Mobile phone shell ink brightness defect detection method based on DeepLab model and mobile phone shell

Abstract

The present application relates to the technical fields of artificial intelligence, printing and the like, and provides a mobile phone shell ink brightness defect detection method based on a DeepLab model and a mobile phone shell. After the ink printed layer images reflecting different ink brightness defects in the mobile phone shell substrate are preprocessed by statistics, ink printed layer image data is obtained. The ink printed layer image data is input into the DeepLab model to train the DeepLab model to identify different ink brightness defects, and a real DeepLab model is obtained. In the printing production process of the mobile phone shell, the ink printed layer of the current mobile phone shell substrate is imaged by a diffuse reflection light source camera system to obtain the ink printed layer image of the current mobile phone shell substrate, and the ink printed layer image of the current mobile phone shell substrate is input into the real DeepLab model to identify the ink brightness defects of the ink printed layer of the current mobile phone shell substrate. Therefore, the detection accuracy and efficiency of the ink brightness of the ink printed layer of the substrate are improved, the detection cost is reduced, and comprehensive detection of the ink brightness of the ink printed layer is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical fields of artificial intelligence, computer vision, image recognition and printing, and particularly relates to a mobile phone shell ink brightness defect detection method based on a DeepLab model and a mobile phone shell. BACKGROUND

[0002] In the printing production process of the mobile phone shell, generally includes printing screen preparation, ink printing, standing, baking and film covering and other processes. After the printing screen is prepared, the configured ink is printed to the base color layer of the substrate (for example, the green film surface of the substrate) through the printing screen, and the ink printing layer is covered by the frosting film after standing and baking. In the process of printing the configured ink to the base color layer of the substrate through the printing screen, it is necessary to detect whether the brightness of the ink printing layer of the substrate meets the requirements. In the prior art, in order to detect the brightness of the ink printing layer of the substrate, there are mainly visual detection, brightness meter or gloss meter detection, spectrophotometer detection and other detection technologies. The detection result of visual detection is easily affected by the experience, vision and environmental light change of the operator, the detection precision is low, the detection efficiency is low, the brightness meter or gloss meter detection can only detect specific points or areas, and it is difficult to fully cover the entire printing layer, and the spectrophotometer detection needs expensive equipment, and the operation is complex, and the detection cost is high.

[0003] In summary, in the printing production process of the existing mobile phone shell, the detection technology of the ink brightness of the ink printing layer of the substrate has the technical problems of low detection precision, low detection efficiency, high detection cost, and incomplete detection. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a mobile phone shell ink brightness defect detection method based on a DeepLab model and a mobile phone shell, so as to improve the detection precision and efficiency of the ink brightness of the ink printing layer of the substrate, reduce the detection cost, and realize the comprehensive detection of the ink brightness of the ink printing layer.

[0005] In the first aspect, the present application provides a mobile phone shell ink brightness defect detection method based on a DeepLab model, comprising:

[0006] Statistical ink printing layer images reflecting different ink brightness defects in a mobile phone shell substrate, and pre-processing the ink printing layer images to obtain pre-processed ink printing layer image data;

[0007] Input the ink printing layer image data into the DeepLab model to train the DeepLab model to identify the ability of the different ink brightness defects, and obtain a real DeepLab model capable of identifying the ink brightness defects of the ink printing layer of the mobile phone shell substrate;

[0008] In the printing production process of the mobile phone shell, the ink printing layer of the current mobile phone shell substrate after standing and baking is imaged by a diffuse reflection light source camera system to obtain an ink printing layer image of the current mobile phone shell substrate, and the ink printing layer image of the current mobile phone shell substrate is input into the real DeepLab model to identify the ink brightness defects of the ink printing layer of the current mobile phone shell substrate.

[0009] In a second aspect, the present application provides a mobile phone shell which is detected by the above-mentioned mobile phone shell ink brightness defect detection method based on the DeepLab model in a printing process.

[0010] Compared with the prior art, the present application has the following beneficial effects:

[0011] The present application provides a mobile phone shell ink brightness defect detection method based on a DeepLab model and a mobile phone shell. By statistically reflecting different ink brightness defects of the ink printing layer image of the mobile phone shell substrate, and preprocessing the ink printing layer image to obtain preprocessed ink printing layer image data, the ink printing layer image data is input into the DeepLab model to train the DeepLab model to identify the ability of the different ink brightness defects, and a real DeepLab model capable of identifying the ink brightness defects of the ink printing layer of the mobile phone shell substrate is obtained. In the printing production process of the mobile phone shell, the ink printing layer of the current mobile phone shell substrate after standing and baking is imaged by a diffuse reflection light source camera system to obtain an ink printing layer image of the current mobile phone shell substrate, and the ink printing layer image of the current mobile phone shell substrate is input into the real DeepLab model to identify the ink brightness defects of the ink printing layer of the current mobile phone shell substrate, thereby improving the detection accuracy and efficiency of the ink brightness of the ink printing layer of the substrate, reducing the detection cost, and realizing the comprehensive detection of the ink brightness of the ink printing layer. BRIEF DESCRIPTION OF DRAWINGS

[0012] The accompanying drawings, which are included to provide a further understanding of the present application, form a part of the present application and illustrate the illustrative embodiments of the present application and the explanation thereof, and do not constitute improper limitations on the present application.

[0013] Figure 1 is a flowchart of the mobile phone shell ink brightness defect detection method based on the DeepLab model of the present application embodiment;

[0014] Figure 2 is a flowchart of the DeepLab model training and ink printing layer image identification in the present application embodiment. DETAILED DESCRIPTION

[0015] In order to make personnel in the technical field better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor should belong to the scope of protection of the present application.

[0016] Referring to Figure 1 , Figure 2 The embodiment of the present application provides a DeepLab model-based mobile phone shell ink brightness defect detection method and a mobile phone shell. The mobile phone shell is detected by using the DeepLab model-based mobile phone shell ink brightness defect detection method in the printing production process. The DeepLab model-based mobile phone shell ink brightness defect detection method comprises steps S101, S102 and S103. The ink printed layer images reflecting different ink brightness defects in the mobile phone shell substrate are counted, and the ink printed layer images are preprocessed to obtain preprocessed ink printed layer image data. The ink printed layer image data is input into a DeepLab model to train the DeepLab model to identify the ability of the different ink brightness defects, and a real DeepLab model capable of identifying the ink brightness defects of the ink printed layer of the mobile phone shell substrate is obtained. In the printing production process of the mobile phone shell, the ink printed layer of the current mobile phone shell substrate after standing and baking is imaged by a diffuse reflection light source camera system to obtain an ink printed layer image of the current mobile phone shell substrate, and the ink printed layer image of the current mobile phone shell substrate is input into the real DeepLab model to identify the ink brightness defects of the ink printed layer of the current mobile phone shell substrate, thereby improving the detection precision and efficiency of the ink brightness of the ink printed layer of the substrate, reducing the detection cost, and realizing the comprehensive detection of the ink brightness of the ink printed layer.

[0017] Step S101, count the ink print layer images reflecting different ink brightness defects in the mobile phone case substrate, and pre-process the ink print layer images to obtain pre-processed ink print layer image data. In step S101, the ink brightness defects existing in the mobile phone case substrate are comprehensively counted to ensure that the training data covers all possible defect types (such as insufficient brightness, excessive brightness, uneven brightness, etc.). It can be understood that in step S101, pre-processing the ink print layer images can ensure that the DeepLab model has high-quality input data, so that it can fully learn the characteristics of ink brightness defects and improve the accuracy of defect recognition. Preferably, when pre-processing the ink print layer images, the image size and brightness range of the ink print layer images can be unified to reduce the interference of data noise on model training. Preferably, when pre-processing the ink print layer images, the data set of the ink print layer images can be expanded by rotating the ink print layer images, cropping the ink print layer images, adjusting the brightness of the ink print layer images, etc., thereby increasing the generalization ability of the model.

[0018] Step S102, input the ink printing layer image data into the DeepLab model to train the ability of the DeepLab model to identify different ink brightness defects, and obtain a real DeepLab model capable of identifying ink brightness defects of the ink printing layer of the mobile phone shell substrate. It should be noted that when training the DeepLab model through the ink printing layer image data, the pre-processed ink printing layer image data and the corresponding pixel-level annotation label can be input into the DeepLab model, and each ink printing layer image corresponds to a pixel-level annotation image, wherein each pixel is marked as "normal" or a specific brightness defect category. The DeepLab model can use DeepLabv3+ as the basic architecture, which has good edge feature capture ability and multi-scale semantic information extraction ability. In addition, the DeepLab model can select ResNet or Xception as the backbone for extracting high-level image features. In addition, the output categories of the DeepLab model classification task can be defined according to the types of ink brightness defects (such as normal, bright spots, brightness unevenness, etc.). In addition, a cross-entropy loss function can be set to evaluate the gap between the model prediction and the actual label, and further add class weight balancing to deal with the problem of insufficient defect class samples in the data set. During specific model training, image data can be processed by the model, features can be extracted layer by layer and pixel classification prediction output can be generated, and dilated convolution (Dilated Convolution) can be used to expand the receptive field to capture large-scale brightness differences. The model prediction result is compared with the annotation data, the cross-entropy loss value is calculated, the current state of the model is evaluated, the enhanced weight is used for small area defects to avoid being ignored by the model, and the model parameters (such as weight, bias) are adjusted according to the loss value through the back propagation algorithm. A certain batch size (batch size) is used to train the data for multiple rounds until the loss function converges. The model performance is evaluated on the validation set after each round (epoch). Independent validation data set is used to evaluate the performance of the model, and the following indicators are observed: pixel accuracy (Pixel Accuracy, PA), which is used to evaluate the overall classification accuracy; intersection over union (Intersecti on over Un i on, I oU), which is used to measure the segmentation accuracy of the model for the defect area; F1 score, which is used to evaluate the balanced performance of the model for each defect category. If the model performance is not good, adjust the training parameters (such as learning rate, optimizer, etc.) and retrain. Save the trained DeepLab model weight to form a "real DeepLab model" to ensure that the model has the ability of real-time detection of industrial production line and supports subsequent deployment. It can be understood that the DeepLab model based on dilated convolution and multi-scale feature fusion technology can accurately capture the complex morphology of the brightness defect area.In addition, the defects such as bright spots or dark spots usually account for a small proportion, and the context awareness capability of the DeepLab model enables it to effectively locate these defects. In addition, the ink brightness defects may have small regional features, and it is necessary to provide accurate training targets through pixel-level supervision to ensure that the model can learn the boundary and shape details of different brightness defects. In addition, ink brightness defect detection requires multi-class classification for each pixel, and cross-entropy loss can quantify the error of model prediction. In addition, due to the fact that normal regions usually account for the vast majority in the data set, cross-entropy loss can prevent the model from ignoring small proportions of defect categories through weight balancing strategy. In addition, through the verification link, it can be detected whether the model is over-fitted on the training set, and the training parameters are adjusted. In addition, the trained and verified DeepLab model can be directly deployed to the production line for real-time detection of ink brightness defects. The saved model can support incremental training and be updated at any time to adapt to new defect types or process conditions.

[0019] In step S103, during the printing production process of the mobile phone shell, the ink printing layer of the current mobile phone shell substrate after standing and baking is imaged by the diffuse reflection light source camera system to obtain the ink printing layer image of the current mobile phone shell substrate, and the ink printing layer image of the current mobile phone shell substrate is input into the real DeepLab model to identify the ink brightness defects of the ink printing layer of the current mobile phone shell substrate. It should be noted that the diffuse reflection light source in the diffuse reflection light source camera system can eliminate the shadows and highlight areas caused by the directionality of light, ensure that the captured image is clearer, and avoid the influence of specular reflection (such as highlight or light spot) on the imaging quality. In addition, diffuse reflection light can make the brightness distribution of the ink printing layer more natural, and enhance the contrast between the brightness defects such as bright spots and dark spots and the normal area. This high-quality imaging can provide good input data for the pixel-level segmentation of the DeepLab model. In addition, the mobile phone shell substrate may have various surface materials (such as smooth, matte or textured surface), and the diffuse reflection light can minimize the influence of material differences on image quality. It should be noted that the ink brightness may change after standing and baking, especially due to the influence of ink leveling, curing or volatilization process. Detecting after standing and baking can reflect the true state of the final product. It should be noted that the DeepLab model has been trained and can identify different types of brightness defects (such as bright spots, dark spots, uneven brightness, etc.), providing accurate defect classification and positioning. Inputting the current image into the DeepLab model can complete the detection in a short time in combination with the automation equipment on the production line, improve the detection efficiency and accuracy, cover the detection of various subtle defects, and make the detection of ink brightness defects more comprehensive.

[0020] In some preferred embodiments, the ink brightness defects of the ink printing layer of the current mobile phone case substrate include: insufficient brightness, excessive brightness, uneven brightness, bright light spots, edge brightness blur, local bright stripes, and dark spots. It should be noted that insufficient brightness refers to the overall brightness value of the ink printing layer being lower than the normal range of the design target, which is a surface dark ink brightness defect. Specifically, the characteristics of insufficient brightness include that the ink printing layer presents a globally low brightness, the ink printing layer has no obvious highlights or dark spots, the brightness of the ink printing layer changes smoothly, and the overall presents a uniform dullness. In addition, excessive brightness refers to the overall brightness value of the ink printing layer being higher than the normal range of the design target, which is too bright. Uneven brightness refers to the brightness value in a region of the ink printing layer presenting obvious fluctuations, which is a transition that is not uniform between different brightness. Specifically, the characteristics of uneven brightness include that the overall brightness of the ink printing layer is within the normal range, but there are brightness gradient differences in local regions. Uneven brightness is a range defect, which emphasizes the fluctuation of the brightness gradient, and is different from the overall brightness deviation of insufficient brightness and excessive brightness, and is clearly distinguished from the specific form of local dark spots or bright stripes. In addition, the bright light spot refers to the appearance of a local strong reflection region in the ink printing layer, with brightness significantly higher than the surrounding area, forming a highlight. The characteristics of the bright light spot mainly reflect the small range of strong light reflection, which is different from the uniform over-brightness of excessive brightness, and is different from the gradient change emphasized by uneven brightness. Edge brightness blur refers to the edge brightness of the printed pattern of the mobile phone case diffusing or transitioning unclearly, which destroys the design boundary clarity. The characteristics of edge brightness blur mainly reflect that the defect only occurs in the edge region of the pattern, which is different from the overall or local insufficient brightness. In addition, the local bright stripe refers to the appearance of a linear region of brightness abnormality in the ink printing layer. The characteristics of the local bright stripe mainly reflect that it presents a bright stripe with brightness significantly higher than the surrounding area, which is different from the local strong light of the bright light spot, and is different from the gradient fluctuation of uneven brightness. In addition, the dark spot refers to the appearance of a local dark region in the ink printing layer, with brightness significantly lower than the surrounding area. The characteristics of the dark spot mainly reflect that it presents a point or small range of dark regions, which is different from the overall insufficient brightness, and is different from the gradient change of uneven brightness.

[0021] In some preferred embodiments, when the ink printing layer image is preprocessed, it includes: unifying the image size of the ink printing layer image, and unifying the brightness range of the ink printing layer image. It should be noted that unifying the image size and brightness range can eliminate irrelevant differences (such as inconsistent size and brightness distribution) in the input data, and improve the generalization ability of the model to new data.

[0022] In some preferred embodiments, when the image of the ink printing layer is preprocessed, the data set of the image of the ink printing layer is expanded by rotating the image of the ink printing layer, cropping the image of the ink printing layer, and adjusting the brightness of the image of the ink printing layer. It should be noted that the data set of the image of the ink printing layer can be expanded by rotation, cropping, brightness adjustment and other methods, which can greatly improve the training effect and detection performance of the model. These methods can enhance the diversity and authenticity of the data from the aspects of spatial distribution, multi-angle adaptation, local feature extraction, and brightness variation, etc., to ensure that the model can stably and efficiently detect the brightness defects in various scenarios.

[0023] In some preferred embodiments, when the real DeepLab model identifies that the ink printing layer of the current mobile phone shell substrate has an ink brightness defect, the specific type of the ink brightness defect is transmitted to the production monitoring end. Further, the production monitoring end matches the cause of the ink brightness defect according to the specific type of the ink brightness defect, and provides corresponding feedback suggestions to the production supervisor after matching the cause of the ink brightness defect. Further, the production monitoring end matches the cause of the ink brightness defect according to the specific type of the ink brightness defect, including: presetting a mapping library of the defect type and the defect cause of the ink brightness; when receiving the specific type of the ink brightness defect, the specific type of the ink brightness defect is used as an index to automatically search in the mapping library of the defect type and the defect cause to match the cause of the ink brightness defect of the specific type. Further, after matching the cause of the ink brightness defect, corresponding feedback suggestions are provided to the production supervisor, including: different feedback suggestions are preset for different defect causes of the ink brightness; after matching the defect cause of the ink brightness defect type, the corresponding feedback suggestions are automatically associated to provide to the production supervisor. Further, after matching the defect cause of the ink brightness defect type, the corresponding feedback suggestions are automatically associated, including: after matching the defect cause of the ink brightness defect type, the adjustment suggestions for adjusting the related equipment are automatically associated, or after matching the defect cause of the ink brightness defect type, the optimization suggestions for optimizing the printing ink are automatically associated, or after matching the defect cause of the ink brightness defect type, the quality inspection suggestions for quality inspecting the mobile phone shell substrate are automatically associated.

[0024] It should be noted that when the ink brightness defect is identified, the defect type is transmitted to the production monitoring end, so that the production monitoring end can obtain the defect information in real time, providing basic data for subsequent processing. In addition, by automatically matching the defect reason, the problem diagnosis time can be shortened, and the delay of manual investigation can be avoided. Moreover, manual investigation may deviate due to lack of experience or subjective judgment errors, while automatic matching can reduce the occurrence of these problems. It should be noted that the mapping library of defect type and reason can be accumulated and optimized according to production experience, covering most common problems and providing reliable diagnostic basis. In addition, different feedback suggestions are preset for different defect reasons of the ink brightness, so that after the defect causing reason matched by the ink brightness defect type, the corresponding feedback suggestion is automatically associated, which is quickly given to the production person in charge for reference, effectively improving the efficiency of defect solving.

[0025] It should be noted that some ink brightness defects (such as uneven brightness, local bright stripes) are usually caused by equipment operation parameters (such as spraying path, nozzle pressure) or precision problems, and by automatically generating equipment adjustment suggestions, it is helpful to quickly locate and solve the equipment problems. In further some preferred embodiments, after the defect causing reason matched by the ink brightness defect type, the adjustment suggestion for adjusting the related equipment is automatically associated, including: after the defect causing reason matched by the brightness stripe is uneven spraying path, the adjustment suggestion for adjusting the spraying trajectory of the spraying equipment is automatically associated; after the defect causing reason matched by the insufficient brightness is insufficient spraying thickness, the adjustment suggestion for improving the spraying pressure or coating thickness of the spraying equipment is automatically associated; after the defect causing reason matched by the bright spot is uneven curing, the adjustment suggestion for improving the coating thickness of the spraying equipment or the temperature of the curing equipment or prolonging the curing time of the curing equipment is automatically associated. It should be noted that the feature of the brightness stripe is usually caused by uneven spraying path, and this defect directly reflects the operation deviation of the equipment on a specific trajectory. Adjusting the spraying trajectory can accurately eliminate this problem and avoid the interference caused by adjusting other irrelevant parameters. By optimizing the spraying trajectory, it can ensure that the equipment uniformly distributes the coating thickness on the entire path, fundamentally solving the brightness stripe problem. In addition, insufficient spraying pressure will result in too thin coating, and then the problem of insufficient brightness will occur. By improving the spraying pressure or increasing the coating thickness, this deficiency can be directly made up. Insufficient spraying thickness not only affects the brightness, but also may cause the protection performance of the coating to decrease. Adjusting the parameters of the spraying equipment can ensure the consistency of the coating quality and enhance the product performance. In addition, the bright spot is usually caused by insufficient curing of the local coating during the curing process or abnormal reflection of the thin coating. Increasing the coating thickness can reduce the reflection tendency, and optimizing the parameters of the curing equipment (such as temperature or time) can ensure the uniformity of curing.

[0026] It is also necessary to point out that some defects (such as insufficient brightness, excessive brightness) can be caused by the physical properties of the ink itself (such as the proportion of high-gloss ink components or dispersibility). By generating optimization suggestions, it is possible to help adjust the ink formula and improve its adaptability. In further preferred embodiments, according to the defect causes matched by the ink brightness defect type, optimization suggestions for optimizing the printing ink are automatically associated, including: according to the defect cause matched by the insufficient brightness as the proportion of high-gloss ink components being too low, an optimization suggestion for increasing the proportion of high-gloss ink components is automatically associated; according to the defect cause matched by the excessive brightness as the proportion of high-gloss ink components being too high, an optimization suggestion for reducing the proportion of high-gloss ink components is automatically associated; according to the defect cause matched by the uneven brightness as the proportion of ink dispersibility components being too low, an optimization suggestion for increasing the proportion of ink dispersibility components is automatically associated. It should be noted that the proportion of high-gloss ink components directly affects the gloss and reflection performance of the printed layer. If the proportion is too low, the problem of insufficient brightness of the ink will be more pronounced. Increasing the proportion of high-gloss ink components can increase the reflectivity of the ink, thereby improving the problem of insufficient brightness. Excessive proportion of high-gloss ink components can cause excessive brightness, thereby damaging the overall visual coordination. By reducing the proportion, the brightness can be adjusted to a reasonable range, thereby improving product consistency. Uneven brightness is usually caused by poor dispersion of pigment particles in the ink, resulting in local fluctuations in coating thickness or gloss. Increasing the proportion of dispersibility components can improve the distribution of pigment particles and improve the consistency of the coating, thereby eliminating the problem of uneven brightness.

[0027] It is also necessary to point out that some defects (such as dark spots, edge brightness blur) can be caused by quality problems of the phone case substrate itself, such as uneven surface roughness or contamination. Quality inspection suggestions can help locate and solve the substrate problem. In further preferred embodiments, according to the defect causes matched by the ink brightness defect type, quality inspection suggestions for quality inspection of the phone case substrate are automatically associated, including: according to the defect cause matched by the edge brightness blur as the deformation of the edge shape of the substrate, a quality inspection suggestion for checking the edge deformation of the phone case substrate is automatically associated; according to the defect cause matched by the dark spot as the surface contamination of the substrate, a quality inspection suggestion for checking the surface contamination of the phone case substrate is automatically associated. It should be noted that edge brightness blur is usually caused by optical deviation due to deformation of the edge shape of the substrate or uneven adhesion of the coating. Automatic checking of the deformation of the edge of the substrate can quickly locate this problem and avoid unnecessary production adjustments. Dark spots are usually caused by surface contamination of the substrate (such as oil, dust, particulate matter), and the contamination can affect the adhesion or light reflection properties of the ink. Checking the surface contamination of the substrate can help directly solve the problem of dark spots.

[0028] In further embodiments, when the real DeepLab model identifies the ink brightness defects of the ink printing layer of the current mobile phone shell base material, the priority order from high to low is: insufficient brightness, excessive brightness, uneven brightness, bright light spots, dark spots, local bright stripes, and edge brightness blur; when a higher priority ink brightness defect is identified, the ink brightness anomaly of the ink printing layer is prompted to the production monitoring end, and the identification of lower priority ink brightness defects is stopped. It should be noted that insufficient brightness is the most significant visual defect, which directly causes the mobile phone shell to appear dull and lackluster, affecting the overall appearance. The identification of insufficient brightness is holistic and can be quickly judged by global brightness distribution without complex local calculation. When insufficient brightness is identified, the real DeepLab model stops identifying other ink brightness defects with lower priority, thereby saving computing resources and improving detection efficiency. Excessive brightness can cause excessive brightness or dazzling visual effects, which can damage the overall product harmony and is an obvious defect affecting the appearance quality, but it is weaker than insufficient brightness in damaging the overall product harmony. The calculation complexity of excessive brightness is similar to that of insufficient brightness, and the detection of excessive brightness is based on the statistics of global brightness values, which has low computational complexity. Uneven brightness can cause obvious color difference or uneven gloss on the product surface, which can damage the visual unity. Detecting uneven brightness requires analyzing the local brightness change gradient, which has slightly higher computational complexity than global detection of insufficient brightness and excessive brightness. Bright light spots can cause local strong light reflection at a specific angle, affecting the appearance, especially in a strong light environment. Bright light spots are usually caused by uneven curing or thin coating, have a low occurrence frequency, and are localized in terms of impact. Moreover, detecting bright light spots requires analyzing the highlight values in a specific area of the image and comparing them with the surrounding area, which has higher computational complexity than brightness problems. Dark spots are local brightness defects that can appear unevenly, but have relatively small overall impact. Dark spots are usually caused by substrate contamination or spraying defects and usually affect a small range. Detecting dark spots requires identifying small-range low-brightness areas, which has slightly higher complexity than other brightness problems. Local bright stripes are usually linear brightness abnormalities that affect specific areas but have some destructive effects on the appearance. Detecting linear features requires combining brightness distribution and geometric characteristics, which has higher computational complexity. Edge brightness blur mainly affects the clarity of the pattern boundary and has relatively small overall impact on the appearance. Detecting edge brightness blur requires analyzing the brightness gradient and clarity of the pattern boundary, which has higher computational complexity.It can be understood that in this embodiment, by setting the priority order of the ink brightness defects identified by the realistic DeepLab model from high to low, the specific priority order from high to low is: insufficient brightness, excessive brightness, uneven brightness, reflective bright spots, dark spots, local bright stripes and blurred edge brightness. When a higher-priority ink brightness defect is identified, the production monitoring end is prompted that the ink brightness of the ink printing layer is abnormal, and the identification of lower-priority ink brightness defects is stopped, thereby improving detection efficiency and saving computing resources.

[0029] It should be pointed out that the above embodiments are only preferred specific implementation methods of the present invention, and the protection scope of the present invention is not limited thereto. Any changes or replacements that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. The protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for detecting the brightness defects of a mobile phone case ink based on a DeepLab model, characterized in that, The method comprises the following steps: statistical mobile phone shell substrate image reflecting different ink brightness defects of ink printing layer is obtained, and the ink printing layer image is preprocessed to obtain preprocessed ink printing layer image data; the ink printing layer image data is input into the DeepLab model to train the DeepLab model to identify the different ink brightness defects, and a real DeepLab model capable of identifying the ink brightness defects of the ink printing layer of the mobile phone shell substrate is obtained; in the printing production process of the mobile phone shell, the ink printing layer of the current mobile phone shell substrate after standing and baking is imaged by a diffuse reflection light source camera system to obtain the ink printing layer image of the current mobile phone shell substrate, and the ink printing layer image of the current mobile phone shell substrate is input into the real DeepLab model to identify the ink brightness defects of the ink printing layer of the current mobile phone shell substrate; the ink brightness defects of the ink printing layer of the current mobile phone shell substrate include insufficient brightness, excessive brightness, uneven brightness, bright spot, edge brightness blur, local bright stripe and dark spot; when the real DeepLab model identifies that the ink printing layer of the current mobile phone shell substrate has ink brightness defects, the specific type of the ink brightness defects is transmitted to the production monitoring end; the production monitoring end matches the cause of the ink brightness defects according to the specific type of the ink brightness defects, and provides corresponding feedback suggestions to the production person in charge after matching the cause of the ink brightness defects; after matching the cause of the ink brightness defects according to the type of the ink brightness defects, the adjustment suggestions for adjusting the related equipment are automatically associated, including: after matching the cause of the defect to uneven spraying path according to the brightness stripe, the adjustment suggestion for adjusting the spraying track of the spraying equipment is automatically associated; after matching the cause of the defect to insufficient spraying thickness according to the insufficient brightness, the adjustment suggestion for improving the spraying pressure or coating thickness of the spraying equipment is automatically associated; after matching the cause of the defect to uneven curing according to the bright spot, the adjustment suggestion for improving the coating thickness of the spraying equipment or the temperature of the curing equipment or prolonging the curing time of the curing equipment is automatically associated; after matching the cause of the defect to the low proportion of high light ink component according to the insufficient brightness, the optimization suggestion for improving the proportion of high light ink component is automatically associated; after matching the cause of the defect to the high proportion of high light ink component according to the excessive brightness, the optimization suggestion for reducing the proportion of high light ink component is automatically associated; after matching the cause of the defect to the low proportion of ink dispersibility component according to the uneven brightness, the optimization suggestion for improving the proportion of ink dispersibility component is automatically associated; after matching the cause of the defect to the deformation of the edge shape of the substrate according to the edge brightness blur, the quality inspection suggestion for checking the edge deformation of the mobile phone shell substrate is automatically associated; after matching the cause of the defect to the surface pollution of the substrate according to the dark spot, the quality inspection suggestion for checking the surface pollution of the mobile phone shell substrate is automatically associated.

2. The DeepLab model-based mobile phone case ink brightness defect detection method of claim 1, wherein, when the ink printing layer image is preprocessed, the image size of the ink printing layer image is unified, and the brightness range of the ink printing layer image is unified. 3.The DeepLab model-based mobile phone shell ink brightness defect detection method of claim 1, wherein, The pre-processing of the ink printing layer image comprises: rotating the ink printing layer image, cutting the ink printing layer image, and adjusting the brightness of the ink printing layer image to expand the data set of the ink printing layer image.

4. The DeepLab model-based mobile phone case ink brightness defect detection method of claim 1, wherein, The production monitoring terminal matches the cause of the ink brightness defect according to the specific type of the ink brightness defect, which comprises: presetting a mapping library of the defect type and the defect cause of the ink brightness; receiving the specific type of the ink brightness defect, taking the specific type of the ink brightness defect as an index, and automatically searching in the mapping library of the defect type and the defect cause to match the cause of the ink brightness defect of the specific type. 5.The DeepLab model-based mobile phone shell ink brightness defect detection method of claim 4, wherein, The feedback suggestion is provided to the production manager after matching the cause of the ink brightness defect, which comprises: presetting different feedback suggestions for different defect causes of the ink brightness; and automatically associating the corresponding feedback suggestion according to the cause of the ink brightness defect matched by the defect type to provide the feedback suggestion to the production manager.

6. The DeepLab model-based mobile phone case ink brightness defect detection method of claim 5, wherein, The feedback suggestion is automatically associated according to the cause of the ink brightness defect matched by the defect type, which comprises: automatically associating the adjustment suggestion of adjusting the related equipment according to the cause of the ink brightness defect matched by the defect type, or automatically associating the optimization suggestion of optimizing the printing ink according to the cause of the ink brightness defect matched by the defect type, or automatically associating the quality inspection suggestion of quality inspecting the mobile phone shell substrate according to the cause of the ink brightness defect matched by the defect type.

7. A phone case, characterized in that, The mobile phone shell is detected by using the DeepLab model-based mobile phone shell ink brightness defect detection method according to any one of claims 1-6 in the printing process.

Images