Finished product batch quality inspection method for purification filter screen production

Abstract

The invention relates to the technical field of image analysis, in particular to a finished product batch quality inspection method for purification filter screen production, which comprises the following steps: acquiring a filter screen grey-scale map; obtaining a plurality of filter screen edge maps through threshold iteration based on the filter screen grey-scale map; obtaining defect features of the newly added edge through edge dynamic changes identified under different thresholds; obtaining an abnormal factor of each edge in each filter screen edge graph according to the similarity between different edges in each filter screen edge graph and the change of the edge in the adjacent filter screen edge graph; and obtaining the possibility that each pixel point is a defective pixel point according to the abnormal factor of each edge in each filter screen edge image and the defect feature of the newly added edge, so as to detect the quality of the finished filter screen. According to the method, the accuracy of detecting the quality of the finished filter screen is improved by analyzing the performance difference between the defects generated when the folds in the filter screen are damaged and the folds in the filter screen under different threshold values.

Description

Technical Field

[0001] The invention relates to the technical field of image analysis, and in particular to a finished product batch quality inspection method for purification filter production. Background Art

[0002] To ensure the quality of the air filters produced, it is necessary to conduct quality inspection on the finished filters. However, due to the dense folds in the air filters, when the filter images are analyzed by machine vision for quality inspection, the angles between the folds at different positions in the filter and the camera are different, resulting in the folds at positions far from the camera being blocked, making the folds at different positions have different appearances in the filter grayscale image, making it difficult to accurately detect defective areas in the filter. Summary of the invention

[0003] The invention provides a finished product batch quality inspection method for purification filter production to solve the existing problem that it is difficult to accurately detect defects in the filter using traditional inspection methods.

[0004] The present invention provides a finished product batch quality inspection method for purification filter production using the following technical scheme: The following steps are involved: Get the filter grayscale image; Based on the filter grayscale image, several filter edge images are obtained through threshold iteration; the diffusion center line in each filter edge image is obtained; according to the distribution of edge pixels in adjacent filter edge images, the edges in each filter edge image are divided into anchor edges and newly added edges, wherein the newly added edges include independent edges and spreading edges; according to the distance between the newly added edge and the diffusion center line, and the similarity between the spreading edge and the anchor edge, the defect characteristics of the newly added edge are obtained; According to the similarity between different edges in each filter edge map, the abnormality degree of each edge in each filter edge map is obtained; according to the abnormality degree of the edge of each filter edge map in different filter edge maps, the abnormality weight of each edge in each filter edge map is obtained, and combined with the abnormality degree of each edge in each filter edge map, the abnormality factor of each edge in each filter edge map is obtained; According to the abnormal factors of each edge in each filter edge map and the defect characteristics of the newly added edges, the possibility of each pixel being a defective pixel is obtained; according to the possibility of each pixel being a defective pixel, the quality of the finished filter is detected.

[0005] Preferably, the method of obtaining a plurality of filter edge images through threshold iteration based on the filter grayscale image includes the following specific methods: Preset an initial threshold , termination threshold and the threshold iteration step size ; Set the threshold in the edge detection operator to Perform edge detection on the filter grayscale image to obtain the first filter edge image; Set the threshold in the edge detection operator to Perform edge detection on the filter grayscale image to obtain a second filter edge image; Set the threshold in the edge detection operator to Perform edge detection on the filter grayscale image to obtain the third filter edge image; And so on, until the threshold in the edge detection operator is set to Perform edge detection on the filter grayscale image and obtain the A filter edge map, where .

[0006] Preferably, the method of obtaining the diffusion center line in each filter edge map; dividing the edge in each filter edge map into anchor edges and newly added edges according to the distribution of edge pixels in adjacent filter edge maps includes the following specific methods: Get the closest position to the camera in any filter edge image, and draw a straight line parallel to the folds along the position as the diffusion center line of the filter edge image; The filter edge map is used The edge pixels in the filter edge map are Mask the edge pixels in the filter edge map. The filter edge graph exists in The edge pixel that does not exist in the filter edge map is recorded as Add new edge pixels in the filter edge map; The filter edge image and the The edge pixels that exist in all filter edge images are recorded as The anchor edge pixel points in the filter edge map; The adjacent newly added edge pixels in the filter edge map are classified as the same newly added edge, and the adjacent anchor edge pixels are classified as the same anchor edge; The first In the filter edge graph, the newly added edges adjacent to the anchor edge are recorded as the spreading edges of the anchor edge; the newly added edges not adjacent to the anchor edge are recorded as independent edges.

[0007] Preferably, the method of obtaining the defect feature of the newly added edge according to the distance between the newly added edge and the diffusion center line and the similarity between the spreading edge and the anchoring edge includes: For The edge of the filter screen independent edge, according to All independent edges in the filter edge graph are The distance between the diffusion midlines of the filter edge graph is obtained. The edge of the filter The specific calculation formula for the defect characteristics of an independent edge is: , In the formula, Indicates The edge of the filter screen Defect characteristics of independent edges; Indicates The number of independent edges in the edge graph of the filter; Indicates The edge of the filter screen An independent edge and The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter An independent edge and The distance between the diffusion center lines of the filter edge map; It represents the absolute value function; represents the linear normalization function; For The edge of the filter screen The spreading edge The first spreading edge and the The anchor edges adjacent to the spreading edges are classified as the same local edge, denoted as local edges; get the The local edge and The Pearson correlation coefficient between the spreading edges, combined with the All the spreading edges in the filter edge map are The distance between the diffusion midlines of the filter edge graph is obtained. The edge of the filter screen The defect characteristics of a spreading edge, the specific calculation formula is: , In the formula, Indicates The edge of the filter screen Defect characteristics of a spreading edge; Indicates The number of spreading edges in the filter edge graph; Indicates The edge of the filter screen The first spreading edge and the The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen The first spreading edge and the The distance between the diffusion center lines of the filter edge map; Indicates The local edge and Pearson correlation coefficient between the spreading edges; represents the linear normalization function.

[0008] Preferably, the method of obtaining the abnormality degree of each edge in each filter edge map according to the similarity between different edges in each filter edge map includes the following specific methods: For The edge of the filter screen The edge and edge; The first pixel point in the edge is taken as the coordinate origin, the horizontal right direction in the filter edge image is taken as the x-axis direction, and the vertical upward direction in the filter edge image is taken as the y-axis direction. The morphological space of the edge is constructed in the same way. The morphological space of the edges; According to The edge and The coordinates of the edges in their respective morphological spaces are obtained. The edge and DTW distance between edges; For The edge of the filter screen edge, get the The edge of the filter screen The edge and The DTW distance between all other edges in the filter edge graph; combined with Each edge in the filter edge graph is The distance between the diffusion midlines of the filter edge graph is obtained. The edge of the filter screen The degree of abnormality of the edge.

[0009] Preferably, the obtaining The edge of the filter screen The specific calculation formula for the abnormal degree of each edge is as follows: , In the formula, Indicates The edge of the filter screen The degree of abnormality of each edge; Indicates The number of edges in the filter edge graph; Indicates The edge of the filter screen The edge and DTW distance between edges; Indicates The edge of the filter screen The edge and The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen The edge and The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen The edge and The distance between the diffusion center lines of the filter edge map; It represents the absolute value function; represents the linear normalization function.

[0010] Preferably, the abnormal weight of each edge in each filter edge map is obtained according to the abnormal degree of the edge of each filter edge map in different filter edge maps, and the specific method includes: Preset a threshold range For the The filter edge map will be The difference in threshold between the filter edge maps is less than The filter edge map is used as the A threshold neighboring edge map of a filter edge map; For The edge of the filter screen edge, The edge of the filter screen The coordinates of the first edge are marked as reference coordinates; The area composed of edge pixels located at the reference coordinates in the threshold neighboring edge map of the filter edge map is recorded as The edge of the filter screen The corresponding area of ​​the edge; according to the abnormality degree between the corresponding areas, obtain the The edge of the filter screen The specific calculation formula for the abnormal weight of an edge is: , In the formula, Indicates The edge of the filter screen The outlier weight of the edge; Indicates The edge of the filter screen The edge of The degree of abnormality of the corresponding area; Indicates The edge of the filter screen The average abnormality of all corresponding areas of the edge; Indicates The edge of the filter screen The number of corresponding regions of each edge; represents the linear normalization function.

[0011] Preferably, the specific method of obtaining the abnormal factor of each edge in each filter edge map includes: For The edge of the filter screen edge, according to The edge of the filter screen The abnormality of the edge, combined with the The edge of the filter screen The anomaly weight of the edge is obtained The edge of the filter screen The specific calculation formula of the abnormal factor of an edge is: , In the formula, Indicates The edge of the filter screen The outlier factor of the edge; Indicates The edge of the filter screen The degree of abnormality of each edge; Indicates The edge of the filter screen The outlier weight of the edge.

[0012] Preferably, the method of obtaining the possibility that each pixel is a defective pixel according to the abnormal factor of each edge in each filter edge map and the defect feature of the newly added edge includes the following specific methods: For The edge of the filter screen Add an edge to the The defect feature of the newly added edge is used as the The defect features of each pixel in the newly added edge are The coordinates of each pixel in the newly added edge are marked as the target coordinates, and the pixels in all other filter edge images located under the target coordinates are taken as the corresponding pixels of the target pixel, and the defect features of the corresponding pixels of the target pixel are set to be equal to the defect features of the target pixel; For The edge of the filter screen edge, The edge of the filter screen The anomaly factor of the edge is used as the The abnormal factor of each pixel within the edge; According to the abnormal factors of each pixel in all filter edge images and the defect characteristics of each pixel, the possibility of each pixel being a defective pixel is obtained. The specific calculation formula is: , In the formula, Indicates The possibility that a pixel is a defective pixel; Indicates the number of filter edge graphs; Indicates The filter edge in the figure The abnormal factor of each pixel; Indicates The filter edge in the figure Defect characteristics of pixels; represents the linear normalization function.

[0013] Preferably, the method of detecting the quality of the finished filter according to the possibility that each pixel is a defective pixel includes the following specific methods: Preset a probability threshold For any pixel, if the probability that the pixel is a defective pixel is less than , then the pixel is a normal pixel; if the probability that the pixel is a defective pixel is greater than or equal to , then the pixel is a defective pixel; Preset a quality threshold ; For any filter, if the ratio of normal pixels to all pixels in the filter is less than or equal to If the ratio of normal pixels to all pixels in the filter is greater than When , the filter quality is qualified.

[0014] The beneficial effects of the technical solution of the present invention are as follows: a plurality of filter edge images are obtained through threshold iteration based on the filter grayscale image; the defect characteristics of the newly added edges are obtained through the dynamic changes of the edges identified under different thresholds, because the farther the distance between the folds in the filter and the industrial camera, the less obvious the grayscale change between the fold and its adjacent folds is due to the loss of transition information between the folds due to being blocked; conversely, the closer the distance between the folds in the filter and the industrial camera, the more obvious the grayscale change between the fold and its adjacent folds; therefore, under normal circumstances, the edges in the filter grayscale image will diffuse in the direction away from the industrial camera as the threshold decreases, and when the folds in the filter are destroyed, the diffusion will be interfered, thereby obtaining the defect characteristics at various positions in the filter; According to the similarity between different edges in each filter edge map and the changes in the edges in the adjacent filter edge maps, the abnormal factor of each edge in each filter edge map is obtained. Since the spacing between the folds in the filter is uniform, the distribution of the edges of the folds in the filter edge image has a certain regularity. When the folds on the filter surface are defective, the regularity of the folds on the filter surface will be destroyed. Since the defective area where the folds are destroyed is stable within a certain threshold range, the abnormal factor of the edge can be obtained according to the stability of the edge abnormality in different filter edge maps. According to the abnormal factor of each edge in each filter edge map and the defective characteristics of the newly added edges, the possibility of each pixel being a defective pixel is obtained, so as to detect the quality of the finished filter, and finally improve the accuracy of detecting the quality of the finished filter. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A flowchart of the steps of a finished product batch quality inspection method for purification filter production according to the present invention; Figure 2 This is the filter grayscale legend; Figure 3 For the The edge and Schematic diagram of the morphological space of the edges. DETAILED DESCRIPTION

[0017] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the method for batch quality inspection of finished products for purification filter production proposed by the present invention, its specific implementation method, structure, characteristics and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The specific scheme of the finished product batch quality inspection method for purification filter production provided by the present invention is described in detail below with reference to the accompanying drawings.

[0020] See also Figure 1 , which shows a flow chart of the steps of a finished product batch quality inspection method for purification filter production provided by an embodiment of the present invention, the method comprising the following steps: Step S001: Obtain a filter grayscale image.

[0021] It should be noted that the role of the air filter is to improve air quality, protect equipment operation safety, extend equipment service life, and reduce the harm of harmful substances to human health and the environment by filtering pollutants such as particulate matter, dust, and pollen in the air. In order to ensure the quality of the produced air filters, this embodiment proposes a finished product batch quality inspection method for purification filter production, which detects defects in the air filter through computer vision, so it is necessary to first collect the surface image of the air filter.

[0022] Specifically, the filter surface image is collected by an industrial camera, and the filter surface image is grayed to obtain a filter grayscale image, such as Figure 2 As shown, Figure 2 This is a grayscale illustration of the filter.

[0023] Step S002: obtaining several filter edge images through threshold iteration based on the filter grayscale image; obtaining the diffusion centerline in each filter edge image; dividing the edges in each filter edge image into anchor edges and newly added edges according to the distribution of edge pixels in adjacent filter edge images, wherein the newly added edges include independent edges and spreading edges; obtaining the defect features of the newly added edges according to the distance between the newly added edges and the diffusion centerline, and the similarity between the spreading edges and the anchor edges.

[0024] It should be noted that due to the dense folds in the air filter, when the grayscale image of the filter is collected by an industrial camera, the angles between the folds at different positions in the filter and the camera are different, resulting in the folds at positions far away from the camera being blocked. This makes the folds at different positions have different appearances in the grayscale image of the filter, making it difficult to accurately detect defective areas in the filter. Therefore, this embodiment proposes a finished product batch quality inspection method for purification filter production, which specifically detects edges in the grayscale image of the filter through iterative threshold detection, analyzes the edge changes of the filter grayscale image detected at different thresholds, and detects defective areas in the filter.

[0025] It should be further explained that, when the distance between the folds in the filter is farther from the industrial camera, the fold will be blocked, resulting in the loss of some transition information between the folds, making the grayscale change between the fold and its adjacent folds less obvious; conversely, the closer the distance between the folds in the filter and the industrial camera, the more obvious the grayscale change between the fold and its adjacent folds; therefore, under normal circumstances, the edge of the filter grayscale image will diffuse away from the industrial camera as the threshold decreases, and when the folds in the filter are destroyed, it will interfere with the diffusion, thereby obtaining the defect characteristics at various positions in the filter.

[0026] Preferably, in a specific embodiment of the present invention, an initial threshold is preset , termination threshold and the threshold iteration step size , , as well as The specific value of can be set according to the actual situation. This embodiment does not make a hard requirement. , , Narrate; set the threshold in the Sobel operator to Perform edge detection on the filter grayscale image to obtain the first filter edge image; Set the threshold in the Sobel operator to Perform edge detection on the filter grayscale image to obtain the second filter edge image; set the threshold in the edge detection operator to Perform edge detection on the filter grayscale image to obtain the third filter edge image; set the threshold in the edge detection operator to Perform edge detection on the filter grayscale image to obtain the fourth filter edge image; and so on, until the threshold in the Sobel operator is set to Perform edge detection on the filter grayscale image and obtain the The filter edge diagram is shown in FIG. .

[0027] It should be noted that the farther the distance between the folds in the filter and the industrial camera, the less obvious the grayscale change between the fold and its adjacent folds, and a small threshold is required to identify them; and the folds close to the industrial camera do not lose the transition information between the folds, so when the folds far from the industrial camera are identified, the transition parts between the folds close to the industrial camera will also be identified, thereby interfering with the feature that the edges in the grayscale image of the filter will diffuse away from the industrial camera as the threshold is lowered, so they need to be analyzed separately.

[0028] Preferably, in a specific embodiment of the present invention, the position closest to the camera in any filter edge image is obtained, and a straight line parallel to the folds is drawn along the position as the diffusion center line of the filter edge image; Filter edge image ( ),use The edge pixels in the filter edge map are Mask the edge pixels in the filter edge map. The filter edge graph exists in The edge pixel that does not exist in the filter edge map is recorded as Add new edge pixels in the filter edge map; The filter edge image and the The edge pixels that exist in all filter edge images are recorded as The anchor edge pixel points in the filter edge map; Adjacent new edge pixels in the filter edge map are classified as the same new edge, and adjacent anchor edge pixels are classified as the same anchor edge.

[0029] Furthermore, In the edge map of the filter screen, the newly added edges adjacent to the anchor edge are recorded as the spreading edges of the anchor edge; the newly added edges not adjacent to the anchor edge are recorded as independent edges; where adjacent means that there are eight-neighborhood adjacent pixels between the two edges.

[0030] It should be noted that the independent edge represents the area where the grayscale change between the folds in the filter grayscale image is most obvious. Since the edges in the filter grayscale image will diffuse away from the industrial camera as the threshold decreases, and since the spacing between the folds in the filter is uniform, the independent edges in the filter edge image will diffuse evenly away from the industrial camera as the threshold decreases. That is, under normal circumstances, each independent edge in the filter edge image is similar to the distance between the diffusion centerline. The greater the difference between the distance between an independent edge and the diffusion centerline and the distance between other independent edges and the diffusion centerline, the more likely the folds corresponding to the independent edge are to be damaged and defects will occur. Therefore, the defect characteristics of the independent edge can be obtained.

[0031] It should be further explained that the spreading edge represents the grayscale transition area between the folds, that is, the spreading edge and its adjacent anchor edge should correspond to the area between the same folds in the filter, so the spreading edge should be similar to its adjacent anchor edge. Therefore, when calculating the defect feature of the spreading edge, a calculation weight should be further assigned to it according to the degree of similarity between the spreading edge and its adjacent anchor edge.

[0032] Preferably, in a specific embodiment of the present invention, for The edge of the filter screen independent edge, according to All independent edges in the filter edge graph are The distance between the diffusion midlines of the filter edge graph is obtained. The edge of the filter screen The specific calculation formula for the defect characteristics of an independent edge is: , In the formula, Indicates The edge of the filter screen Defect characteristics of independent edges; Indicates The number of independent edges in the edge graph of the filter; Indicates The edge of the filter screen An independent edge and The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen An independent edge and The distance between the diffusion center lines of the filter edge map; It represents the absolute value function; Represents a linear normalization function, the normalization range is all independent edges in all filter edge graphs , For The edge of the filter screen The spreading edge The first spreading edge and the The anchor edges adjacent to the spreading edges are classified as the same local edge, denoted as local edges; get the The local edge and The Pearson correlation coefficient between the spreading edges is a well-known prior art, so it will not be described in detail in this embodiment. All the spreading edges in the filter edge map are The distance between the diffusion midlines of the filter edge graph is obtained. The edge of the filter screen The defect characteristics of a spreading edge, the specific calculation formula is: , In the formula, Indicates The edge of the filter screen Defect characteristics of a spreading edge; Indicates The number of spreading edges in the filter edge graph; Indicates The edge of the filter screen The first spreading edge and the The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen The first spreading edge and the The distance between the diffusion center lines of the filter edge map; Indicates The local edge and Pearson correlation coefficient between the spreading edges; Represents a linear normalization function, and the normalization range is all the spreading edges in all filter edge maps. .

[0033] It should be noted that for independent edges, the greater the difference in distance between an independent edge area and other independent edges on the diffusion centerline, the more likely the fold corresponding to the independent edge is to be damaged. The larger the value of , the more likely the wrinkles corresponding to the independent edge are to be damaged, and the greater the defect feature is. For the spreading edge, unlike the independent edge, the spreading edge corresponds to the grayscale transition part between the wrinkles. Therefore, the morphology of the spreading edge should be similar to its local edge. The less similar the spreading edge is to its local edge, the more likely the corresponding wrinkles are to be damaged. Therefore, it is necessary to assign a negatively correlated calculation weight based on the similarity between the morphology of the spreading edge and its local edge to obtain the defect feature of the spreading edge.

[0034] At this point, the defect features of the newly added edges are obtained.

[0035] Step S003: According to the similarity between different edges in each filter edge map, obtain the abnormality degree of each edge in each filter edge map; according to the abnormality degree of the edge of each filter edge map in different filter edge maps, obtain the abnormality weight of each edge in each filter edge map, and combine the abnormality degree of each edge in each filter edge map to obtain the abnormality factor of each edge in each filter edge map.

[0036] It should be noted that since the spacing between the folds in the filter is uniform, the distribution of the edges of the folds in the filter edge image has a certain regularity. When defects appear on the folds on the filter surface, the regularity of the folds on the filter surface will be destroyed. Therefore, the greater the difference between the edge in the filter edge image and other edges, the more likely the folds corresponding to the edge are to be damaged and defective, thereby obtaining the degree of edge abnormality. Since the defective area where the folds are damaged is stable within a certain threshold range, the edge abnormality weight can be obtained according to the stability of the edge abnormality degree in different filter edge images, and the edge abnormality factor can be further obtained in combination with the edge abnormality degree.

[0037] Preferably, in a specific embodiment of the present invention; for The edge of the filter screen The edge and edge; The first pixel point in the edge is taken as the coordinate origin, the horizontal right direction in the filter edge image is taken as the x-axis direction, and the vertical upward direction in the filter edge image is taken as the y-axis direction. The morphological space of the edge is constructed in the same way. The morphological space of edges, such as Figure 3 As shown, Figure 3 For the The edge and Schematic diagram of the morphological space of the edges; According to The edge and The coordinates of the edges in their respective morphological spaces are obtained. The edge and The DTW distance between the edges. Since the DTW distance is a well-known prior art, it will not be described in detail in this embodiment. For The edge of the filter screen edge, get the The edge of the filter screen The edge and The DTW distance between all other edges in the filter edge graph; combined with Each edge in the filter edge graph is The distance between the diffusion midlines of the filter edge graph is obtained. The edge of the filter screen The specific calculation formula for the abnormal degree of an edge is: , In the formula, Indicates The edge of the filter screen The degree of abnormality of each edge; Indicates The number of edges in the filter edge graph; Indicates The edge of the filter screen The edge and DTW distance between edges; Indicates The edge of the filter screen The edge and The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen The edge and The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen The edge and The distance between the diffusion center lines of the filter edge map; It represents the absolute value function; Represents a linear normalization function, whose normalization range is All the edges in the filter edge graph .

[0038] It should be noted that, since the spacing between the folds in the filter is uniform, the distribution of the edges of the folds in the filter edge image has a certain regularity. In the filter edge image, the closer the distance between the two edges and the diffusion center line is, the more similar the two edges should be. and The larger the value of The edge of the filter screen The more likely the wrinkles corresponding to the first edge are to be damaged, The edge of the filter screen The greater the abnormality of the edge.

[0039] It should be further explained that since the defective area where the folds are destroyed is stable within a certain threshold range, the first The edge of the filter screen The outlier weight of the edge.

[0040] Preferably, in a specific embodiment of the present invention, a threshold range is preset , The specific value of can be set according to the actual situation. This embodiment does not make a hard requirement. To narrate; The filter edge map will be The difference in threshold value between the filter edge images is less than The filter edge map is used as the A threshold neighboring edge map of a filter edge map; Furthermore, for the The edge of the filter screen edge, The edge of the filter screen The coordinates of the first edge are marked as reference coordinates; The area composed of edge pixels located at the reference coordinates in the threshold neighboring edge map of the filter edge map is recorded as The edge of the filter screen The corresponding area of ​​the edge; according to the abnormality degree between the corresponding areas, obtain the The edge of the filter screen The specific calculation formula for the abnormal weight of an edge is: , In the formula, Indicates The edge of the filter screen The outlier weight of the edge; Indicates The edge of the filter screen The edge of The degree of abnormality of the corresponding area; Indicates The edge of the filter screen The average abnormality of all corresponding areas of the edge; Indicates The edge of the filter screen The number of corresponding regions of each edge; Represents a linear normalization function, whose normalization range is All the edges in the filter edge graph .

[0041] It should be noted that It means the The edge of the filter screen The instability of the abnormal degree of the corresponding area of ​​the edge; since the defective area where the fold is destroyed is stable within a certain threshold range, if the first The edge of the filter screen The more unstable the abnormality of the corresponding area of ​​the first edge is, the The edge of the filter screen The less likely the edge is to be a defect area, The larger the value of The edge of the filter screen The more likely the wrinkles corresponding to the edges are to be damaged; further combined with the degree of abnormality, the abnormality factor is obtained.

[0042] In a specific embodiment of the present invention, for The edge of the filter screen edge, according to The edge of the filter screen The abnormality of the edge, combined with the The edge of the filter screen The anomaly weight of the edge is obtained The edge of the filter screen The specific calculation formula of the abnormal factor of an edge is: , In the formula, Indicates The edge of the filter screen The outlier factor of the edge; Indicates The edge of the filter screen The degree of abnormality of each edge; Indicates The edge of the filter screen The outlier weight of the edge.

[0043] It should be noted that the larger the abnormality factor and the degree of abnormality of the edge, the more likely the edge is to be damaged; therefore, the larger the value of the abnormality factor of the edge, the more likely the edge is to be damaged; and since the destruction of the folds in the filter will cause its grayscale transition to be destroyed and thus produce an edge, that is, defects will produce an edge, so for the pixel points in the filter edge map that are not edge pixels, the abnormality factor of the pixel point is zero.

[0044] At this point, the abnormal factors of the edges in the filter edge graph are obtained.

[0045] Step S004: According to the abnormal factors of each edge in each filter edge map and the defect characteristics of the newly added edges, the possibility of each pixel being a defective pixel is obtained; according to the possibility of each pixel being a defective pixel, the quality of the finished filter is detected.

[0046] It should be noted that, the defect characteristics of the newly added edge and the abnormality factor of the edge are obtained respectively through step S002 and step S003; that is, the defect characteristics and abnormality factors of each pixel point can be obtained through the defect characteristics of the newly added edge and the abnormality factors of the edge, and the possibility of each pixel point being a defective pixel point can be further obtained, so as to detect the quality of the filter.

[0047] Preferably, in a specific embodiment of the present invention, for The edge of the filter screen Add an edge to the The defect feature of the newly added edge is used as the The defect features of each pixel in the newly added edge are The coordinates of each pixel in the newly added edge are marked as the target coordinates, and the pixels in all other filter edge images located under the target coordinates are taken as the corresponding pixels of the target pixel, and the defect features of the corresponding pixels of the target pixel are set to be equal to the defect features of the target pixel; For example: The edge of the filter screen The defect characteristic of the newly added edge is 0.7, then The edge of the filter screen The defect characteristics of all pixels in the newly added edge are 0.7. The edge of the filter screen The coordinates of all pixels in the newly added edge are the target coordinates; the defect characteristics of all other pixels in the filter edge map located under the target coordinates are 0.7.

[0048] It should be noted that since the folds in the filter are destroyed, its grayscale transition will be destroyed and edges will be produced, that is, defects will produce edges. Therefore, in all pixel points in the filter edge map that are not located at the target coordinates, the defect characteristics of the pixel points are zero.

[0049] Furthermore, for the The edge of the filter screen edge, The edge of the filter screen The anomaly factor of the edge is used as the The abnormal factor of each pixel within the edge; For example: The edge of the filter screen The outlier factor of the edge is 0.6, then The edge of the filter screen The anomaly factor of each pixel within the edge is 0.6.

[0050] According to the abnormal factors of each pixel in all filter edge images and the defect characteristics of each pixel, the possibility of each pixel being a defective pixel is obtained. The specific calculation formula is: , In the formula, Indicates The possibility that a pixel is a defective pixel; Indicates the number of filter edge graphs; Indicates The filter edge in the figure The abnormal factor of each pixel; Indicates The filter edge in the figure Defect characteristics of pixels; Represents a linear normalization function, whose normalization range is all pixels .

[0051] It should be noted that the larger the abnormal factor and defect feature of a pixel point, the more likely the pixel point is to be located at the place where the fold is damaged; that is, based on the possibility that the pixel point is a defective pixel point, the defective pixel point is obtained and used to detect the quality of the filter; since there is no new edge in the first filter edge map, starting from the second filter edge map, the possibility of the pixel point at each position being a defective pixel point is calculated.

[0052] Specifically, set a probability threshold , The specific value of can be set according to the actual situation. This embodiment does not make a hard requirement. =0.8; for any pixel, if the probability that the pixel is a defective pixel is less than , then the pixel is a normal pixel; if the probability that the pixel is a defective pixel is greater than or equal to , then the pixel is a defective pixel; Furthermore, a quality threshold is preset , The specific value of can be set according to the actual situation. This embodiment does not make a hard requirement. For any filter, if the ratio of normal pixels to all pixels in the filter is less than or equal to If the ratio of normal pixels to all pixels in the filter is greater than When , the filter quality is qualified.

[0053] At this point, this embodiment is completed.

[0054] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A finished product batch quality inspection method for purification filter production, characterized in that: The method comprises the following steps: Get the filter grayscale image; Based on the filter grayscale image, several filter edge images are obtained through threshold iteration; the diffusion center line in each filter edge image is obtained; according to the distribution of edge pixels in adjacent filter edge images, the edges in each filter edge image are divided into anchor edges and newly added edges, wherein the newly added edges include independent edges and spreading edges; according to the distance between the newly added edge and the diffusion center line, and the similarity between the spreading edge and the anchor edge, the defect characteristics of the newly added edge are obtained; According to the similarity between different edges in each filter edge map, the abnormality degree of each edge in each filter edge map is obtained; according to the abnormality degree of the edge of each filter edge map in different filter edge maps, the abnormality weight of each edge in each filter edge map is obtained, and combined with the abnormality degree of each edge in each filter edge map, the abnormality factor of each edge in each filter edge map is obtained; According to the abnormal factors of each edge in each filter edge map and the defect characteristics of the newly added edges, the possibility of each pixel being a defective pixel is obtained; according to the possibility of each pixel being a defective pixel, the quality of the finished filter is detected.

2. A finished product batch quality inspection method for purification filter production according to claim 1, characterized in that: The specific method of obtaining a plurality of filter edge images based on the filter grayscale image by threshold iteration is as follows: Preset an initial threshold , termination threshold and the threshold iteration step size ; Set the threshold in the edge detection operator to Perform edge detection on the filter grayscale image to obtain the first filter edge image; Set the threshold in the edge detection operator to Perform edge detection on the filter grayscale image to obtain a second filter edge image; Set the threshold in the edge detection operator to Perform edge detection on the filter grayscale image to obtain the third filter edge image; And so on, until the threshold in the edge detection operator is set to Perform edge detection on the filter grayscale image and obtain the A filter edge map, where .

3. A finished product batch quality inspection method for purification filter production according to claim 1, characterized in that: The method of obtaining the diffusion center line in each filter edge map and dividing the edge in each filter edge map into anchor edges and newly added edges according to the distribution of edge pixels in adjacent filter edge maps includes: Get the closest position to the camera in any filter edge image, and draw a straight line parallel to the folds along the position as the diffusion center line of the filter edge image; The filter edge map is used The edge pixels in the filter edge map are Mask the edge pixels in the filter edge map. The filter edge graph exists in The edge pixel that does not exist in the filter edge map is recorded as Add new edge pixels in the filter edge map; The filter edge image and the The edge pixels that exist in all filter edge images are recorded as The anchor edge pixel points in the filter edge map; The adjacent newly added edge pixels in the filter edge map are classified as the same newly added edge, and the adjacent anchor edge pixels are classified as the same anchor edge; The first In the filter edge graph, the newly added edges adjacent to the anchor edge are recorded as the spreading edges of the anchor edge; the newly added edges not adjacent to the anchor edge are recorded as independent edges.

4. A finished product batch quality inspection method for purification filter production according to claim 1, characterized in that: The method of obtaining the defect feature of the newly added edge according to the distance between the newly added edge and the diffusion center line and the similarity between the spreading edge and the anchoring edge includes: For The edge of the filter independent edge, according to All independent edges in the filter edge graph are The distance between the diffusion midlines of the filter edge graph is obtained. The edge of the filter screen The specific calculation formula for the defect characteristics of an independent edge is: ， In the formula, Indicates The edge of the filter screen Defect characteristics of independent edges; Indicates The number of independent edges in the edge graph of the filter; Indicates The edge of the filter screen An independent edge and The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen An independent edge and The distance between the diffusion center lines of the filter edge map; It represents the absolute value function; represents the linear normalization function; For The edge of the filter screen The spreading edge The spreading edge and The anchor edges adjacent to the spreading edges are classified as the same local edge, denoted as local edges; get the The local edge and The Pearson correlation coefficient between the spreading edges, combined with the All the spreading edges in the filter edge map are The distance between the diffusion midlines of the filter edge graph is obtained. The edge of the filter screen The defect characteristics of a spreading edge, the specific calculation formula is: ， In the formula, Indicates The edge of the filter screen Defect characteristics of a spreading edge; Indicates The number of spreading edges in the filter edge graph; Indicates The edge of the filter screen The spreading edge and The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen The spreading edge and The distance between the diffusion center lines of the filter edge map; Indicates The local edge and Pearson correlation coefficient between the spreading edges; represents the linear normalization function.

5. A finished product batch quality inspection method for purification filter production according to claim 1, characterized in that: The specific method of obtaining the abnormality degree of each edge in each filter edge map according to the similarity between different edges in each filter edge map is as follows: For The edge of the filter screen The edge and edge; The first pixel point in the edge is taken as the coordinate origin, the horizontal right direction in the filter edge image is taken as the x-axis direction, and the vertical upward direction in the filter edge image is taken as the y-axis direction. The morphological space of the edge is constructed in the same way. The morphological space of the edges; According to The edge and The coordinates of the edges in their respective morphological spaces are obtained. The edge and DTW distance between edges; For The edge of the filter screen edge, get the The edge of the filter screen The edge and The DTW distance between all other edges in the filter edge graph; combined with the Each edge in the filter edge graph is The distance between the diffusion midlines of the filter edge graph is obtained. The edge of the filter screen The degree of abnormality of the edge.

6. A finished product batch quality inspection method for purification filter production according to claim 5, characterized in that: The acquisition The edge of the filter screen The specific calculation formula for the abnormal degree of each edge is as follows: ， In the formula, Indicates The edge of the filter screen The degree of abnormality of each edge; Indicates The number of edges in the filter edge graph; Indicates The edge of the filter screen The edge and DTW distance between edges; Indicates The edge of the filter screen The edge and The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen The edge and The distance between the diffusion center lines of the filter edge map; Indicates The edge of the filter screen The edge and The distance between the diffusion center lines of the filter edge map; It represents the absolute value function; represents the linear normalization function.

7. A finished product batch quality inspection method for purification filter production according to claim 1, characterized in that: The specific method of obtaining the abnormal weight of each edge in each filter edge map according to the abnormal degree of the edge of each filter edge map in different filter edge maps includes: Preset a threshold range For the The filter edge map will be The difference in threshold between the filter edge maps is less than The filter edge map is used as the A threshold neighboring edge map of a filter edge map; For The edge of the filter screen edge, The edge of the filter screen The coordinates of the first edge are marked as reference coordinates; The area composed of edge pixels located at the reference coordinates in the threshold neighboring edge map of the filter edge map is recorded as The edge of the filter screen The corresponding area of ​​the edge; according to the abnormality degree between the corresponding areas, obtain the The edge of the filter screen The specific calculation formula for the abnormal weight of an edge is: ， In the formula, Indicates The edge of the filter screen The outlier weight of the edge; Indicates The edge of the filter screen The edge of The degree of abnormality of the corresponding area; Indicates The edge of the filter screen The average abnormality of all corresponding areas of the edge; Indicates The edge of the filter screen The number of corresponding regions for each edge; represents the linear normalization function.

8. A finished product batch quality inspection method for purification filter production according to claim 1, characterized in that: The specific method of obtaining the abnormal factor of each edge in each filter edge map includes: For The edge of the filter screen edge, according to The edge of the filter screen The abnormality of the edge, combined with the The edge of the filter screen The anomaly weight of the edge is obtained The edge of the filter screen The specific calculation formula of the abnormal factor of an edge is: ， In the formula, Indicates The edge of the filter screen The outlier factor of the edge; Indicates The edge of the filter screen The degree of abnormality of each edge; Indicates The edge of the filter screen The outlier weight of the edge.

9. A finished product batch quality inspection method for purification filter production according to claim 1, characterized in that: The method of obtaining the possibility that each pixel is a defective pixel according to the abnormal factor of each edge in each filter edge map and the defect feature of the newly added edge includes: For The edge of the filter screen Add an edge to the The defect feature of the newly added edge is used as the The defect features of each pixel in the newly added edge are The coordinates of each pixel in the newly added edge are marked as the target coordinates, and the pixels in all other filter edge images located under the target coordinates are taken as the corresponding pixels of the target pixel, and the defect features of the corresponding pixels of the target pixel are set to be equal to the defect features of the target pixel; For The edge of the filter screen edge, The edge of the filter screen The anomaly factor of the edge is used as the The abnormal factor of each pixel within the edge; According to the abnormal factors of each pixel in all filter edge images and the defect characteristics of each pixel, the possibility of each pixel being a defective pixel is obtained. The specific calculation formula is: ， In the formula, Indicates The possibility that a pixel is a defective pixel; Indicates the number of filter edge graphs; Indicates The filter edge in the figure The abnormal factor of each pixel; Indicates The filter edge in the figure Defect characteristics of pixels; represents the linear normalization function.

10. A finished product batch quality inspection method for purification filter production according to claim 1, characterized in that: The method of detecting the quality of the finished filter according to the possibility that each pixel is a defective pixel includes: Preset a probability threshold For any pixel, if the probability that the pixel is a defective pixel is less than , then the pixel is a normal pixel; if the probability that the pixel is a defective pixel is greater than or equal to , then the pixel is a defective pixel; Preset a quality threshold ; For any filter, if the ratio of normal pixels to all pixels in the filter is less than or equal to If the ratio of normal pixels to all pixels in the filter is greater than When , the filter quality is qualified.

Images