Cable surface defect detection method

Abstract

This invention provides a method for detecting surface defects in cables, applied in the field of machine vision technology. The method includes: identifying defects in each image from a sequence of images of the same batch and roll of cables continuously acquired within a preset time period, and determining the physical location of the defects on the cable surface; determining the timestamp of the defect's occurrence on the production line based on the cable's traction speed and physical location; extracting multiple process parameter values ​​to be analyzed within an estimated time interval that is expected to affect the defect from the production line based on the timestamp, obtaining a sequence of process parameters to be analyzed, with the timestamp falling within the estimated time interval; determining the parameter deviation degree based on the difference between the sequence of process parameters to be analyzed and a reference sequence, where the reference sequence is a sequence of multiple reference process parameter values ​​under a defect-free production state corresponding to the estimated time interval; and determining the cause type leading to the formation of defects on the cable surface based on the parameter deviation degree.

Description

Technical Field

[0001] This invention relates to the field of machine vision technology, and more specifically to a method for detecting surface defects in cables. Background Technology

[0002] With the rapid development of industries such as power and communications, the quality requirements for cable products are becoming increasingly stringent. Surface defects in cables, such as scratches, dents, scorching, and off-center cores, not only affect the product's appearance but also pose a potential threat to its insulation performance, mechanical strength, and long-term reliability, potentially leading to serious quality accidents. Traditional cable surface defect detection mainly relies on manual visual inspection, which is inefficient, subjective, and prone to fatigue-induced missed defects, and can no longer meet the quality control needs of modern high-speed production lines.

[0003] Currently, machine vision technology has been introduced into the field of cable production quality inspection. Industrial cameras automatically capture images of cable surfaces, and image processing algorithms are used to identify defects. However, most existing machine vision-based inspection systems are still at the rudimentary stage of identification and alarm functions. Summary of the Invention

[0004] In view of this, the present invention provides a method for detecting surface defects in cables.

[0005] This invention provides a method for detecting surface defects in cables, comprising: identifying defects in each image from a sequence of images of the same batch and roll of cables continuously acquired within a preset time period, and determining the physical location of the defects on the cable surface; determining the timestamp of the defects' occurrence on the production line based on the traction speed and physical location of the cable; extracting multiple process parameter values ​​to be analyzed within an estimated time interval that is expected to affect the defects from the production line based on the timestamp, obtaining a sequence of process parameters to be analyzed, wherein the timestamp of the defects is located within the estimated time interval; determining the parameter deviation degree based on the difference between the sequence of process parameters to be analyzed and a reference sequence, wherein the reference sequence is a sequence of multiple reference process parameter values ​​under a defect-free production state corresponding to the estimated time interval; and determining the cause type of defects on the cable surface based on the parameter deviation degree.

[0006] According to an embodiment of the present invention, the cable surface defect detection method further includes: when it is determined that there are multiple defects on the cable surface and the multiple defects have the same defect appearance category, determining the temporal distribution density of the multiple defects in the time dimension based on the timestamps of their respective generation on the production line; determining the spatial distribution density of the multiple defects in the spatial dimension based on their respective physical locations on the cable surface; determining the spatiotemporal clustering degree of the defects based on the temporal distribution density and the spatial distribution density; and determining the cause type of the formation of the multiple defects on the cable surface based on the parameter deviation and the spatiotemporal clustering degree of the defects.

[0007] According to an embodiment of the present invention, determining the temporal distribution density of multiple defects in the time dimension based on the timestamps of their respective generation on the production line includes: determining the generation time interval between adjacent defects based on the timestamps of their respective generation on the production line; obtaining the quantified temporal clustering of multiple defects based on the variance of the generation time intervals, and determining it as the temporal distribution density.

[0008] According to an embodiment of the present invention, determining the spatial distribution density of multiple defects in a spatial dimension based on the physical location of each defect on the cable surface includes: clustering multiple defects using a clustering algorithm based on the physical location of each defect on the cable surface to identify at least one defect cluster; taking the cable length range covered by each defect cluster as the cluster range, and determining the spatial distribution density by the ratio of the number of defects in the cluster to the corresponding cluster range.

[0009] According to an embodiment of the present invention, the cause type of multiple defects on the cable surface is determined based on parameter deviation and defect spatiotemporal clustering, including: normalizing the parameter deviation and defect spatiotemporal clustering respectively to obtain normalized parameter deviation and normalized defect spatiotemporal clustering; weighted summing the normalized parameter deviation and normalized defect spatiotemporal clustering to obtain a defect type tendency value; if the defect type tendency value is greater than or equal to a predetermined threshold, the cause type is determined to be a systematic process anomaly; if the defect type tendency value is less than the predetermined threshold, the cause type is determined to be an instantaneous random occurrence.

[0010] According to an embodiment of the present invention, the cable surface defect detection method further includes: extracting appearance features of defect areas in each image of an image sequence from multiple visual dimensions, wherein the defect is determined based on the difference between the image and a standard image in at least one of the image's sharpness, brightness, and color difference; and, when it is determined that multiple defects exist on the cable surface, classifying the appearance features corresponding to the multiple defects to obtain defect appearance categories of the multiple defects.

[0011] According to an embodiment of the present invention, the cable surface defect detection method further includes: when the cause type is determined to be a systematic process anomaly, identifying abnormal process parameters from the sequence of process parameters to be analyzed, and determining the abnormal process parameters as the cause of the defect; when the cause type is determined to be an instantaneous random occurrence, triggering the manufacturing execution system marking and isolation process.

[0012] According to an embodiment of the present invention, the cable surface defect detection method further includes: when the cause of the defect is determined, determining an accuracy assessment value for estimating the accuracy level of the cause of the defect based on the number of causes of the defect and the duration of the abnormal process parameters corresponding to the causes of the defect; and determining a processing strategy that matches the accuracy assessment value so as to process the cable with defects based on the processing strategy.

[0013] According to an embodiment of the present invention, determining a processing strategy matching the accuracy assessment value includes: when the accuracy assessment value is determined to be a first accuracy level, the processing strategy matching the accuracy assessment value is to send a first type of instruction to the manufacturing execution system, the first type of instruction instructing the manufacturing execution system to issue a process parameter adjustment command to the production line control system, and to track and verify the effect of the adjustment; when the accuracy assessment value is determined to be a second accuracy level, the processing strategy matching the accuracy assessment value is to send a second type of instruction to the manufacturing execution system, the second type of instruction instructing the manufacturing execution system to generate a repair work order containing the cause of the defect and the corresponding abnormal process parameters, and to push it to the equipment maintenance terminal; when the accuracy assessment value is determined to be a third accuracy level, the processing strategy matching the accuracy assessment value is to send a third type of instruction to the manufacturing execution system, the third type of instruction instructing the manufacturing execution system to trace back the cause of the first occurrence of the defect; wherein, the first accuracy level is greater than the second accuracy level, and the second accuracy level is greater than the third accuracy level.

[0014] According to an embodiment of the present invention, the manufacturing execution system traces the cause of the first occurrence of a defect by performing the following operations: determining the temporal stability of the cause of the defect based on the abnormal duration and abnormal fluctuation frequency; if the temporal stability is determined to be greater than or equal to a preset temporal stability, corresponding processing is performed on the cause of the first occurrence of the defect; if the temporal stability is determined to be less than the preset temporal stability, the cause of the defect that occurs at the next time point is traced.

[0015] According to embodiments of the present invention, the physical location of a defect is identified from a sequence of images of the same batch and roll of cable continuously acquired within a preset time period. Then, using this physical location and the traction speed of the corresponding production line, the timestamp of the defect's occurrence on the production line is calculated in reverse. This allows for a precise correspondence between a spatial defect point and the production process state on the timeline, breaking down the data barrier between visual inspection and process control. Defects are no longer viewed in isolation; instead, the sequence of process parameters before and after the timestamp is extracted. Based on the complete production timeline context, the defect is analyzed and detected, thereby establishing a traceable chain from the surface result to the process cause. This enables tracing the causal type of the defect on the cable surface, facilitating accurate location of the defect's root cause, targeted handling of anomalies, improved production quality control efficiency, and reduced losses. Attached Figure Description

[0016] The above and other objects, features and advantages of the present invention will become clearer from the following description of embodiments of the invention with reference to the accompanying drawings.

[0017] Figure 1 A flowchart of a cable surface defect detection method according to an embodiment of the present invention is shown.

[0018] Figure 2 A flowchart of a cable surface defect detection method according to an embodiment of the present invention is shown.

[0019] Figure 3 A flowchart of a cable surface defect detection method according to an embodiment of the present invention is shown.

[0020] Figure 4 A block diagram of a cable surface defect detection device according to an embodiment of the present invention is shown. Detailed Implementation

[0021] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the invention. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the invention for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0022] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The terms “comprising,” “including,” etc., as used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0023] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0024] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0025] The following will be through Figures 1-3 The cable surface defect detection method according to an embodiment of the present invention will be described in detail.

[0026] Figure 1 A flowchart of a cable surface defect detection method according to an embodiment of the present invention is shown.

[0027] like Figure 1As shown, the cable surface defect detection method includes steps S110 to S150.

[0028] In step S110, defects in each image are identified from a sequence of images of the same batch and roll of cable that are continuously collected within a preset time period, and the physical location of the defects on the cable surface is determined.

[0029] In step S120, the timestamp of the defect on the production line is determined based on the traction speed and physical location of the cable on the production line.

[0030] In step S130, based on the generated timestamp, multiple process parameter values ​​to be analyzed within the estimated time interval that is expected to affect the existence of defects are extracted from the production line to obtain the process parameter sequence to be analyzed.

[0031] In step S140, the parameter deviation is determined based on the difference between the process parameter sequence to be analyzed and the reference sequence.

[0032] In step S150, the cause type of defects that lead to the formation of the cable surface is determined based on the parameter deviation.

[0033] In embodiments of the present invention, the preset time period is not specifically limited, but may be predetermined based on the detection accuracy. The presence of defects in the image can be understood as a difference in clarity and / or brightness and / or color difference between the cable surface at a certain location in the image and the standard image.

[0034] For example, a calibration mapping relationship between the pixel coordinates of the image and the physical length of the cable can be established in advance; by combining the pixel position of the defect in the image, the cable travel speed and the image acquisition frame sequence, the physical length of the defect from the cable production start point can be calculated, which is the precise physical position of the defect on the cable surface.

[0035] A timestamp is generated to indicate when the defect occurred during cable production.

[0036] For example, the time elapsed from the start of cable production to the occurrence of the defect can be obtained from the ratio of physical location to traction speed; the time is then added to the reference time of the start of production of the cable segment to obtain the timestamp of the defect's occurrence on the production line.

[0037] The generated timestamp is within the estimated time interval.

[0038] For example, the estimated time interval can be determined based on the preset backward tracing duration and preset forward tracing duration according to process knowledge. The preset backward tracing duration can be used to observe the continued impact of defects, and the preset forward tracing duration can be used to capture the potential causes of defects. For example, if the timestamp is T0, the estimated time interval can be [T0-T1, T0+T2], where T1 represents the preset backward tracing duration and T2 represents the preset forward tracing duration.

[0039] A baseline sequence can be a sequence of multiple baseline process parameter values ​​under defect-free production conditions, corresponding to an estimated time interval. For example, a baseline sequence can be obtained through one or a combination of the following methods: calling up similar process parameter values ​​from historical gold production batches stored in the manufacturing execution system within the same time period as the baseline sequence; and / or dynamically generating a baseline sequence based on process parameter values ​​from a stable operating period prior to the defect occurrence timestamp of the current production order. Historical gold production batches can be understood as production batches that have completed production, warehousing, inspection, and shipment, and have entered the historical database.

[0040] Parameter deviation can be used to characterize the degree to which a sequence of process parameters to be analyzed deviates from a reference sequence. For example, it can be determined by the arithmetic mean and standard deviation of both the sequence of process parameters to be analyzed and the reference sequence.

[0041] The causes can include, but are not limited to, systemic process anomalies and transient random occurrences. For example, parameter deviations can be compared with a predetermined deviation threshold range to determine the cause of defects on the cable surface. The predetermined deviation threshold range can be determined based on practical experience regarding systemic process anomalies and transient random occurrences.

[0042] According to embodiments of the present invention, the physical location of a defect is identified from a sequence of images of the same batch and roll of cable continuously acquired within a preset time period. Then, using this physical location and the traction speed of the corresponding production line, the timestamp of the defect's occurrence on the production line is calculated in reverse. This allows for a precise correspondence between a spatial defect point and the production process state on the timeline, breaking down the data barrier between visual inspection and process control. Defects are no longer viewed in isolation; instead, the sequence of process parameters before and after the timestamp is extracted. Based on the complete production timeline context, the defect is analyzed and detected, thereby establishing a traceable chain from the surface result to the process cause. This enables tracing the causal type of the defect on the cable surface, facilitating accurate location of the defect's root cause, targeted handling of anomalies, improved production quality control efficiency, and reduced losses.

[0043] According to an embodiment of the present invention, the parameter deviation can be determined based on the deviation from the mean, the deviation from volatility, and the deviation from morphological characteristics.

[0044] For example, the deviation from the mean, D0, can be obtained using the following formula (1):

[0045] D0=|D1-D2| / C0(1)

[0046] Where D1 is the arithmetic mean of the process parameter sequence to be analyzed, and D2 and C0 are the arithmetic mean and standard deviation of the benchmark sequence, respectively.

[0047] The volatility deviation D' can be obtained by the following formula (2):

[0048] D'=C1 / C0(2)

[0049] Where C1 is the standard deviation of the process parameter sequence to be analyzed.

[0050] The minimum warped path distance Dd between the analyzed process parameter sequence and the reference sequence can be calculated using a dynamic time warping algorithm to measure the difference in the overall waveform between the two sequences. The minimum warped path distance Dd is then defined as the morphological deviation. The deviation from the mean, the volatility, and the morphological deviation can be weighted and fused to obtain the parameter deviation degree.

[0051] This invention integrates statistical and morphological features. Mean deviation can capture persistent parameter drift, such as a slow temperature rise caused by heater aging. Fluctuation deviation can detect the loss of system stability, such as a violent speed jump caused by a speed controller failure. Morphological feature deviation can identify complex abnormal waveforms that traditional statistical methods would ignore, such as a brief spike or a specific oscillation pattern. The multi-angle evaluation makes the capture of process anomalies more comprehensive and sensitive, reducing the false negative rate.

[0052] Even so, if there are multiple defects, the lack of macroscopic analysis capabilities on the distribution patterns of defects in the spatiotemporal dimensions makes it difficult to accurately distinguish between systemic process anomalies and instantaneous random events, thus failing to provide more accurate decision-making basis for subsequent processing.

[0053] For example, in a related example, the cable is supported within the annular limiting grooves of each limiting wheel. The cable limiting unit can adjust the distance between the upper and lower limiting wheels to accommodate cables of different sizes, increasing the applicability of the cable surface defect detection and classification device. Two industrial cameras comprehensively and meticulously capture images of the cables passing through the inspection mechanism, and the control mechanism analyzes the images transmitted by the two industrial cameras. When a problem appears on the cable surface, the control mechanism controls two cable-carrying mechanisms to transport the problematic portion of the cable to the vicinity of the cable folding and labeling machine. The cable folding and labeling machine then labels the area near the problematic portion of the cable. This serves two purposes: firstly, it proves that there is a problem on the surface of the bundle of cables to be inspected; secondly, it marks the location of the problem for later repair. However, this is only a simple classification based on image features, and the aforementioned problems still exist.

[0054] Figure 2 A flowchart of a cable surface defect detection method according to an embodiment of the present invention is shown.

[0055] In view of this, in another embodiment of the present invention, the cable surface defect detection method includes, in addition to the above-described methods, the following: Figure 1 In addition to steps S110 to S150 shown, the following may also be included: Figure 2 Steps S210 to S240 are shown.

[0056] In step S210, if it is determined that there are multiple defects on the cable surface and the defects have the same defect appearance category, the time distribution density of the multiple defects in the time dimension is determined based on the timestamps of their respective generation on the production line.

[0057] In step S220, the spatial distribution density of the multiple defects in the spatial dimension is determined based on the physical location of each defect on the cable surface.

[0058] In step S230, the spatiotemporal clustering degree of defects is determined based on the temporal distribution density and spatial distribution density.

[0059] In step S240, based on parameter deviation and spatiotemporal clustering of defects, the cause type of multiple defects on the cable surface is determined.

[0060] In this embodiment, the defect appearance category can be used to characterize surface morphology categories such as scratches, dents, and scorching. For example, defects can be classified based on image features to determine the defect appearance category.

[0061] For example, the time distribution density can be obtained by statistically analyzing the distribution of defects on the time axis and calculating the number of defects per unit time.

[0062] Spatial distribution density can be obtained by statistically analyzing the distribution of defect points along the length of the cable and calculating the number of defects per unit length.

[0063] Defect spatiotemporal clustering can be obtained by weighted summation of temporal and spatial distribution densities. For example, high defect spatiotemporal clustering indicates that defects repeatedly and concentratedly occur at specific locations or within a specific time period, which is a typical characteristic of systemic and root-cause problems, such as mold damage or periodic equipment vibration. Low defect spatiotemporal clustering indicates that defects are isolated and random, pointing to transient and accidental causes, such as falling objects or accidental scratches.

[0064] The defect type tendency value can be determined based on the parameter deviation and the spatiotemporal clustering of defects. By comparing the defect type tendency value with a predetermined threshold, the cause type of multiple defects on the cable surface can be determined.

[0065] This invention calculates the spatiotemporal clustering degree of defects, so that the system no longer views each defect in isolation, but can understand the distribution pattern of defects in time and space. It can directly determine the root cause type of defects from the macroscopic pattern, providing a key decision basis for subsequent targeted treatment strategies.

[0066] For example, determining the temporal distribution density of multiple defects in the time dimension based on their respective timestamps on the production line may include: determining the time interval between adjacent defects based on their respective timestamps on the production line; obtaining the quantified temporal clustering of multiple defects based on the variance of the time intervals, and determining it as the temporal distribution density.

[0067] In this example, a smaller variance indicates that the defects occur more concentratedly and at more uniform intervals, reflecting a concentrated outbreak of defects within a short period of time and a higher temporal distribution density. A larger variance indicates that the defects occur more dispersedly and with greater randomness, reflecting a lack of obvious temporal clustering characteristics and a lower temporal distribution density.

[0068] For example, consider a machine vision system performing surface quality inspection on optical fiber cables produced on a high-speed extrusion production line. The system can be configured to analyze a sequence of cables produced within the past 30 minutes. Within this time window, the system successfully identified 15 defects from a continuous sequence of cable surface images. Image feature analysis categorized these defects as surface scratches. Analyzing the timestamps of each defect revealed a highly concentrated occurrence: 12 scratches were generated within the last 8 minutes, forming a high-incidence period, while the other 3 occurred earlier. Based on this time information, the system calculated the temporal distribution density of these 15 defects. In this example, the high temporal distribution density is due to the concentrated occurrence of defects within a specific timeframe.

[0069] This invention determines the time distribution density based on the variance of the time intervals, which can sensitively quantify the temporal clustering of defects and improve the accuracy and stability of the time distribution density calculation.

[0070] For example, determining the spatial distribution density of multiple defects in a spatial dimension based on their respective physical locations on the cable surface may include: clustering the multiple defects using a clustering algorithm based on their respective physical locations on the cable surface to identify at least one defect cluster; using the cable length range covered by each defect cluster as the cluster range, and determining the spatial distribution density as the ratio of the number of defects in the cluster to the corresponding cluster range.

[0071] For example, continuing with the example above of a machine vision system inspecting the surface quality of a communication optical cable produced on a high-speed extrusion production line, the machine vision system can obtain the specific physical locations of the 15 scratches on the cable surface, such as their distance in meters from the cable's starting point. Analysis reveals that the scratches are not uniformly distributed along the entire cable length; up to 10 scratches appear densely within a relatively short length, such as within a continuous 50-meter cable segment, while the remaining 5 are scattered in other locations. Based on this location information, the machine vision system can calculate the spatial distribution density of these 15 defects. In this example, because the defects are highly concentrated in a specific area, the spatial distribution density is also relatively high.

[0072] This invention uses a clustering algorithm to calculate spatial distribution density, which can adaptively identify defect clustering areas and improve the accuracy and reliability of spatial distribution density calculation.

[0073] For example, determining the cause type of multiple defects on the cable surface based on parameter deviation and defect spatiotemporal clustering can include: normalizing the parameter deviation and defect spatiotemporal clustering respectively to obtain normalized parameter deviation and normalized defect spatiotemporal clustering; weighted summing the normalized parameter deviation and normalized defect spatiotemporal clustering to obtain a defect type tendency value; determining the cause type as a systematic process anomaly if the defect type tendency value is greater than or equal to a predetermined threshold; and determining the cause type as an instantaneous random occurrence if the defect type tendency value is less than the predetermined threshold.

[0074] In this example, the dimensions and ranges of parameter deviation and defect spatiotemporal clustering may differ. Normalization can map them to the same numerical range, such as between 0 and 1. Weighting can be applied, for example, by assigning higher weight to defect spatiotemporal clustering if the system tends to consider spatiotemporal clustering patterns as the primary basis for judging systemic problems. Similarly, higher weight can be assigned to parameter deviation if direct evidence of anomalies in process parameters is considered more important. The predetermined threshold can be an empirical value.

[0075] Systemic process anomalies can be understood as being caused by any continuous and systematic anomaly, such as equipment failure, workstation wear, or deviation of process parameters. For example, systemic process anomalies tend to be highly concentrated in space and occur continuously and in a concentrated manner in time.

[0076] Instantaneous random-initiated types can be understood as those caused by accidental and random factors such as instantaneous impurities in raw materials and instantaneous environmental disturbances. For example, instantaneous random-initiated types tend to be scattered in space and have no obvious aggregation in time.

[0077] This invention normalizes the calculated spatial and temporal distribution densities to bring them under a unified dimension, and finally obtains a comprehensive defect type tendency value by weighted summation. This enables the quantitative characterization of the type tendency of cable surface defects, accurately reflects the aggregation characteristics and anomaly degree of defects in the spatiotemporal dimension, and provides a quantitative basis for tracing the causes of defects and judging production anomalies.

[0078] According to embodiments of the present invention, the cable surface defect detection method includes, in addition to the above-described, Figure 1 In addition to steps S110 to S150, the method may further include: extracting the appearance features of defective regions in each image of the image sequence from multiple visual dimensions, wherein the defects are determined based on the differences between the image and the standard image in at least one of the image's sharpness, brightness, and color difference; and classifying the appearance features corresponding to the multiple defects to obtain the defect appearance categories of the multiple defects when it is determined that there are multiple defects on the cable surface.

[0079] In this embodiment, the image sequence consists of consecutive frames. Each frame can be analyzed using object detection or image segmentation algorithms to identify all potential defect regions in the image. The location and extent of each defect are then precisely marked using bounding boxes or pixel-level masks. For each detected defect region, morphological geometric features, texture features, color and brightness features, and context and intensity features can be extracted in parallel. All extracted feature values ​​are normalized and scaled to the [0,1] interval to eliminate the influence of dimensions. The normalized features are then concatenated into a one-dimensional vector in a preset order, representing the appearance features of a single defect. Classifying these appearance features determines the defect's appearance category.

[0080] This invention does not specifically limit the classification. For example, for multiple defects detected on the same batch and the same roll of cable, their respective appearance feature vectors, arranged in chronological order, can constitute a feature sequence for macroscopic analysis. Classifying this feature sequence allows determination of the defect appearance category of the multiple defects detected on the same batch and the same roll of cable. "Same roll of cable" can refer to cable products produced continuously and uninterruptedly on the same production equipment using the same raw materials, the same formula, and the same process parameters, and ultimately wound onto the same spool / bollard. Each roll of cable can have a unique roll number, serving as the smallest unit for quality traceability, defect recording, and production process data binding. "Same batch of cables" can refer to a group of cable products produced continuously or centrally using the same raw material batch number, the same product specifications, the same process formula, and the same production conditions, and assigned the same batch number by the system. "Same batch" can contain one or more rolls of cable.

[0081] If the actual detected defect is not a single defect appearance category, but a combination of multiple defects with different shapes, then the different defects can be initially distinguished and classified according to their appearance characteristics. Then, based on each type of defect, the subsequent operation provided by this invention to determine the cause type of the multiple defects on the cable surface can be performed.

[0082] This invention, by classifying defects into preliminary defect appearance categories, can distinguish defects of different morphologies, avoiding interference from mixed defect types in subsequent spatiotemporal distribution density calculations, and ensuring more accurate spatial and temporal clustering analysis of each defect. Furthermore, it is compatible with detection scenarios where multiple defects coexist, eliminating the need for separate processing procedures for each defect, thus improving the system's versatility, robustness, and detection efficiency.

[0083] Figure 3 A flowchart of a cable surface defect detection method according to an embodiment of the present invention is shown.

[0084] According to embodiments of the present invention, the cable surface defect detection method includes, in addition to the above-described, Figure 1In addition to steps S110 to S150 shown, the following may also be included: Figure 3 Steps S310 to S320 are shown.

[0085] In step S310, if the cause type is determined to be a systematic process anomaly, abnormal process parameters are identified from the sequence of process parameters to be analyzed, and the abnormal process parameters are identified as the cause of the defect.

[0086] In step S320, if the cause type is determined to be transient random triggering type, the manufacturing execution system mark isolation process is triggered.

[0087] In this embodiment, the isolation process may refer to marking defects in the corresponding production period and the corresponding cable section, and implementing isolation control to prevent defective products from flowing into subsequent processes.

[0088] In this invention, for defects caused by systemic process anomalies, the abnormal process parameters can be thoroughly investigated to identify the root cause of the defect, ensuring that the root cause is fully analyzed and resolved, and preventing the defect from recurring. For defects caused by transient, random occurrences, less analytical resources are required; the manufacturing execution system's marking and isolation process is directly triggered, greatly improving processing efficiency and avoiding wasting time on sporadic problems. This allows engineers to focus on more important systemic failures.

[0089] According to embodiments of the present invention, the cable surface defect detection method includes, in addition to the above-described, Figure 1 In addition to steps S110 to S150, the method may further include: if the cause of the defect is determined, determining an accuracy assessment value for estimating the accuracy level of the cause of the defect based on the number of causes of the defect and the duration of the abnormal process parameters corresponding to the causes of the defect; and determining a processing strategy that matches the accuracy assessment value so as to process the defective cable based on the processing strategy.

[0090] In this example, the number of defect causes and the reciprocal of the duration of the abnormal process parameter corresponding to each defect cause can be determined. Then, the reciprocals are weighted and summed to obtain the accuracy assessment value. It should be noted that here, an abnormal process parameter can be considered as one defect cause.

[0091] For example, the accuracy assessment value P can be obtained by calculating formula (3):

[0092] P=α*(1 / N)+β*(1 / L)(3)

[0093] Where N is the number of abnormal process parameters, L is the minimum, maximum or average time length of the abnormal process parameters, α and β are preset weighting coefficients, and α+β=1.

[0094] Indicatively, multiple positioning accuracy levels and their corresponding numerical ranges can be set. The calculated accuracy assessment value is matched with the preset numerical range to determine the accuracy level of the defect cause. Accuracy levels can include at least high, medium, and low accuracy. Different processing strategies can be set based on different accuracy levels.

[0095] For example, determining a processing strategy that matches the accuracy assessment value may include: if the accuracy assessment value is determined to be a first accuracy level, the processing strategy that matches the accuracy assessment value is to send a first type of instruction to the manufacturing execution system; if the accuracy assessment value is determined to be a second accuracy level, the processing strategy that matches the accuracy assessment value is to send a second type of instruction to the manufacturing execution system; and if the accuracy assessment value is determined to be a third accuracy level, the processing strategy that matches the accuracy assessment value is to send a third type of instruction to the manufacturing execution system.

[0096] The first precision level is higher than the second precision level, and the second precision level is higher than the third precision level. For example, the first precision level is high precision, the second precision level is medium precision, and the third precision level is low precision.

[0097] The first type of instruction can instruct the manufacturing execution system to issue process parameter adjustment commands to the production line control system and track and verify the effects of the adjustments.

[0098] The second type of instruction can instruct the manufacturing execution system to generate a repair work order containing the cause of the defect and the corresponding abnormal process parameters, and push it to the equipment maintenance terminal.

[0099] The third type of instruction can instruct the manufacturing execution system to trace the cause of the first occurrence of a defect.

[0100] This invention, when the cause of a defect is fully identified (e.g., high precision), bypasses manual approval through a first type of instruction, directly controlling the production line for adjustments. For process issues requiring rapid response, such as temperature and speed drift, this can quickly curb the continued generation of defects, prevent the expansion of scrap, and achieve second-level process self-healing. When the cause of a defect is clear but not fully identified (e.g., medium precision), a structured maintenance work order is generated through a second type of instruction, significantly shortening on-site troubleshooting time and improving the efficiency and success rate of maintenance work.

[0101] According to an embodiment of the present invention, the manufacturing execution system traces the cause of the first occurrence of a defect by performing the following operations: determining the temporal stability of the cause of the defect based on the abnormal duration and abnormal fluctuation frequency; if the temporal stability is determined to be greater than or equal to a preset temporal stability, corresponding processing is performed on the cause of the first occurrence of the defect; if the temporal stability is determined to be less than the preset temporal stability, the cause of the defect that occurs at the next time point is traced.

[0102] In this embodiment, the temporal stability can be determined as follows: the duration of the anomaly and the frequency of the anomaly fluctuations are normalized and assigned preset weights, and then a comprehensive temporal stability is obtained through a fusion method such as weighted summation or product. The duration of the anomaly is positively correlated with the temporal stability, while the frequency of the anomaly fluctuations is negatively correlated with the temporal stability.

[0103] The preset time-series stability level can be determined in the following ways: Process parameters from a large number of historical gold production batches can be retrieved from the Manufacturing Execution System (MES) database, and their normal fluctuation range under stable operating conditions can be analyzed. Simultaneously, historically confirmed and successful defect cause analysis cases can be analyzed. For historically proven abnormal process parameters that are the root cause, the duration and frequency of abnormal fluctuations before the incident can be determined. By comparing and analyzing the normal fluctuation range, the duration and frequency of abnormal fluctuations, key patterns that can effectively distinguish between stable anomalies and unstable noise can be identified. Based on these patterns, statistical methods, such as cluster analysis and percentile division, can be used to calculate one or more threshold intervals, ultimately comprehensively determining a preset time-series stability level value or score.

[0104] In this invention, by judging the timing stability, the system can proactively avoid invalid or inefficient operation cycles. If the processing flow is immediately started upon detecting an unstable anomaly, such as generating work orders or automatic adjustments, it may lead to a large number of invalid work orders based on false signals, or cause equipment wear and malfunction due to frequent adjustments. Furthermore, when the timing stability is determined to be less than the preset timing stability, tracing the cause of the defect at the next time node can be seen as a resource protection strategy. It suspends the currently uncertain analysis branch and reserves computing and execution resources for anomaly events with clearer and more certain patterns, thereby improving the overall system's operating efficiency and processing success rate.

[0105] Figure 4 A block diagram of a cable surface defect detection device according to an embodiment of the present invention is shown.

[0106] like Figure 4 As shown, the cable surface defect detection device 400 includes an identification module 410, a first determination module 420, an extraction module 430, a second determination module 440, and a third determination module 450.

[0107] The identification module 410 is used to identify defects in each image from a sequence of images of the same batch and the same roll of cable that are continuously collected within a preset time period, and to determine the physical location of the defects on the cable surface.

[0108] The first determining module 420 is used to determine the timestamp of the defect on the production line based on the traction speed and physical location of the cable on the corresponding production line.

[0109] The extraction module 430 is used to extract multiple process parameter values ​​to be analyzed from the production line within the estimated time interval that is expected to have an impact on defects, based on the generated timestamp, to obtain the process parameter sequence to be analyzed, with the generated timestamp located within the estimated time interval.

[0110] The second determining module 440 is used to determine the parameter deviation based on the difference between the process parameter sequence to be analyzed and the reference sequence. The reference sequence is a sequence consisting of multiple reference process parameter values ​​under a defect-free production state corresponding to the estimated time interval.

[0111] The third determining module 450 is used to determine the cause type of defects that lead to the formation of the cable surface based on the parameter deviation.

[0112] According to embodiments of the present invention, any plurality of modules among the identification module 410, the first determining module 420, the extraction module 430, the second determining module 440, and the third determining module 450 may be combined into one module, or any one of these modules may be split into multiple modules. Alternatively, at least a portion of the functionality of one or more of these modules may be combined with at least a portion of the functionality of other modules and implemented in one module. According to embodiments of the present invention, at least one of the identification module 410, the first determining module 420, the extraction module 430, the second determining module 440, and the third determining module 450 may be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any one of the three implementation methods or a suitable combination of any of them. Alternatively, at least one of the identification module 410, the first determination module 420, the extraction module 430, the second determination module 440, and the third determination module 450 may be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0113] According to an embodiment of the present invention, the cable surface defect detection device further includes: a fourth determining module, a fifth determining module, a sixth determining module, and a seventh determining module. The fourth determining module is used to determine the temporal distribution density of the multiple defects in the time dimension based on their respective timestamps on the production line, when it is determined that multiple defects exist on the cable surface and the defects have the same defect appearance category. The fifth determining module is used to determine the spatial distribution density of the multiple defects in the spatial dimension based on their respective physical locations on the cable surface. The sixth determining module is used to determine the spatiotemporal clustering degree of the defects based on the temporal distribution density and the spatial distribution density. The seventh determining module is used to determine the cause type leading to the formation of the multiple defects on the cable surface based on the parameter deviation and the spatiotemporal clustering degree of the defects.

[0114] According to an embodiment of the present invention, the cable surface defect detection device further includes a feature extraction module and a classification module. The feature extraction module is used to extract the appearance features of defective regions in each image of an image sequence from multiple visual dimensions. The defect is determined based on a difference between the image and a standard image in at least one of the following: image sharpness, brightness, and color difference. The classification module is used to classify the appearance features corresponding to multiple defects when it is determined that multiple defects exist on the cable surface, thereby obtaining defect appearance categories for the multiple defects.

[0115] According to an embodiment of the present invention, the cable surface defect detection device further includes an abnormal parameter identification module and a triggering module. The abnormal parameter identification module is used to identify abnormal process parameters from the sequence of process parameters to be analyzed, and to determine the abnormal process parameters as the cause of the defect, when the cause type is determined to be a systemic process anomaly. The triggering module is used to trigger the manufacturing execution system's marking and isolation process when the cause type is determined to be an instantaneous random occurrence.

[0116] According to an embodiment of the present invention, the cable surface defect detection device further includes an eighth determining module and a ninth determining module. The eighth determining module is used to determine, upon determining the cause of the defect, an accuracy assessment value for estimating the accuracy level of the defect cause based on the number of defect causes and the duration of the abnormal process parameters corresponding to the defect causes. The ninth determining module is used to determine a processing strategy matching the accuracy assessment value, so as to process the defective cable based on the processing strategy.

[0117] It should be noted that the cable surface defect detection device part in the embodiments of the present invention corresponds to the cable surface defect detection method part in the embodiments of the present invention. For a detailed description of the cable surface defect detection device part, please refer to the cable surface defect detection method part, which will not be repeated here.

[0118] Those skilled in the art will understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention can be combined and / or combined in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or combinations fall within the scope of the present invention.

[0119] The embodiments of the present invention have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of the invention. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of the invention, and all such substitutions and modifications should fall within the scope of the invention.

Claims

1. A method for detecting surface defects in cables, characterized in that, The method includes: From a sequence of images of the same batch and roll of cable surface collected continuously within a preset time period, defects in each image are identified, and the physical location of the defects on the cable surface is determined. Based on the traction speed of the cable on the production line and the physical location, determine the timestamp of the defect's occurrence on the production line; Based on the generated timestamp, multiple process parameter values ​​to be analyzed within the estimated time interval that is expected to affect the defect are extracted from the production line to obtain a sequence of process parameters to be analyzed, wherein the generated timestamp is located within the estimated time interval. Based on the differences between the process parameter sequence to be analyzed and the baseline sequence, the parameter deviation is determined. The baseline sequence is a sequence consisting of multiple baseline process parameter values ​​under a defect-free production state, corresponding to the estimated time interval. If it is determined that there are multiple defects on the surface of the cable, and the defects have the same defect appearance category, the temporal distribution density of the multiple defects in the time dimension is determined based on the timestamp of each of the multiple defects on the production line. Based on the physical location of each of the defects on the cable surface, the spatial distribution density of the defects in the spatial dimension is determined. Based on the temporal distribution density and the spatial distribution density, the spatiotemporal clustering degree of the defect is determined; Based on the parameter deviation and the spatiotemporal clustering of the defects, the causal types that cause the formation of multiple defects on the surface of the cable are determined.

2. The method according to claim 1, characterized in that, The step of determining the temporal distribution density of the multiple defects in the time dimension based on the timestamps of their respective occurrences on the production line includes: Based on the timestamps of the generation of each of the multiple defects on the production line, the time interval between the generation of adjacent defects is determined; Based on the variance of the generation time interval, the temporal clustering of multiple defects is quantified and determined as the temporal distribution density.

3. The method according to claim 1, characterized in that, The step of determining the spatial distribution density of the multiple defects in a spatial dimension based on their respective physical locations on the cable surface includes: Based on the physical location of each of the defects on the cable surface, a clustering algorithm is used to cluster the defects and identify at least one defect cluster. The cable length range covered by each defect cluster is taken as the cluster range, and the ratio of the number of defects in the cluster to the corresponding cluster range is determined as the spatial distribution density.

4. The method according to claim 1, characterized in that, The determination of the causal types leading to the formation of multiple defects on the cable surface based on the parameter deviation and the spatiotemporal clustering degree of the defects includes: The parameter deviation and the defect spatiotemporal clustering are normalized respectively to obtain normalized parameter deviation and normalized defect spatiotemporal clustering. The defect type tendency value is obtained by weighted summing the normalized parameter deviation and the normalized defect spatiotemporal clustering. If the defect type tendency value is determined to be greater than or equal to a predetermined threshold, the cause type is determined to be a systemic process anomaly. If the defect type tendency value is determined to be less than the predetermined threshold, the cause type is determined to be transient random triggering type.

5. The method according to claim 1, characterized in that, The method further includes: The appearance features of defective regions in each image of the image sequence are extracted from multiple visual dimensions. The defects are determined based on differences between the image and a standard image in at least one of the following: image sharpness, brightness, and color difference. If it is determined that there are multiple defects on the surface of the cable, the appearance features corresponding to the multiple defects are classified to obtain the defect appearance categories of the multiple defects.

6. The method according to claim 1, characterized in that, The method further includes: If the cause type is determined to be a systematic process anomaly, abnormal process parameters are identified from the sequence of process parameters to be analyzed, and the abnormal process parameters are determined as the cause of the defect. If the cause type is determined to be transient random triggering type, the manufacturing execution system is triggered to mark and isolate the process.

7. The method according to claim 6, characterized in that, The method further includes: Once the cause of the defect is determined, an accuracy assessment value for estimating the accuracy level of the cause of the defect is determined based on the number of the causes of the defect and the duration of the abnormal process parameters corresponding to the causes of the defect. A processing strategy matching the accuracy assessment value is determined so that defective cables can be processed based on the processing strategy.

8. The method according to claim 7, characterized in that, The processing strategy for determining the accuracy assessment value includes: If the accuracy assessment value is determined to be the first accuracy level, the processing strategy matching the accuracy assessment value is to send a first type of instruction to the manufacturing execution system. The first type of instruction instructs the manufacturing execution system to issue a process parameter adjustment command to the production line control system and to track and verify the effect of the adjustment. If the accuracy assessment value is determined to be the second accuracy level, the processing strategy matching the accuracy assessment value is to send a second type of instruction to the manufacturing execution system. The second type of instruction instructs the manufacturing execution system to generate a repair work order containing the cause of the defect and the corresponding abnormal process parameters, and push it to the equipment maintenance terminal. If the accuracy assessment value is determined to be the third accuracy level, the processing strategy matching the accuracy assessment value is to send a third type of instruction to the manufacturing execution system, which instructs the manufacturing execution system to trace the cause of the first occurrence of the defect. Wherein, the first accuracy level is greater than the second accuracy level, and the second accuracy level is greater than the third accuracy level.

9. The method according to claim 7, characterized in that, The manufacturing execution system traces the cause of the first occurrence of a defect by performing the following operations: The temporal stability of the defect cause is determined based on the abnormal duration and abnormal fluctuation frequency of the defect cause. If the timing stability is determined to be greater than or equal to the preset timing stability, the cause of the first occurrence of the defect is handled accordingly. If the timing stability is determined to be less than the preset timing stability, the cause of the defect at the next time point is traced back.

Images