Performing image preprocessing on the essence bottle seal image to obtain a detail-enhanced essence bottle seal area image; The detail-enhanced essence bottle sealing area image is passed through a sealing quality feature extractor based on a bidirectional pyramid network to obtain the multi-scale essence bottle sealing quality feature map.

10. The automated production method of essence according to claim 9, characterized in that: Performing image preprocessing on the essence bottle seal image to obtain a detail-enhanced essence bottle seal area image, including: Passing the essence bottle sealing image through a sealing area detector based on a YOLO model to obtain an essence bottle sealing area image; The image of the seal area of ​​the essence bottle is subjected to histogram equalization preprocessing and CLAHE color correction to obtain the detail-enhanced image of the seal area of ​​the essence bottle.

Images