FPC reinforcing and bonding quality detection method and system

Abstract

The present application relates to the technical field of FPC bonding detection, in particular to an FPC reinforcing bonding quality detection method and system, comprising: acquiring a layout area image, extracting a defect feature vector according to a difference image after multiple superimpositions of the layout area image; cross-solution of the defect feature vector is performed, the intersection of each defect feature vector is determined as a target defect area, and the hierarchical order of each target defect area is determined according to the priority order of each target defect area; according to the establishment of the hierarchical order of each target defect area, the bonding judgment result of each target defect area is extracted in turn; based on the bonding judgment result of each target defect area, the distribution points of defects after bonding are mapped, and the defect set during production is determined; the automation of FPC bonding detection is realized, and the accuracy and efficiency of defect positioning are improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of FPC bonding detection, and in particular to an FPC reinforcing bonding quality detection method and system. BACKGROUND

[0002] With the increasing miniaturization of electronic products, flexible circuit (FPC) boards appear, and FPC board reinforcing can have different reinforcing methods according to different material types, such as polyimide reinforcing and polyester reinforcing. The FPC bonding condition is usually detected by checking the offset and thickness of the bonding, and whether the reinforcing material bonded is qualified is identified.

[0003] For example, Chinese Patent Application Publication No. CN112649446A discloses a detection method, bonding method and device for FPC, which is applied to a bonding device equipped with a camera. The method comprises: controlling the camera to capture images when the soft material to be detected moves to two preset positions, respectively, to obtain two first images; obtaining first detection information from the two first images, wherein the first detection information includes a first contour image, a first angle parameter, a first length parameter and a first gray scale parameter; according to the preset reference information and the first detection information, material defect detection is performed on the soft material to be detected to obtain the detection result of the soft material to be detected.

[0004] For example, Chinese Patent Application Publication No. CN111879777A discloses a soft material bonding defect detection method, device, equipment and storage medium. The method comprises: collecting an FPC image to be detected; performing template matching on the FPC image to be detected to obtain a matching area matched with a preset template; according to the position information of the matching area and the position relationship between the matching area and the target area, extracting a target area image from the FPC image to be detected; and judging whether the soft material is missing by judging whether the soft material exists in the target area image.

[0005] In the prior art, the alignment angle of the gray scale image is identified, the identified angle, length and gray scale are compared with the contour, and it is determined whether the reinforcing material bonded has deviation. However, the prior art is inclined to angle alignment and the length after angle alignment, and these alignment comparison methods are inclined to cold pressing in FPC bonding. When hot pressing is used to bond the reinforcing material to the FPC, the original structure contour of the reinforcing material is difficult to be used as a standard for subsequent comparison of contour deviation, and the continuity fluctuation and relative deviation aggregation of defects need to be identified to determine whether the position of the reinforcing material after bonding is qualified. SUMMARY

[0006] In order to solve the above technical problems, the technical scheme adopted by the present application is: a FPC reinforcing and laminating quality detection method, comprising: S1, acquiring a laying area image in response to the laying position of reinforcing material on the FPC.

[0007] S2, extracting a defect feature vector according to a difference image after multiple superimpositions of the laying area image, and taking the corresponding position of the difference image as a guide.

[0008] S3, cross-solution of the defect feature vector is performed, the intersection of each defect feature vector is determined as a target defect area, and the hierarchical order of each target defect area is determined according to the priority order represented by each target defect area.

[0009] S4, when the hierarchical order of each target defect area is established, the deviation data of each target defect area is located according to the corresponding contour deviation and thickness of each target defect area, and the lamination judgment result of each target defect area is set using the deviation data.

[0010] S5, when the hierarchical order of each target defect area is not established, the trend of each target defect area is analyzed in isolation, and the lamination judgment result of each target defect area is set.

[0011] S6, based on the lamination judgment result of each target defect area, the distribution points of defects after lamination are mapped, and the defect set during production is determined.

[0012] A FPC reinforcing and laminating quality detection system, comprising: an image acquisition module for acquiring a laying area image according to the laying position of reinforcing material on the FPC.

[0013] A feature extraction module for extracting a defect feature vector according to a difference image after multiple superimpositions of the laying area image, and taking the corresponding position of the difference image as a guide.

[0014] A hierarchical judgment module for cross-solution of the defect feature vector, determining the intersection of each defect feature vector as a target defect area, and determining the hierarchical order of each target defect area according to the priority order represented by each target defect area.

[0015] A lamination judgment module for determining the hierarchical order of each target defect area according to the hierarchical order of each target defect area, when the hierarchical order of each target defect area is established, locating the deviation data of each target defect area according to the corresponding contour deviation and thickness of each target defect area, and setting the lamination judgment result of each target defect area using the deviation data; when the hierarchical order of each target defect area is not established, analyzing the trend of each target defect area in isolation, and setting the lamination judgment result of each target defect area.

[0016] The defect output module is used for mapping the distribution points of defects after lamination according to the lamination judgment result of each target defect area, and determining the defect set during production.

[0017] The present application has the following advantages: 1. The present application extracts a defect feature vector from a multi-angle gray scale histogram and a differential image collected continuously, combines defect edge trend density analysis and area deviation calculation, describes the defect area recognized in the current image in the form of a defect point, and further maps the temperature and stress under the corresponding defect in combination with the temperature and stress when the defect area is laminated with the FPC, so as to avoid the sensitivity of single data in identifying defects, enhance the integrated processing of the lamination process, and prevent the defect from being missed.

[0018] 2. The present application extracts a target defect area through the intersection of a defect feature vector, and introduces the hierarchical order dependence of the target defect area combination in combination with state transition chain and process parameter correlation mapping, combines the distribution of the defect area on the FPC, describes the distribution of the target defect area on the neighborhood position, and the combination of multiple target defect areas, so as to adapt to the defect area in the neighborhood analysis mode in multiple scenarios.

[0019] 3. The present application summarizes the data of multiple target defect areas under overall fluctuation and extreme deviation by meeting the hierarchical nature of the hierarchical order dependence and the isolated area not meeting the hierarchical order, describes the distribution of the current hierarchical and isolated processing, maps the lamination judgment result calculated to the specific position on the FPC by using spatial distribution deviation calculation, outputs the current maximum distribution deviation, realizes the rapid response to abnormal defects during FPC lamination, and realizes the quantitative positioning of response processing, completes the defect set on the FPC after the current lamination judgment and position tracing, and improves the positioning effect of different defect types. BRIEF DESCRIPTION OF DRAWINGS

[0020] The present application will be further described below in combination with the drawings and embodiments.

[0021] Figure 1 It is a flowchart of a FPC reinforcing lamination quality detection method.

[0022] Figure 2 It is a structural schematic diagram of a FPC lamination in a FPC reinforcing lamination quality detection method.

[0023] Figure 3 It is a flowchart of step S2 of a FPC reinforcing lamination quality detection method.

[0024] Figure 4 It is a flowchart of step S3 of a FPC reinforcing lamination quality detection method.

[0025] Figure 5 is a flowchart of step S4 of the FPC reinforcing and laminating quality detection method.

[0026] Figure 6 is a system framework diagram of the FPC reinforcing and laminating quality detection system. DETAILED DESCRIPTION

[0027] Embodiments of the present application are described in detail below. The embodiments described below are exemplary only, and are not to be construed as limiting the present application. Unless otherwise noted, technical or conditions not specified in the embodiments are performed according to the techniques or conditions described in the literature in the art or according to the product instructions.

[0028] Reference Figure 1 An FPC reinforcing and laminating quality detection method includes: S1, in response to the layout position of the reinforcing material on the FPC, obtaining a layout area image.

[0029] S2, according to the difference image after multiple superimpositions of the layout area image, taking the corresponding position of the difference image as a guide, extracting a defect feature vector.

[0030] S3, cross-solution of the defect feature vector, determining the intersection of each defect feature vector as a target defect area, and determining the hierarchical order of each target defect area according to the priority order represented by each target defect area.

[0031] S4, when the hierarchical order of each target defect area is established, positioning the deviation data of each target defect area according to the corresponding contour deviation and thickness of each target defect area, and setting the lamination judgment result of each target defect area using the deviation data.

[0032] S5, when the hierarchical order of each target defect area is not established, performing isolated analysis on the trend of each target defect area, and setting the lamination judgment result of each target defect area.

[0033] S6, based on the lamination judgment result of each target defect area, mapping the distribution points of the defects after lamination, and determining the defect set during production.

[0034] In the present application, FPC refers to a flexible circuit board, which is made of polyimide or polyester film as a substrate, and is a highly reliable and excellent flexible printed circuit board. In general, in order to improve the deformation resistance and pressure resistance of the circuit board, reinforcing materials such as reinforcing film are used to reinforce the areas to be reinforced on the circuit board in the form of reinforcing film, so as to improve the stability and service life of the circuit board.

[0035] The layout region image described above is used to illustrate the position of the reinforcing material after being attached to the FPC. The relevant image of this part of the position is collected to identify the attachment of the windowed region of the reinforcing material and the reinforcing region on the FPC. After obtaining the corresponding attachment, the defects such as the deviation, the contrast angle, and the bubble can be identified to determine the relative effect of the reinforcing material after being attached.

[0036] It should be noted that the collected layout region image can be obtained in multiple angles by the image collection device, and the image can be processed by gray scale to form multiple images about the current reinforcing material attachment process.

[0037] The implementation mode of step S1 includes: S11, taking the region of the reinforcing material attached to the FPC as an attachment point based on the continuous image data of the layout position, and obtaining an attachment point array.

[0038] S12, obtaining the gray scale histogram corresponding to the attachment point array based on the change of the shooting angle of the attachment point array in the continuous time window, and taking the gray scale histogram in multiple angles as the output layout region image.

[0039] At this time, the relative angle of the layout position is used to perform the contour scanning on each layout position. The attachment of the windowed region of the reinforcing material on each layout position and the reinforcing region on the FPC is identified. The attached position is regarded as an attachment point. Then, the gray scale histogram formed on the attachment position is continuously viewed. The gray scale histogram in multiple angles is taken as the output layout region image.

[0040] The high-resolution industrial camera such as the CCD or CMOS camera and the special light source such as the ring light source, the coaxial light source, or the laser light source can be used for image collection. The laser is obliquely shot to the surface of the reinforcing material. The displacement of the reflected light spot is captured by the CCD camera to calculate the surface topography and the attachment state. Then, the data of the FPC attached to the reinforcing material is described based on the images collected in different angles. The Canny, Sobel, or other operators are used to detect the boundary of the reinforcing material and the FPC to obtain the image of the relevant defects of the reinforcing material after being attached.

[0041] Preferably, in order to cover the image of the FPC attached in the attachment process, the continuous time window can be set to 2-3 seconds to ensure that the collected image can cover the FPC attachment process in the current continuous time window.

[0042] For example, Figure 2As shown, the FPC is fixed and reinforced using the reinforcing window layer, the discharge layer, the PIN pin fixing layer and the base layer. At this time, the base and the discharge layer are made with a 4.1mm hole, then the PIN pin fixing layer is made with a 2.1mm hole, and the reinforcing window layer is made with a 2.1mm hole and a single-sided pre-release 0.05mm or less than 0.05mm reinforcing window. Then, the reinforcing window layer, the discharge layer, the PIN pin fixing layer and the base layer are sequentially bound into a jig using the PIN pin. After the FPC is sleeved into the discharge layer, the reinforcing window layer is closed, the reinforcing material is attached to the window position of the reinforcing window layer, then the discharge layer is lifted to take out the FPC, to complete the attachment of the reinforcing material. At this time, the obtained gray scale histogram is based on the image of the FPC in the area to be reinforced in the current attachment process, and is obtained in multiple angles such as vertical, horizontal and diagonal. Multiple attachment points are set according to the actual attachment window position. After the images extracted at these attachment points are mapped using the attachment point array, the current processing gray scale histogram is composed.

[0043] In an embodiment of the present application, during the processing of step S2, the gray scale histogram obtained on the layout area needs to be compared with the standard attachment image in the form of multiple acquisitions, to identify the differential images after superimposing multiple images, which represent the defects that can exist during attachment. The standard attachment image is a standard image set in advance according to the current FPC, which is used to assist in identifying the defects in the differential image and labeling the position and size of the corresponding defects.

[0044] It should be noted that when the current multiple images are superimposed, the pixel-level gray difference between the three continuously collected layout area images and the standard image is performed, and the absolute value is taken and superimposed to highlight the stable defect area.

[0045] As shown in Figure 3 The implementation of step S2 also includes: S21, analyzing the differential image according to the distribution of the reinforcing material on the FPC to obtain defect edge direction data.

[0046] S22, extracting multiple defect regions corresponding to the defect direction density according to the density analysis method of the defect edge direction data.

[0047] S23, calculating the area deviation of the defect region according to the position and size of each defect region, and regarding the calculated absolute area deviation as a defect feature vector.

[0048] Preferably, the above-mentioned defect edge direction data is used to obtain a plurality of edge points containing defects from the differential image, and the positions of the edge points are discretized into eight directions, such as north, northeast, northwest, east, southeast, southwest, west, and south. According to the number of edge points in each direction, the direction with the largest number of edge points is regarded as the current direction distribution, and the data of the plurality of points contained in the direction distribution is taken as the defect edge direction data. These obtained data can indicate the size of the surface profile and the side thickness difference of the differential image, so as to determine whether the differential image has image deviation in the profile aspect and deviation in the thickness aspect.

[0049] As for the process of performing density analysis, the number of edge points in each direction is divided by the value of the length of the corresponding direction pixels to indicate the number of edge points per pixel length. This value is taken as the defect direction density, and the corresponding area is taken as the current analyzed defect area. When obtaining the defect area, the edge points in each direction are clustered, and the minimum circumscribed matrix obtained by clustering is taken as the corresponding defect area. The clustering method can adopt the K-means clustering method. When a new edge point is introduced, the average value of the clustering cluster does not change as the stop condition, to obtain a plurality of clustering clusters about the edge points, and the minimum circumscribed matrix corresponding to the corresponding clustering cluster is taken as the output defect area.

[0050] When calculating the area deviation of the defect area, the implementation of step S23 also includes: S231, performing surface state mapping with the current input defect area, and extracting temperature limit values and stress limit values in different directions. At this time, the temperature and stress limit values generated by the defect are mapped in combination with the surface-identified defect. When quantifying the FPC reinforcement, whether there is stress or temperature abnormality, which causes partial areas to appear defects. At this time, temperature sensors and force sensors can be set during reinforcement to view the state of the current FPC during the reinforcement process.

[0051] The temperature peak and valley values and the stress peak and valley values on the defect area are extracted, and the direction of the current defect area is combined to indicate the reinforced area on the FPC.

[0052] S232, radius match the temperature limit value and the stress limit value with the current defect area to determine the radius matching data of the defect area and the temperature and stress; when radius matching, set a radius with the geometric center of the defect area as the center, search the limit value of stress and temperature within the radius, since the defect area in different directions is mapped with the temperature limit value and the stress limit value in step S231, the radius set at this time is just to include the corresponding temperature limit value and the stress limit value, that is, radius matching based on the minimum circumscribed circle of the current defect area; at this time, temperature matching and stress matching will be performed respectively, and the temperature standard deviation and the stress standard deviation in the area contained in the current radius are compared with the preset temperature standard deviation and the preset stress standard deviation in turn, if greater, it indicates that there is an abnormality about temperature or stress in the corresponding area, the preset temperature standard deviation and the preset stress standard deviation will be set based on the average of the standard deviation of stress and temperature when FPC is normally reinforced, and the part with temperature abnormality or stress abnormality at this time is marked, and the corresponding temperature, stress data is combined with the mark as the radius matching data output.

[0053] S233, based on the correlation between the radius matching data and the defect direction density, the area deviation corresponding to the defect area is sequentially calculated.

[0054] Preferably, when the defect feature vector is obtained, the current processed differential image can be further obtained by using surface state mapping to obtain the defect feature vector related temperature, stress and other data; at this time, the surface state mapping is biased to obtain the temperature and stress existing in the corresponding defect area of the reinforcing material in the FPC bonding process, and this part of data represents the data collected on each area in the bonding process, these values can use thermal imaging, force sensor, vibration sensor and other data to describe the reinforcing material bonding FPC process, to realize multi-dimensional surface state mapping.

[0055] Preferably, when the area deviation is calculated subsequently, the part data with temperature and stress abnormality needs to be recorded, the radius matching data is associated with the defect direction density in the current direction, and the recognized temperature value, stress value and defect direction density are standardized to eliminate their dimensions, and the deviations of the three kinds of data are calculated, and the weighted sum of the deviations is described.

[0056] Wherein, the deviation of temperature is used to indicate the difference between the temperature standard deviation and the preset temperature standard deviation, to indicate the process under the fixed FPC bonding, the preset temperature standard deviation is calculated according to the corresponding temperature value in the historical data; the deviation of stress is the same as the deviation of temperature, and the defect trend density is directly weighted. The defect trend density represents the defect situation existing in the current process, and the weights of temperature and stress are set according to the ratio of the frequency of abnormality of temperature and stress in the current defect area to the frequency of abnormality of all areas. The weight of defect trend density can be set according to the ratio of the current defect trend density to the average value of defect trend density in the historical data, to indicate the form of the current defect relative to the average situation of the historical data.

[0057] It should be noted that the processing of mapping temperature and stress is biased to indicate that the FPC bonding reinforcing material is completed based on the hot pressing process at this time. At this time, the reinforcing material can adopt hot pressing process, for example, when the reinforcing material is 0.1mm and 0.2mm, because the thickness is thin, ordinary fast pressing machine hot pressing can be adopted, or cold pressing can be directly realized. When using cold pressing processing, the current extracted defect feature vector will record the pixel points with defects in the form of feature matrix on the difference image; That is, in the specific processing process of step S23, the defect area is directly matched with radius, and after checking the pixel points within the current radius, the difference between these points and the pixel points of the standard image is output as the area deviation of the region.

[0058] The finally output defect feature vector represents the position, angle and size of the defect, and the combined content of these data is displayed in the form of vector to indicate the form of current defect processing.

[0059] It should be noted that when the defect area of the layout area image is processed, a plurality of layout area images shot by visible light, a thermal imaging image describing the glue covering area of the reinforcing material, and a three-dimensional point cloud image scanned by the current FPC using an industrial camera will be processed in turn to obtain a plurality of defect feature vectors of the defect area. These defect feature vectors will record the position, angle and size of the recognized content, to facilitate finding the intersection part that can be recognized on the current FPC after comparing a plurality of groups of data, and taking the content of the intersection part as the target defect area.

[0060] In the above processing, the temperature is generally kept stable during the normal heat pressing process, and the deviation of the temperature is added to prevent the current equipment from working abnormally, to prevent the reduction of data recognition accuracy due to single profile deviation recognition, and to change the different settings of the FPC to be reinforced. At this time, multiple stress data are checked to prevent some materials from having defects, which may cause misadjustment of the process parameters during the subsequent adjustment of the heat pressing process. Finally, the density of the recognized defects in each direction is associated to further describe whether the different temperatures and stresses of the current heat pressing setting cause the difference and appearance of the defect distribution density, thereby completing the combination of defects and process parameters in multiple batches of FPC bonding recognition.

[0061] In one embodiment of the present application, as shown in Figure 4 The implementation of step S3 includes: S31, mapping each group of defect feature vectors to the same space, and regarding the intersection of the defect feature vectors in the same space as the target defect area; at this time, multiple defect area intersections obtained by analyzing multiple images are regarded as the current target defect area for multiple sources or for the current FPC bonding site, to prevent the problem of abnormal area recognition caused by single image analysis.

[0062] If the current defect feature vector cannot directly obtain the relevant intersection in the physical space, the feature similarity is calculated based on the value of the current defect feature vector, such as the cosine similarity calculation method. When the cosine similarity is greater than the preset similarity threshold, the defect feature vector corresponding to the similarity is regarded as the target defect area at this time. At this time, the intersection is obtained by calculating the feature similarity, and this intersection represents the potential associated intersection in addition to the direct intersection. At this time, the cosine similarity value can be based on the average value of the cosine similarity in the potential associated intersection scenario in the historical data as the preset similarity, to obtain part of the data that does not exist directly associated, to improve the accuracy of subsequent analysis of the target defect area.

[0063] S32, draw the state transition chain of the target defect region using the size of the target defect region at each cross-solution, the state transition chain of the target defect region indicates that the defect feature vectors obtained using multiple groups of images are mapped to the same spatial coordinate system, and the target defect region is obtained by cross-comparison in two groups each time. At this time, the size of the target defect region after comparing the two images may be different in size. At this time, the size of the area obtained by cross-solution each time is used to calculate the probability value of the current target defect region under Gaussian distribution when the size of the current target defect region conforms to Gaussian distribution, and the conditional probability corresponding to the state transition is calculated according to the change of the region size after each solution, that is, the probability value of the corresponding target defect region size obtained by the next solution after the current target defect region size is satisfied.

[0064] S33, after the state transition chain of the target defect region is associated and mapped with the data representing the target defect region, the priority order is set according to the position of the target defect region belonging to the FPC.

[0065] At this time, the main purpose of the association and mapping is to associate the current target defect region with the image and the value of the defect feature vector, and to set the priority of the current target defect region in the form of label description of the position of the flexible circuit board. For example, the number of times of identifying abnormal circuit components of the current FPC representing the flexible circuit board accounts for the total number of abnormalities, and the priority is set so that the circuit component directly corresponding to the current target defect region is preferentially processed in the part where the abnormality occurs more frequently.

[0066] As for the hierarchical order dependence, the highest priority target defect region is expanded based on its spatial position, and the hierarchical order dependence is set in sequence according to the dependence of each target defect region in space.

[0067] S34, the highest priority target defect region is used as the initial level, and the target defect region closest to the initial level is used as the secondary level, and the hierarchical order dependence of each target defect region is gradually divided. At this time, the defect regions are layered according to the priority and spatial proximity to form a clear defect path.

[0068] At this time, the image can be directly represented as the image before and after the lamination. For example, the lamination position offset, reinforcing material bubbles, etc. are captured in real time by high-precision cameras before lamination, and the overflow, indentation, etc. are detected by AOI after lamination, and the lamination is detected by three-dimensional point cloud to form a three-dimensional point cloud image. The corresponding defects on the current FPC are viewed to see if the corresponding defects are in the process parameters corresponding to the lamination pressure and temperature, and the relative relationship of the corresponding defect forms, and the forms of these defect changes are described using the state transition chain.

[0069] The implementation manner of the state transition chain in the step S33 includes: associating the defect state with the process parameters to form a mapping relationship, such as the combination of the detected defect form, temperature, stress, pressure, and the content of the single defect region on the same position under multiple pictures, in which the defect form is transformed into other defect forms, and the actual hot-pressing process data such as the bonding pressure and temperature.

[0070] The state transition probability value corresponding to the defect state is counted by using a sliding time window, and the priority weight is extracted from the position of the current target defect region according to the confidence degree of the state transition probability and the process parameter, and the moving average value of the priority weight is used as the priority order of the current target defect region. The extracted priority weight is set with a position weight according to the key functional area and the edge area on the FPC board, and then the two values are weighted and summed or are used as a retrieval condition to query the most relevant priority weight from the database, so as to indicate the priority of the FPC to be set in the current time period. The moving average value reduces short-term fluctuations and stabilizes the priority order.

[0071] Preferably, when the hierarchical order dependency is formed in the step S34, the termination condition of the hierarchical order dependency also needs to be determined, for example, five as the termination point to prevent infinite recursion, and five levels as the set level can reduce the complexity of the subsequent analysis of the bonding condition under normal analysis. The potential relationship between the bonding conditions is set while the hierarchical structure association is reserved; if the remaining area exists independently, the isolated analysis is not used to describe the hierarchical order dependency, and if the circular dependency is recognized under the hierarchical order dependency, the related data is marked, and the area with the smallest target defect region area in the nearest area is selected as the direct termination condition to obtain the hierarchical order dependency between the multiple target defect regions.

[0072] After the hierarchical order dependency is met, that is, the current data belongs to the data connected by the hierarchical order dependency, the bonding judgment result of the current target defect region is mainly based on the defects on the profile bias and the thickness.

[0073] In an embodiment of the present application, as shown in Figure 5As shown, the implementation of step S4 further includes: S41, according to the target defect area obtained, extracting the deviation of the profile deviation and the thickness of all hierarchical sequence dependencies, at this time, the deviation of the profile deviation and the thickness is obtained by extracting the image of the thickness difference in the differential image, and the image of the planar profile difference, and the pixel point deviation identified in these images is taken as the deviation of the profile deviation and the thickness at this time. The profile deviation represents the deviation existing when the profile deviation occurs in the target defect area, and the deviation of the thickness is divided in order to further quantify the deviation of the FPC bonding on the plane and the side surface.

[0074] S42, the number of target areas corresponding to the maximum deviation of the profile deviation and the thickness is set as the maximum iteration number, at this time, the deviation of the profile deviation and the thickness is checked from all target defect areas, when there are multiple target defect areas that can reach the maximum thickness deviation, the iteration number under each hierarchical sequence dependency is set, and then the deviation existing in each area in multiple iterations is judged to determine the form of the deviation existing on the corresponding area after the current hierarchical dependency is constructed.

[0075] S43, the bonding judgment result of the target defect area is set according to the fitness of each target defect area under the maximum iteration number.

[0076] Preferably, when the maximum iteration number is set, the target defect area that can reach the maximum deviation is set, for example, there is only one target defect area that is the maximum deviation, at this time, the deviation of the profile deviation and the thickness exists in one area, or there are two target defect areas, the deviation of the profile deviation and the thickness exists in two areas, at this time, the basic iteration number is selected, the basic iteration number is the average of the iteration number of data analysis in the historical data, if the area corresponding to the maximum deviation is more than two, the maximum iteration number is directly obtained from the data recorded in the database; at this time, the calculation method of particle swarm optimization is adopted, the deviation of the profile deviation and the thickness is normalized, and the pixel point value producing the deviation is counted, for example, the difference between the pixel point and the standard image pixel point is counted, the profile deviation of the identified reinforcement area under the profile deviation is counted, and the total profile deviation of the reinforcement area is counted, as for the deviation on the thickness, the deviation value of the pixel point on the corresponding area is counted in the image recognition, the two values are respectively calculated by the mean square error according to the profile deviation and the thickness deviation of all target defect areas under the current hierarchical dependency, then the sum of the two mean square errors is added to the minimum as the termination iteration condition, otherwise the minimum mean square error added under the current maximum iteration number is obtained.

[0077] Then the minimum mean square error is divided by the number of pixels with deviation in the target defect area under the current hierarchical order dependency, so as to obtain the fitness value of each target defect area under the current hierarchical order dependency. The value is based on the calculation method of SE standard error, and the standard error is divided by the square root of the number of pixels to describe the deviation fluctuation of the current target defect area, to describe the fluctuation under the fitting quality detection using multiple samples, and to emphasize the relative situation of the deviation in multiple positions under the fitting quality problem.

[0078] It should be noted that the above-mentioned deviation is the difference value calculated by comparing the current acquired image with the standard image, to indicate the difference value of the corresponding pixel points.

[0079] As for the isolated area that does not meet the hierarchical order dependency, the trend of the target defect area is analyzed, that is, the implementation mode of step S5 further includes: obtaining the maximum deviation and the minimum deviation of the target defect area in the same trend, and regarding the distance between the maximum deviation and the minimum deviation as the output fitting judgment result.

[0080] In step S5, the fitting judgment result will be output according to the distance between the regions representing the maximum deviation and the minimum deviation in the current target defect area; at this time, the concentration of isolated target defect areas in the same trend will be emphasized, and the distance value will be calculated based on the Euclidean distance of the pixel points, emphasizing the size of the deviation of a single pixel point; at this time, the deviation calculation method will be based on the distance value between each pixel point with deviation and the corresponding pixel point on the standard template, and then the pixel point representing the maximum deviation and the pixel point representing the minimum deviation will be found; at this time, the processing method will cover the contour deviation and thickness in the hierarchical dependency, and will focus on the specific pixel points with deviation in the fitting process after correlating and mapping the parts related to temperature and stress, and will quantify the extreme deviation risk of each target defect area.

[0081] In an embodiment of the present application, in step S6, the extreme deviation and the deviation fluctuation of the target defect area are combined, and the recognized data is further mapped to the production process to explain the fitting quality during production. The implementation mode of step S6 includes: extracting the index value in the fitting judgment result, normalizing the index value according to the acquisition method, and converting the fitting judgment result into a target variable; at this time, the values of the target defect areas with hierarchical order dependency and not hierarchical order dependency are normalized according to the output fitting judgment result.

[0082] With the distribution points of the target variable on the FPC, the distribution points are extracted according to the specification information of the FPC to obtain the distribution frequency at the corresponding position. At this time, the content calculated in the fitting judgment result, that is, the specific position of the target variable mapped to the FPC, such as the connector pin, the wiring path and the like, is set according to the number of times of mapping to the corresponding position of multiple fitting judgment results, and the distribution frequency is set.

[0083] Based on the obtained distribution frequency, the spatial solution of the distribution points of the target variable is performed, the distribution deviation of the distribution points compared with the overall distribution is calculated, and the data corresponding to the maximum distribution deviation is regarded as the output defect set.

[0084] When performing spatial solution, the specific position of the distribution points directly mapped to the FPC is needed, and the difference between the distribution frequency and the overall distribution frequency at each specific position is used to obtain the distribution deviation compared with the overall distribution. Then, the part with the maximum distribution deviation is taken as the output defect set.

[0085] The output defect set is used to quickly locate the most serious defect area on the FPC, find out the distribution of the defect area under the current process parameters, and combine these data in the form of temperature, stress and specific deviation for output, so as to prevent errors caused by single parameter fluctuation and the binding relationship between specific deviation and process parameters, facilitate finding out the main problem form under the current process parameters and FPC fitting condition, and extract data basis for adjusting FPC fitting production for workers.

[0086] As shown in Figure 6 The application also provides an FPC reinforcing fitting quality detection system, which comprises an image acquisition module, a feature extraction module, a hierarchical judgment module, a fitting judgment module and a defect output module. The output end of the image acquisition module is connected with the feature extraction module, the output end of the feature extraction module is connected with the hierarchical judgment module, the output end of the hierarchical judgment module is connected with the fitting judgment module, and the output end of the fitting judgment module is connected with the defect output module.

[0087] The image acquisition module is used to acquire the layout area image according to the layout position of the reinforcing material on the FPC.

[0088] The feature extraction module is used to extract the defect feature vector according to the difference image after multiple superimpositions of the layout area image, and the corresponding position of the difference image is used as the guide.

[0089] The hierarchical judgment module is used to cross-solve the defect feature vectors, determine the intersection of each defect feature vector as a target defect area, and determine the hierarchical order of each target defect area according to the priority order represented by each target defect area.

[0090] The fitting judgment module is configured to: according to the hierarchical sequence dependency of each target defect region, when the hierarchical sequence dependency of each target defect region is established, locate deviation data of each target defect region according to the profile offset and the thickness corresponding to each target defect region, and set a fitting judgment result of each target defect region by using the deviation data; and when the hierarchical sequence dependency of each target defect region is not established, perform isolated analysis on the trend of each target defect region, and set the fitting judgment result of each target defect region.

[0091] The defect output module is configured to: according to the fitting judgment result of each target defect region, perform result mapping on the distribution points of the defects after fitting, and determine a defect set in production.

[0092] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application, which are still covered by the protection scope of the present application.

Claims

1. A method for detecting FPC reinforcing bonding quality, characterized in that, The method comprises the following steps: S1, acquiring a laying area image in response to the laying position of the reinforcing material on the FPC; S2, extracting a defect feature vector according to a differential image obtained after multiple superimpositions of the laying area image, and taking the corresponding position of the differential image as a guide; The implementation mode of step S2 further comprises the following steps: S21, analyzing the differential image to obtain defect edge direction data according to the distribution of the reinforcing material on the FPC; S22, extracting a plurality of defect regions corresponding to defect direction density according to the density analysis mode of the defect edge direction data; S23, calculating the area deviation of the defect region according to the position and size of each defect region, and taking the calculated absolute area deviation as the defect feature vector; S3, cross-solving the defect feature vectors to determine the intersection of each defect feature vector as a target defect region, and determining the hierarchical order dependency of each target defect region according to the priority order represented by each target defect region; S4, when the hierarchical order dependency of each target defect region is established, positioning the deviation data of each target defect region according to the contour deviation and thickness corresponding to each target defect region, and setting the fitting judgment result of each target defect region using the deviation data; S5, when the hierarchical order dependency of each target defect region is not established, performing isolated analysis on the direction of each target defect region, and setting the fitting judgment result of each target defect region; S6, based on the fitting judgment result of each target defect region, mapping the distribution points of the defects after fitting to determine the defect set during production.

2. The FPC reinforcing and bonding quality detection method according to claim 1, characterized in that, The implementation mode of step S1 comprises the following steps: S11, taking the area of the reinforcing material on the FPC as a fitting point to obtain an array of fitting points using continuous image data of the laying position; S12, obtaining a gray histogram corresponding to the array of fitting points according to the change of the shooting angle of the array of fitting points in a continuous time window, and taking the gray histogram at multiple angles as the output laying area image.

3. The FPC reinforcing and bonding quality detection method according to claim 1, characterized in that, The implementation mode of step S23 further comprises the following steps: S231, performing surface state mapping on the current input defect region to extract temperature limit values and stress limit values under different directions; S232, performing radius matching of the temperature limit values and the stress limit values with the current defect region to determine the radius matching data of the defect region and the temperature and the stress; S233, based on the correlation between the radius matching data and the defect direction density, sequentially calculating the area deviation corresponding to the defect region.

4. The FPC reinforcing and bonding quality detection method according to claim 1, characterized in that, The implementation mode of step S3 comprises the following steps: S31, mapping each group of defect feature vectors to the same space, and taking the intersection of the defect feature vectors in the same space as the target defect region; S32, drawing a state transition chain of the target defect region using the area size of the target defect region in each cross-solution; S33, after associating and mapping the state transition chain of the target defect region with the data represented by the target defect region, setting the priority order according to the position of the target defect region belonging to the FPC; S34, taking the target defect region with the highest priority as the initial level, taking the target defect region closest to the initial level as the secondary level, and gradually dividing the hierarchical order dependency of each target defect region.

5. The FPC reinforcing and bonding quality detection method according to claim 4, characterized in that, The implementation of the state transition chain in the association mapping in step S33 further includes: associating the defect state with the process parameter according to the defect state described in the state transition chain of the current target defect area; statistically obtaining the state transition probability value corresponding to the defect state in a sliding time window, and extracting the priority weight from the position of the current target defect area according to the confidence degree of the state transition probability value and the process parameter, and taking the moving average value of the priority weight as the priority order of the current target defect area.

6. The FPC reinforcing and bonding quality detection method according to claim 1, characterized in that, The implementation of step S4 further includes: S41, extracting the deviation of the contour offset and the thickness in all hierarchical order dependencies according to the obtained target defect area; S42, setting the maximum traversal number according to the target area number corresponding to the maximum deviation of the contour offset and the thickness; S43, setting the fitting judgment result of the target defect area according to the fitness of each target defect area in the maximum traversal number.

7. The FPC reinforcing and bonding quality detection method according to claim 1, characterized in that, The implementation of step S5 further includes: taking the distance between the maximum deviation and the minimum deviation of the target defect area in the same direction as the output fitting judgment result.

8. The FPC reinforcing and bonding quality detection method according to claim 1, characterized in that, The implementation of step S6 includes: extracting the index value in the fitting judgment result, converting the fitting judgment result into a target variable by normalizing the index value according to the acquisition method; taking the distribution point of the target variable on the FPC, extracting the distribution frequency at the corresponding position according to the specification information of the FPC; based on the obtained distribution frequency, spatially solving the distribution point of the target variable, calculating the distribution deviation of the distribution point compared to the overall distribution, and taking the data corresponding to the maximum distribution deviation as the output defect set.

9. An FPC reinforcement bonding quality detection system for performing the steps in the FPC reinforcement bonding quality detection method according to any one of claims 1-8, characterized in that, It includes: an image acquisition module configured to acquire a layout area image according to the layout position of the reinforcing material on the FPC; a feature extraction module configured to extract a defect feature vector according to a difference image obtained by multiple superimpositions of the layout area image, and take the corresponding position of the difference image as a guide; a hierarchical judgment module configured to cross-solve the defect feature vectors to determine the intersection of each defect feature vector as a target defect area, and determine the hierarchical order dependency of each target defect area according to the priority order represented by each target defect area; a fitting judgment module configured to, when the hierarchical order dependency of each target defect area is established, position the deviation data of each target defect area according to the contour offset and the thickness corresponding to each target defect area, and set the fitting judgment result of each target defect area using the deviation data; when the hierarchical order dependency of each target defect area is not established, isolate the direction of each target defect area for analysis, and set the fitting judgment result of each target defect area; a defect output module configured to map the distribution point of the defect after fitting according to the fitting judgment result of each target defect area, and determine the defect set during production.

Images