A method and system for quality supervision in injection mold production

Abstract

This invention belongs to the field of image processing technology, specifically relating to a method and system for quality supervision in injection mold production. The method includes: performing two-dimensional discrete wavelet transform on a standard image of a defect-free product and a target image corresponding to the product to be inspected, respectively, to obtain the energy of each sub-band; calculating the energy deviation characteristic value of the target image relative to the standard image; selecting the sub-band with the largest energy deviation characteristic value to construct a wavelet coefficient deviation map, dividing it into multiple regions according to the key physical location of the mold, and calculating the degradation concentration degree reflecting the degree of spatial concentration of degradation; analyzing the temporal change of the degradation concentration degree to determine the degradation trend degree; and normalizing and fusing the sensitive sub-band energy deviation, degradation concentration degree, and degradation trend degree to obtain a comprehensive early warning score for early warning. This invention can identify degradation trends before defects form a visible morphology, shifting the quality control checkpoint from post-screening to pre-intervention, thereby reducing the scrap rate and improving production stability.

Description

Technical Field

[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to a method and system for quality supervision in injection mold production. Background Technology

[0002] In the injection molding industry, mold condition has a decisive impact on product quality. As the production process continues, molds inevitably experience progressive wear, leading to gradual defects on the product surface such as changes in microtexture, increased roughness, or decreased gloss. These defects differ from obvious structural defects; their development is slow and insidious, often difficult to detect with the naked eye or conventional inspection methods in their early stages. However, as production volumes accumulate, they can eventually lead to batch-wide quality problems.

[0003] Currently, the industry mainly relies on manual sampling or automated vision inspection systems for quality checks of finished products. Manual sampling relies on subjective experience and judgment, which is not only inefficient but also unable to capture microscopic surface changes. While automated vision inspection systems enable batch screening, their core logic still focuses on the macroscopic appearance characteristics of the final product, lacking the ability to sensitively capture microscopic changes on the product surface. They cannot identify minor deterioration of molds in the early stages and can only perform post-screening after defects have become visible. This post-screening model has a serious lag; usually, by the time the inspection system detects obvious defects, hundreds or even thousands of defective products have already been produced, seriously affecting product quality and production efficiency.

[0004] Therefore, there is an urgent need for a method to supervise the production quality of injection molds, so as to achieve real-time monitoring and early warning of the mold status during the product production process. Summary of the Invention

[0005] To address the technical problem that existing detection technologies lack the ability to sensitively detect early-stage minor deterioration of molds, and cannot provide early warnings before gradual defects become visible, resulting in batches of defective products that can only be screened after defects have become obvious, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for quality supervision in injection mold production, comprising:

[0007] The most representative standard image is selected from images of defect-free, qualified products. Images of the product to be inspected are acquired in real-time as the target image. Two-dimensional discrete wavelet transforms are performed on both the standard and target images to obtain the energy of each sub-band at each level. Based on the difference between the energy of each sub-band at each level of the target image and the corresponding sub-band at the corresponding level of the standard image, the energy deviation characteristic value of each sub-band in the target image is determined. Sensitive sub-bands are selected based on the magnitude of the energy deviation characteristic value. A wavelet coefficient deviation map is constructed based on the difference between the target and standard images in the sensitive sub-bands. This map is divided into multiple key regions, and the degradation concentration is determined based on the wavelet coefficient deviation in the key regions. The degradation trend is determined based on the temporal change of the degradation concentration. A comprehensive warning score is obtained based on the energy deviation characteristic value of the target image in the sensitive sub-bands, the degradation concentration of the target image, and the degradation trend. A warning is issued based on the magnitude of the comprehensive warning score.

[0008] This invention selects the most representative standard image from defect-free, qualified product images as a benchmark, and performs two-dimensional discrete wavelet transform on the standard and target images to obtain the energy of each sub-band. This allows for precise quantification of the texture and gloss characteristics of the product surface at different scales and directions. By calculating the energy deviation characteristic values ​​between the target and standard images on corresponding sub-bands, this invention can convert microscopic changes in the product surface (such as increased roughness and decreased gloss) into measurable numerical signals, thereby capturing early signs of degradation that are difficult to detect with the naked eye. Furthermore, this invention filters sensitive sub-bands based on energy deviation characteristic values ​​and constructs a wavelet coefficient deviation map, dividing the wavelet coefficient deviation map into multiple sub-bands corresponding to key physical locations of the mold. By analyzing key areas and calculating the degradation concentration, this invention can effectively distinguish between systemic local degradation and global random fluctuations, thus avoiding misjudgment. Furthermore, by analyzing the temporal changes in degradation concentration to determine degradation trend, this invention integrates the energy deviation characteristic value of the target image in the sensitive sub-band, degradation concentration, and degradation trend to obtain a comprehensive early warning score. This achieves a three-dimensional degradation characterization of "instantaneous-space-time," enabling the system to provide early, sensitive, and reliable warnings for gradual degradation. This avoids the lag problem of traditional detection methods that can only identify visible defects, shifting the quality control checkpoint from post-screening to pre-intervention, significantly reducing scrap rates and improving the stability of the production process.

[0009] Preferably, the method further includes: preprocessing the standard image and the target image respectively, wherein the preprocessing includes image registration and illumination normalization, the image registration adopts the SIFT feature matching algorithm combined with the RANSAC algorithm, and the illumination normalization adopts histogram equalization.

[0010] Preferably, the energy deviation characteristic value satisfies the expression: In the formula, The first part represents the target image. The first level Energy deviation characteristic value of each sub-band; The first part represents the target image. The first level The energy carried by a person; The first standard image The first level The energy carried by a person; This is a hyperparameter used to prevent the denominator from being zero.

[0011] This invention transforms invisible surface degradation into a measurable numerical signal by quantifying the relative changes in wavelet domain energy, thereby enabling early warning of mold conditions and avoiding the lag of relying solely on the final product appearance inspection.

[0012] Preferably, the step of selecting sensitive sub-bands based on the magnitude of energy deviation feature values ​​includes: selecting the sub-band with the largest energy deviation feature value as the sensitive sub-band from all sub-bands at all levels of the target image.

[0013] This invention selects the sub-band with the largest energy deviation feature value from all sub-bands at all levels of the target image as the sensitive sub-band, which can accurately capture the most significant degradation features of the current product. This allows the system to automatically focus on the feature scale and direction that best reflects the current main degradation mode, avoiding information redundancy and computational burden caused by processing all sub-bands equally.

[0014] Preferably, constructing the wavelet coefficient deviation map includes: calculating the difference between the wavelet coefficient values ​​of the target image and the standard image at the same position in the sensitive sub-band, which is taken as the wavelet coefficient deviation between the target image and the standard image at the corresponding position in the sensitive sub-band, and constructing the wavelet coefficient deviation map by combining the wavelet coefficient deviations of the target image and the standard image at all positions in the sensitive sub-band.

[0015] Preferably, dividing the wavelet coefficient deviation map into multiple key regions includes: dividing the wavelet coefficient deviation map into several key regions according to the geometric regions corresponding to the key physical locations of the mold.

[0016] Preferably, determining the degradation concentration based on the wavelet coefficient deviation of the key region includes: calculating the mean of the wavelet coefficient deviation of all positions in each key region as the deviation index of each key region, and using the variance of the deviation indices of all key regions as the degradation concentration.

[0017] This invention uses the variance of the deviation index of all key regions as the degradation concentration, which can effectively quantify the unevenness of degradation in spatial distribution. When degradation is a systematic local problem, the wavelet coefficient deviation will be concentrated in a few key regions, resulting in significant differences in the deviation index of different key regions, thus making the degradation concentration value large. When degradation is a global problem or random noise, the deviation index of each key region tends to be consistent, making the degradation concentration value close to 0. Therefore, by calculating the degradation concentration, this invention can effectively distinguish between systematic degradation and random fluctuations, reducing the false alarm rate.

[0018] Preferably, determining the degradation trend degree based on the temporal change of degradation concentration includes: taking the most recent The degradation concentration of each image of the product to be inspected constitutes a degradation sequence; the degradation trend of the target image is determined based on the degradation sequence. ;in, Indicates the degree of degradation trend of the target image; Indicates the index of the degradation concentration in the degradation sequence; Indicates the first deterioration in the sequence Individual deterioration concentration; Indicates the window size.

[0019] This invention, by calculating the degradation trend degree, can capture the development trend of the degradation process over time, thereby distinguishing between real systematic degradation and random fluctuations, avoiding false alarms caused by relying solely on a single measurement, and improving the reliability of mold condition monitoring and the accuracy of early warning.

[0020] Preferably, the comprehensive early warning score satisfies the expression: ;in, This indicates the overall early warning score; Indicates the concentration of degradation in the target image; Indicates the degree of degradation trend of the target image; This represents the energy deviation characteristic value of the target image in the sensitive sub-band; Represents a linear rectified function; This indicates normalization.

[0021] This invention integrates the energy deviation characteristics, degradation concentration, and degradation trend of the target image in the sensitive sub-band, achieving a three-dimensional degradation characterization of "instantaneous-space-time". This enables the early warning system to comprehensively assess the mold condition and identify the true trend in the early stage of gradual degradation, avoiding the lag problem of traditional detection methods that can only identify visible defects. It shifts the quality control checkpoint from post-screening to pre-intervention, providing a decision-making basis for mold maintenance, effectively reducing the scrap rate and improving the stability of the production process.

[0022] Secondly, the present invention provides an injection mold production quality supervision system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned injection mold production quality supervision method is implemented.

[0023] By adopting the above technical solution, a computer program for the above-mentioned injection mold production quality supervision method is generated and stored in a memory so that it can be loaded and executed by a processor. This allows for the creation of terminal equipment based on the memory and processor, making it convenient to use.

[0024] The beneficial effects of this invention are as follows:

[0025] This invention acquires the sub-band energy of each level of the target image and the standard image, enabling precise quantification of the texture and gloss characteristics of the product surface at different scales and directions. It utilizes energy deviation characteristic values ​​to quantify microscopic changes on the product surface, capturing early signs of degradation that are difficult to detect with the naked eye. By calculating degradation concentration, it effectively distinguishes between systematic local degradation and global random fluctuations, thus avoiding misjudgments. By integrating energy deviation characteristic values, degradation concentration, and degradation trend, this invention achieves a three-dimensional degradation characterization encompassing "instantaneous-spatial-temporal" dimensions. This allows the system to provide early, sensitive, and reliable warnings for gradual degradation, avoiding the lag problem of traditional detection methods that can only identify visible defects. It shifts the quality control checkpoint from post-screening to pre-intervention, significantly reducing scrap rates and improving the stability of the production process. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a method for monitoring the production quality of injection molds according to the present invention;

[0027] Figure 2 It is a standard image diagram;

[0028] Figure 3 This is a schematic diagram of the low-frequency approximate subband of the first level of the standard image;

[0029] Figure 4 This is a schematic diagram of the high-frequency detail subband in the horizontal direction of the first level of the standard image;

[0030] Figure 5 This is a schematic diagram of the high-frequency detail subband in the vertical direction of the first level of the standard image;

[0031] Figure 6 This is a schematic diagram of the high-frequency detail subband in the diagonal direction of the first level of the standard image;

[0032] Figure 7 This is a schematic diagram of the low-frequency approximate subband of the second level of the standard image;

[0033] Figure 8 This is a schematic diagram of the high-frequency detail subband in the horizontal direction of the second level of the standard image;

[0034] Figure 9 This is a schematic diagram of the high-frequency detail subband in the vertical direction of the second level of the standard image;

[0035] Figure 10 This is a schematic diagram of the high-frequency detail subband in the diagonal direction of the second level of the standard image;

[0036] Figure 11 This is a schematic diagram of the target image;

[0037] Figure 12 This is a schematic diagram of the low-frequency approximate subband of the first level of the target image;

[0038] Figure 13 This is a schematic diagram of the high-frequency detail subband in the horizontal direction of the first level of the target image;

[0039] Figure 14 This is a schematic diagram of the high-frequency detail subband in the vertical direction of the first level of the target image;

[0040] Figure 15 This is a schematic diagram of the high-frequency detail subband in the diagonal direction of the first level of the target image;

[0041] Figure 16 This is a schematic diagram of the low-frequency approximate subband of the second level of the target image;

[0042] Figure 17 This is a schematic diagram of the high-frequency detail subband in the horizontal direction of the second level of the target image;

[0043] Figure 18 This is a schematic diagram of the high-frequency detail subband in the vertical direction of the second level of the target image;

[0044] Figure 19 This is a schematic diagram of the high-frequency detail subband in the diagonal direction of the second level of the target image;

[0045] Figure 20 This is a waveform coefficient deviation diagram. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0048] This invention discloses a method for quality supervision in injection mold production, referring to... Figure 1 This includes steps S1-S5:

[0049] S1. Select the most representative standard image from the defect-free qualified product images, perform two-dimensional discrete wavelet transform on the standard image, and obtain the energy of each sub-band of each level of the standard image.

[0050] Specifically, in the initial production phase after a new mold is put into use or after deep maintenance, an industrial camera on the production line is used to capture images of a set (exemplary, such as 100 to 1000) of continuously produced and confirmed defect-free qualified products to form a benchmark sample set.

[0051] Each image in the benchmark sample set is preprocessed. The preprocessing process includes image registration and illumination normalization. Image registration is used to eliminate minor position and pose changes of the product in the camera's field of view. In this embodiment, the SIFT feature matching algorithm combined with the RANSAC algorithm is used. In other embodiments, the implementer can choose the image registration algorithm according to the actual implementation situation. Illumination normalization is used to reduce the influence of ambient light fluctuations on the image grayscale distribution. In this embodiment, histogram equalization is used for illumination normalization. In other embodiments, the implementer can choose the algorithm for illumination normalization according to the actual implementation situation, such as the Retinex algorithm.

[0052] Furthermore, from the preprocessed benchmark sample set, the most representative image is selected as the standard image.

[0053] In one embodiment, relevant personnel select standard images from a preprocessed set of benchmark samples based on their experience.

[0054] In another embodiment, the standard image is selected as follows: for any image in the preprocessed benchmark sample set, the sum of differences between the image and all other images is calculated, and the image with the smallest sum of differences is selected as the standard image.

[0055] Furthermore, a two-dimensional discrete wavelet transform is performed on the standard image to decompose it into... The system comprises three levels, each containing a low-frequency approximation sub-band and three high-frequency detail sub-bands in different directions (horizontal, vertical, and diagonal). This represents the number of levels in the wavelet transform, in this invention, The value can be set from 3 to 5. Implementers can adjust the number of layers according to the actual implementation situation. For example, for products with high gloss requirements, the number of layers can be appropriately increased to capture more subtle texture variations. For example, Figure 2 This is a schematic diagram of a standard image. Figure 3 This is the low-frequency approximate subband of the first level of the standard image. Figure 4 , Figure 5 , Figure 6 These are the high-frequency detail subbands in the horizontal, vertical, and diagonal directions of the first layer of the standard image, respectively. Figure 7 This is the low-frequency approximate subband of the second level of the standard image. Figure 8 , Figure 9 , Figure 10 These are the high-frequency detail subbands in the horizontal, vertical, and diagonal directions of the second level of the standard image, respectively.

[0056] Obtain the energy of the standard image in each subband:

[0057] ;

[0058] in, The first standard image The first level The energy carried by a person; The first standard image The first level Coordinates in each sub-band Wavelet coefficient values ​​at; The first standard image The size of each sub-band at each level, The first standard image The number of rows in each sub-band of each level. The first standard image The number of columns in each sub-band of each level.

[0059] It should be noted that energy refers to the energy density of a subband, which reflects the intensity of local changes in the image at the corresponding level and direction. The higher the energy, the richer the texture or edge information contained in that level. When there are micro-changes on the product surface (such as increased roughness), the energy of the high-frequency detail subband usually increases. When the gloss of the product surface decreases or becomes blurred, the energy of the low-frequency approximation subband usually decreases.

[0060] S2. Real-time acquisition of images of the product to be tested as target images. Two-dimensional discrete wavelet transform is performed on the target images to obtain the energy of each sub-band of each level of the target images. Based on the difference between the energy of each sub-band of each level of the target images and the energy of the corresponding sub-band of the corresponding level of the standard image, the energy deviation characteristic value of each sub-band of the target images is determined.

[0061] Specifically, during the continuous operation of the production line, images of each product to be inspected are acquired in real time. Each image is preprocessed, and the preprocessing process is the same as that in step S1. The preprocessed image of the current product to be inspected is used as the target image. A two-dimensional discrete wavelet transform is performed on the target image to obtain all sub-bands at each level, and the energy of each sub-band at each level of the target image is obtained. Figure 11 This is a schematic diagram of a standard image. Figure 12 This is the low-frequency approximate subband of the first level of the standard image. Figure 13 , Figure 14 , Figure 15 These are the high-frequency detail subbands in the horizontal, vertical, and diagonal directions of the first layer of the standard image, respectively. Figure 16 This is the low-frequency approximate subband of the second level of the standard image. Figure 17 , Figure 18 , Figure 19 These are the high-frequency detail subbands in the horizontal, vertical, and diagonal directions of the second level of the standard image, respectively.

[0062] It should be noted that when microscopic changes occur on the product surface, such as increased surface roughness or decreased gloss, a redistribution of wavelet domain energy will inevitably occur. Therefore, this invention obtains the energy deviation characteristic value of the corresponding sub-band by calculating the energy difference between the current product to be detected and the standard image in the corresponding sub-band, which is used to quantify the slight deviation of the current product state from the ideal healthy state.

[0063] Specifically, the energy deviation eigenvalue satisfies the expression:

[0064] ;

[0065] In the formula, The first part represents the target image. The first level Energy deviation characteristic value of each sub-band; The first part represents the target image. The first level The energy carried by a person; The first standard image The first level The energy carried by a person; These are hyperparameters used to prevent the denominator from being zero. In this invention... Implementers can set the parameters according to the actual implementation situation, but to avoid exceeding the parameter limit... An excessively large value will affect the energy deviation eigenvalue. Need to be Within the range. It should be noted that this invention refers to the energy of the sub-band corresponding to the standard image. Using relative deviation as the denominator can eliminate the influence of absolute energy dimensions.

[0066] It should be further explained that when micro-texture changes occur on the product surface, such as increased surface roughness due to mold wear, the energy of the high-frequency detail subband will rise slightly, resulting in a small positive energy deviation characteristic value for the corresponding high-frequency detail subband. Conversely, when the product surface gloss decreases or becomes blurred, the energy of the low-frequency approximate subband will decrease, causing the energy deviation characteristic value for the low-frequency approximate subband to be negative. Therefore, the energy deviation characteristic value can reflect different types of early degradation.

[0067] S3. Select sensitive sub-bands based on the magnitude of energy deviation characteristic values. Construct a wavelet coefficient deviation map based on the difference between the target image and the standard image in the sensitive sub-bands. Divide the wavelet coefficient deviation map into multiple key regions. Determine the degradation concentration based on the wavelet coefficient deviation of the key regions.

[0068] It should be noted that gradual degradation usually does not occur uniformly throughout the entire mold, but rather begins in specific local areas, such as gates, ejector pin holes, or the edge of the parting line. To effectively capture this spatial unevenness, this invention constructs a wavelet coefficient deviation map and divides the wavelet coefficient deviation map into multiple key regions, determining the degradation concentration based on the wavelet coefficient deviation of the key regions.

[0069] Specifically, the subband with the largest energy deviation eigenvalue best reflects the current main degradation mode. Therefore, among all subbands at all levels of the target image, the subband with the largest energy deviation eigenvalue is selected as the sensitive subband.

[0070] For example, Figure 11 target image and Figure 2 The standard image has the largest energy deviation eigenvalue in the high-frequency detail subband in the diagonal direction of the first level, therefore the high-frequency detail subband in the diagonal direction of the first level is the sensitive subband.

[0071] Furthermore, based on the differences between the target image and the standard image in the sensitive sub-bands, a wavelet coefficient deviation map is constructed:

[0072] ;

[0073] in, Represents the coordinates of the target image and the standard image in the sensitive sub-band. Wavelet coefficient deviation at; Represents the coordinates of the target image in the sensitive sub-band. Wavelet coefficient values ​​at; Represents the coordinates of the standard image in the sensitive subband. Wavelet coefficient values ​​at; The sign indicates the absolute value. A wavelet coefficient deviation map is constructed by comparing the wavelet coefficients of the target image and the standard image at all coordinates within the sensitive sub-band. Figure 20 This is a graph showing the deviation of wavelet coefficients.

[0074] Furthermore, the wavelet coefficient deviation map is divided into several key regions according to the geometric regions corresponding to the critical physical locations of the mold (such as gates, ejector pin positions, parting line edges, etc.). The mean value of the wavelet coefficient deviation at all locations in each key region is calculated as the deviation index for each key region. It should be noted that the number and shape of the key regions can be set by the implementers according to the mold structure and product characteristics.

[0075] Furthermore, based on the deviation index of all key regions, the degradation concentration of the target image is determined:

[0076] ;

[0077] in, It represents the concentration of degradation in the target image, used to characterize the degree of non-uniformity in the spatial distribution of degradation signals; The first wavelet coefficient deviation plot represents the... Deviation index for key areas; This represents the mean of the wavelet coefficient deviations at all positions in the wavelet coefficient deviation plot; This indicates the number of critical regions. If the degradation is a systemic, localized problem, such as degradation caused by localized wear of the mold, the wavelet coefficient deviation will be concentrated in a few critical regions, resulting in differences in the wavelet coefficient deviation between different critical regions, thus increasing the degradation concentration. The degradation is relatively large; if the degradation becomes a global problem (such as overall color difference due to batch variations in raw materials) or random noise, the wavelet coefficient deviations in each key area will tend to be consistent, resulting in a higher concentration of degradation. The value is close to 0.

[0078] S4. Determine the degradation trend degree based on the temporal changes in degradation concentration, and obtain a comprehensive early warning score based on the energy deviation characteristic value of the target image in the sensitive sub-band, the degradation concentration degree of the target image, and the degradation trend degree.

[0079] It should be noted that the actual physical degradation process has temporal continuity and inertia, and its effects accumulate with the increase in production frequency. For example, mold wear gradually intensifies with the increase in production volume. Therefore, this invention obtains the degradation trend degree based on the temporal changes in degradation concentration in order to confirm the authenticity of the degradation trend from a time dimension.

[0080] Specifically, a sliding time window of size W is constructed, and the degradation concentration corresponding to the images of the most recent W products to be inspected is used to form a one-dimensional sequence, denoted as the degradation sequence. Here, W is a preset window size. In this embodiment, W=50. In other embodiments, the implementer can set W according to the actual implementation situation. For example, for molds with rapid degradation development, a smaller window (e.g., W=30) can be set to improve the response speed; for molds with slow degradation development, a larger window (e.g., W=100) can be set to improve the stability of trend estimation.

[0081] Determine the degradation trend of the target image based on the degradation sequence:

[0082] ;

[0083] in, Indicates the degree of degradation trend of the target image; Indicates the index of the degradation concentration in the degradation sequence; Indicates the first deterioration in the sequence The degradation concentration is represented by W, where W represents the window size. This invention uses a univariate linear regression slope estimation formula to obtain the degradation trend of the mold. The degradation trend of the target image is equal to the fitting slope of the degradation sequence. If the change in degradation concentration is random, the fitting slope of the degradation sequence will be close to 0. Conversely, if there is a continuous degradation process, the degradation concentration in the degradation sequence will show a monotonically increasing trend, making the fitting slope of the degradation sequence a large positive value.

[0084] It should be noted that a true and urgent degradation trend should simultaneously exhibit large instantaneous deviations, uneven spatial distribution, and continuous deterioration over time. The energy deviation characteristic values ​​of the target image in each sub-band at each level reflect the instantaneous deviation of the target image, with the energy deviation characteristic value in the sensitive sub-band reflecting the most significant instantaneous deviation. The degradation concentration of the target image reflects the degree of concentration of degradation in spatial distribution, while the degradation trend reflects the rate of continuous deterioration over time. Therefore, this invention obtains a comprehensive early warning score based on the energy deviation characteristic values ​​of the target image in the sensitive sub-band, the degradation concentration of the target image, and the degradation trend.

[0085] Specifically, the comprehensive early warning score satisfies the expression:

[0086] ;

[0087] in, This indicates the overall early warning score; Indicates the concentration of degradation in the target image; Indicates the degree of degradation trend of the target image; This represents the energy deviation characteristic value of the target image in the sensitive sub-band; Represents a linear rectification function, when the degradation trend is... When it is a negative value or 0, A value of 0 indicates a deterioration trend. When it is a positive value, Values This effectively suppresses false alarms caused by random noise; This represents the normalization function. In this embodiment, based on historical data from the first 1000 products, a linear normalization method is used to normalize the energy deviation eigenvalues. Deterioration concentration and the mapping value of the degradation trend. Normalization is performed; in other embodiments, the implementer can choose the normalization method according to the actual implementation situation; when the energy deviation characteristic value Deterioration concentration and the degree of degradation trend When both are large, it indicates that the target image has a significant instantaneous deviation, and the degradation is concentrated in space and continues to worsen over time. In this case, a large comprehensive warning score indicates a serious degradation trend in the mold. When the energy deviation characteristic value... Deterioration concentration and the degree of degradation trend When any one of the values ​​is relatively small, it indicates that the degradation characteristics of the target image are not obvious or are only random fluctuations. In this case, the overall warning score is relatively small, indicating that the mold is in normal condition.

[0088] S5. Issue warnings based on the overall warning score.

[0089] Specifically, an alert is generated in response to a comprehensive alert score exceeding a preset alert threshold. The alert information includes: the spatial location of the degradation (the critical area with the highest deviation index in the wavelet coefficient deviation graph) and the urgency of the degradation (degradation trend degree T). It should be noted that the alert threshold is set by the implementers based on production experience and risk tolerance. In this embodiment, the alert threshold is set to 0.3. Implementers can adjust the alert threshold according to actual implementation conditions. For example, for products with extremely high quality requirements, the alert threshold can be appropriately lowered to improve sensitivity; for products that allow for a certain degree of degradation, the alert threshold can be appropriately increased to reduce false alarms.

[0090] Furthermore, when an early warning message is generated, the system automatically generates a maintenance work order and notifies relevant personnel to conduct mold inspections based on the warning message.

[0091] This invention also discloses an injection mold production quality supervision system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, an injection mold production quality supervision method according to the present invention is implemented.

[0092] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for quality supervision in injection mold production, characterized in that, include: Select the most representative standard image from the images of defect-free qualified products, and acquire images of the product to be inspected in real time as the target image; Two-dimensional discrete wavelet transforms are performed on the standard image and the target image respectively to obtain the energy of each sub-band at each level of the standard image and the target image. Based on the energy difference between each sub-band of each level in the target image and the corresponding sub-band of the corresponding level in the standard image, the energy deviation characteristic value of each sub-band in the target image is determined, which satisfies the expression: ; In the formula, The first part represents the target image. The first level Energy deviation characteristic value of each sub-band; The first part represents the target image. The first level The energy carried by a person; The first standard image The first level The energy carried by a person; This is a hyperparameter used to prevent the denominator from being zero; Sensitive sub-bands are selected based on the magnitude of energy deviation eigenvalues. A wavelet coefficient deviation map is constructed based on the differences between the target image and the standard image within these sensitive sub-bands. This wavelet coefficient deviation map is then divided into multiple key regions. The degradation concentration is determined based on the wavelet coefficient deviations within these key regions, including: Calculate the mean of the wavelet coefficient deviations at all locations in each key region as the deviation index for each key region, and use the variance of the deviation indices of all key regions as the deterioration concentration. The degradation trend is determined by the temporal changes in degradation concentration. A comprehensive early warning score is obtained based on the energy deviation characteristic value of the target image in the sensitive sub-band, the degradation concentration of the target image, and the degradation trend. An early warning is issued based on the magnitude of the comprehensive early warning score.

2. The method for quality supervision in injection mold production according to claim 1, characterized in that, Also includes: The standard image and the target image are preprocessed separately. The preprocessing includes image registration and illumination normalization. Image registration uses the SIFT feature matching algorithm combined with the RANSAC algorithm, and illumination normalization uses histogram equalization.

3. The method for quality supervision in injection mold production according to claim 1, characterized in that, The process of selecting sensitive sub-bands based on the magnitude of energy deviation characteristic values ​​includes: Among all subbands at all levels of the target image, the subband with the largest energy deviation eigenvalue is selected as the sensitive subband.

4. The method for quality supervision in injection mold production according to claim 1, characterized in that, The construction of the wavelet coefficient deviation map includes: The difference between the wavelet coefficient values ​​at the same position in the sensitive sub-band between the target image and the standard image is calculated as the wavelet coefficient deviation between the target image and the standard image at the corresponding position in the sensitive sub-band. The wavelet coefficient deviation between the target image and the standard image at all positions in the sensitive sub-band is used to construct a wavelet coefficient deviation map.

5. The method for quality supervision in injection mold production according to claim 1, characterized in that, The wavelet coefficient deviation map is divided into several key regions, including: The wavelet coefficient deviation map is divided into several key regions according to the geometric regions corresponding to the key physical locations of the mold.

6. The method for quality supervision in injection mold production according to claim 1, characterized in that, The determination of degradation trend degree based on the temporal changes in degradation concentration includes: Recently The degradation concentration of each image of the product to be inspected constitutes a degradation sequence; the degradation trend of the target image is determined based on the degradation sequence. ； in, Indicates the degree of degradation trend of the target image; Indicates the index of the degradation concentration in the degradation sequence; Indicates the first deterioration sequence Individual deterioration concentration; Indicates the window size.

7. The method for quality supervision in injection mold production according to claim 1, characterized in that, The comprehensive early warning score satisfies the expression: ； in, This indicates the overall early warning score; Indicates the concentration of degradation in the target image; Indicates the degree of degradation trend of the target image; This represents the energy deviation characteristic value of the target image in the sensitive sub-band; Represents a linear rectified function; This indicates normalization.

8. A quality supervision system for injection mold production, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a method for monitoring the production quality of injection molds according to any one of claims 1-7.

Images