A printed product defect detection method and system based on deep learning

Abstract

The present invention discloses a method and system for detecting printed product defects based on deep learning, and relates to the technical field of printed product defect detection. The method solves the problem that traditional detection methods often miss detection and make false detections when processing tiny broken defects on the surface of high-precision printed products, especially when processing complex surface structures and unclear defect features, and the detection accuracy is difficult to guarantee. The present invention establishes a set of efficient and intelligent defect detection processes through technical means such as area division, historical data classification, image comparison and precise positioning of abnormal points. By utilizing the powerful computing power and professional rule setting of the deep learning model, the method can quickly identify key defect areas, accurately locate abnormal points and damaged areas, greatly improve the degree of automation and reliability of detection, and effectively reduce false detection and missed detection.

Description

Technical Field

[0001] The present invention relates to the technical field of printed product defect detection, and specifically to a printed product defect detection method and system based on deep learning. Background Art

[0002] With the rapid development of the printing industry, the quality requirements of printed products are constantly increasing, especially in the field of high-precision printing, such as steel plate printing, precision woodblock printing, etc., the requirements for the surface quality of printed products are becoming more and more stringent. At present, the detection of surface defects of printed products mainly relies on manual detection or traditional machine vision detection methods, but these methods have obvious shortcomings in practical applications.

[0003] Although manual inspection methods can rely on the experience of inspectors to inspect the surface of printed products, they have problems such as low efficiency, fatigue, and inconsistent inspection standards. Although traditional machine vision inspection methods can improve inspection efficiency, they have the following technical difficulties when dealing with broken defect inspection of high-precision printed products:

[0004] First, the broken defects on the surface of high-precision printed products are often random and diverse, and traditional detection methods are difficult to accurately identify the characteristics of different types of defects. Secondly, since the location of damage on the surface of printed products has a certain historical regularity, the existing technology cannot fully utilize historical data to conduct targeted detection of key areas, resulting in low detection accuracy and efficiency. Thirdly, in the actual production process, the importance of broken defects in different areas is different. The existing technology lacks an effective regional classification detection mechanism, making it difficult to achieve differentiated detection.

[0005] In addition, traditional detection methods often miss or misdetect tiny broken defects on the surface of high-precision printed products, especially when the surface structure is complex and the defect features are not obvious, so the detection accuracy is difficult to guarantee. At the same time, the existing technology also lacks accurate calculation and positioning methods for the area of ​​the defect area, and cannot provide accurate data support for subsequent quality control.

[0006] Therefore, it is urgent to develop a new method that can make full use of deep learning technology to realize intelligent detection of surface breakage defects of high-precision printed products, so as to improve the accuracy and efficiency of detection and meet the strict requirements of the modern printing industry for product quality control. Summary of the invention

[0007] In view of the shortcomings of the prior art, the present invention provides a printed product defect detection method and system based on deep learning, which solves the problems in the background technology.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: A printed product defect detection method based on deep learning, comprising:

[0009] Step 1: obtaining the entire printing surface area of ​​the printed matter, and dividing the entire printing surface area into a plurality of sub-areas;

[0010] Step 2: According to the historical damage defect position records of the printed surface of the printed product, several sub-areas are classified into key defect sub-areas and non-key defect sub-areas;

[0011] Step 3: Capture images of subsequent printed products, and simultaneously obtain a printing standard image, compare the captured image with the printing standard image, determine the image abnormality area of ​​the captured image, and mark the printed product corresponding to the image abnormality area to obtain the abnormal area of ​​the defective printed product;

[0012] Step 4: based on the key defect sub-regions and the non-key defect sub-regions, determining in turn whether several sub-regions of the defective printed product are marked as key defect identification sub-regions or non-key defect identification sub-regions, and determining abnormal points for each sub-region marked as key defect identification sub-region or non-key defect identification sub-region;

[0013] The specific contents of the abnormal point determination include:

[0014] BS1: Determine one of the defect key identification sub-areas or defect non-key identification sub-areas, and establish a horizontal reference lower plane on the horizontal plane where the defect key identification sub-area or defect non-key identification sub-area is located, and the height from the horizontal reference lower plane is The position of the parallel reference upper plane is established, where It is the preset standard height;

[0015] BS2: According to the specific location of the defect key identification sub-area or defect non-key identification sub-area, make a perpendicular line segments, where It is determined according to the defect focus identification sub-region or defect non-focus identification sub-region. If the sub-region is recorded as a defect focus identification sub-region, the larger k is, and if the sub-region is recorded as a defect non-focus identification sub-region, the larger k is. The smaller;

[0016] BS3: Then count the height of each perpendicular line segment , get the height of each perpendicular line segment With standard height The absolute value of ,in, for Any one of the perpendicular segments, ; Where m = the area of ​​the defect focus identification sub-region or the defect non-focus identification sub-region × ;

[0017] Then the absolute value With the preset threshold For comparison:

[0018] like , mark the perpendicular line segment as an abnormal perpendicular line segment;

[0019] like , no processing is done;

[0020] BS4: Count the total number of abnormal perpendicular segments and record it as , and calculate and Ratio , if the ratio Greater than the preset judgment value When the vertical points of all the vertical lines marked as abnormal foot segments that touch the defect key identification sub-area or the defect non-key identification sub-area are marked as abnormal points;

[0021] Step 5: Determine the area of ​​the abnormal region based on the abnormal points of each defect key identification sub-region or defect non-key identification sub-region, and determine whether each defect key identification sub-region or defect non-key identification sub-region is a printing breakage defect region based on the abnormal region area, and determine the specific location of the breakage defect in the printing breakage defect region.

[0022] As a further solution of the present invention: In the step 2, the specific method of classifying the plurality of sub-regions into key defect sub-regions and non-key defect sub-regions according to the historical defect position record of the printed surface of the printed product is as follows:

[0023] AS1: Starting from the current time, obtain the historical damage and defect location records of the printed surface of the printed products in the previous 30 days, and mark this period of time as the recording period;

[0024] AS2: Count the total number of times each of several sub-areas is marked as a damage defect during the recording period and record it as ,in, , i represents the i-th sub-region, is the total number of sub-regions;

[0025] AS3: Then pass the formula , determine the number of times each of several sub-areas was marked as a broken defect each day during the recording period , and Make a judgment:

[0026] like , then the i-th sub-region is classified as a key defect region;

[0027] like , then the i-th sub-region is classified as a non-key defect region, where Indicates the preset threshold.

[0028] As a further solution of the present invention: in the step 3, determining the content of the abnormal area of ​​the image includes:

[0029] An image that is inconsistent with a printing standard image is identified and marked as an abnormal printing image, and the printed product is marked as a defective print. The image inconsistency area between the abnormal printing image and the printing standard image is then determined and marked as an image abnormality area, and the image abnormality area is determined in the entire printing surface area of ​​the defective print.

[0030] As a further solution of the present invention: the step BS4 further includes:

[0031] BS5: Repeat steps BS1-BS4 to determine each abnormal point marked as a defect focus identification sub-region or a defect non-focus identification sub-region.

[0032] As a further solution of the present invention: in the step 5, the abnormal area is determined according to the abnormal points of each defect key identification sub-region or defect non-key identification sub-region, and whether each defect key identification sub-region or defect non-key identification sub-region is a printing breakage defect region is determined according to the abnormal area, and the specific position of the breakage defect in the printing breakage defect region is determined in the following specific manner:

[0033] CS1: After obtaining the abnormal points in one of the defect key identification sub-areas or defect non-key identification sub-areas, a rectangular coordinate system is established and the distance between two abnormal points is calculated using the following formula :

[0034] ;

[0035] In the formula, and For any two outliers, Expressed as Point horizontal coordinate, Expressed as The vertical coordinate of the point, Expressed as Point horizontal coordinate, Expressed as Point vertical coordinate;

[0036] CS2: Then get the two outliers with the largest distance and mark the distance as , then determine the distance The center point of Draw a circle with the radius in the defect key identification sub-area or defect non-key identification sub-area, and determine the area of ​​the abnormal area by the following formula :

[0037] ;

[0038] In the formula, is an influencing factor, and its value is different in the defect focus identification sub-area and the defect non-focus identification sub-area;

[0039] CS3: Get the preset reference threshold area and mark it as , the area of ​​the benchmark threshold region The area of ​​abnormal area Comparing and determining whether the defect key identification sub-region or the defect non-key identification sub-region is a printing breakage defect region, and determining the specific location of the breakage defect in the printing breakage defect region;

[0040] CS4: Repeat steps CS1-CS3 to determine whether each defect key identification sub-region or defect non-key identification sub-region is a printing breakage defect region, and determine the specific location of the breakage defect in the printing breakage defect region.

[0041] As a further solution of the present invention: in step CS3, the specific method of determining whether the defect focus identification sub-region or the defect non-focus identification sub-region is a printing breakage defect region, and determining the specific position of the breakage defect in the printing breakage defect region is:

[0042] like When , it indicates the area of ​​abnormal region If the defect is too large, the defect key identification sub-region or the defect non-key identification sub-region is determined as a printing broken region; at the same time, the area of ​​the abnormal region is Determine the specific location of the broken defect as the printing broken defect area;

[0043] like When , it indicates the area of ​​abnormal region Within an acceptable range, the defect focus identification sub-region or the defect non-focus identification sub-region is determined as a normal printing area.

[0044] A method for detecting printed product defects based on deep learning, comprising:

[0045] A region division module is used to obtain the entire printing surface area of ​​the printed matter and divide the entire printing surface area into a plurality of sub-areas;

[0046] A region classification module is used to classify several sub-regions into key defect sub-regions and non-key defect sub-regions according to the historical damage defect position records of the printed surface of the printed product;

[0047] The image comparison module is used to collect images of subsequent printed products and simultaneously obtain a printing standard image, compare the collected image with the printing standard image, determine the image abnormality area of ​​the collected image, and mark the printed product corresponding to the image abnormality area to obtain the abnormal area of ​​the defective printed product;

[0048] An abnormality judgment module is used to judge whether several sub-regions of defective printed products are marked as defect key identification sub-regions or defect non-key identification sub-regions based on the key defect sub-regions and non-key defect sub-regions, and determine abnormal points for each sub-region marked as a defect key identification sub-region or a defect non-key identification sub-region;

[0049] The specific contents of the abnormal point determination include:

[0050] Determine one of the defect key identification sub-regions or defect non-key identification sub-regions, and establish a horizontal reference lower plane on the horizontal plane where the defect key identification sub-region or defect non-key identification sub-region is located, and the height from the horizontal reference lower plane is The position of the parallel reference upper plane is established, where It is the preset standard height;

[0051] According to the specific location of the defect key identification sub-region or the defect non-key identification sub-region, a plane parallel to the reference is made from the plane parallel to the defect key identification sub-region or the defect non-key identification sub-region per area unit. perpendicular line segments, where It is determined according to the defect focus identification sub-region or defect non-focus identification sub-region. If the sub-region is recorded as a defect focus identification sub-region, the larger k is, and if the sub-region is recorded as a defect non-focus identification sub-region, the larger k is. The smaller;

[0052] Then count the height of each perpendicular line segment , get the height of each perpendicular line segment With standard height The absolute value of ,in, for Any one of the perpendicular segments, ; Where m = the area of ​​the defect focus identification sub-region or the defect non-focus identification sub-region × ;

[0053] Then the absolute value With the preset threshold For comparison:

[0054] like , mark the perpendicular line segment as an abnormal perpendicular line segment;

[0055] like , no processing is done;

[0056] Count the total number of abnormal perpendicular segments and record it as , and calculate and Ratio , if the ratio Greater than the preset judgment value When the vertical points of all the vertical lines marked as abnormal foot segments that touch the defect key identification sub-area or the defect non-key identification sub-area are marked as abnormal points;

[0057] The defect positioning module is used to determine the area of ​​the abnormal area according to the abnormal points of each defect key identification sub-area or defect non-key identification sub-area, and determine whether each defect key identification sub-area or defect non-key identification sub-area is a printing breakage defect area according to the area of ​​the abnormal area, and determine the specific location of the breakage defect in the printing breakage defect area.

[0058] The present invention provides a method and system for detecting defects in printed products based on deep learning. Compared with the prior art, the method and system have the following beneficial effects:

[0059] The printed product defect detection method and system based on deep learning provided by the present invention establishes an efficient and intelligent defect detection process through technical means such as area division, historical data classification, image comparison and accurate positioning of abnormal points. By utilizing the powerful computing power and professional rule setting of the deep learning model, it is possible to quickly identify key defect areas, accurately locate abnormal points and damaged areas, greatly improve the automation and reliability of detection, and effectively reduce false detection and missed detection.

[0060] This method realizes the dynamic optimization of the detection process and the rational allocation of resource utilization, and adapts to the detection needs of diversified printed products. The overall system improves the detection accuracy and efficiency significantly, provides a scientific basis for the quality control of printed products, further reduces production costs, and ensures high-quality output of products. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The present invention will be further described below in conjunction with the accompanying drawings.

[0062] Figure 1 This is a flowchart of the steps of a printed product defect detection method based on deep learning of the present invention.

[0063] Figure 2This is a structural framework diagram of a printed product defect detection system based on deep learning in the present invention. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] Embodiment 1

[0066] See also Figure 1 ,The present invention provides a method for detecting printed product defects based on deep learning, comprising:

[0067] Step 1: obtaining the entire printing surface area of ​​the printed matter, and dividing the entire printing surface area into a plurality of sub-areas;

[0068] It should be noted that the division of the overall area of ​​the printed surface is specifically set by professional staff. For different printed products, the overall area of ​​the printed product is different. Through the pre-set area division rules, the corresponding printed product can be quickly divided into regions. The printed product described in this embodiment is a determined target printed product, for example, it can be a printed steel plate surface or a printed wood board surface;

[0069] Step 2: According to the historical damage defect position records of the printed surface of the printed product, several sub-areas are classified into key defect sub-areas and non-key defect sub-areas;

[0070] It should be noted that the historical damage defect location record indicates the specific location of the damage problem on the printed surface detected by the corresponding printed product defect detection device after the printed product is printed;

[0071] The specific method of classifying several sub-regions into key defect sub-regions and non-key defect sub-regions according to the historical defect position records of the printed surface of the printed product is as follows:

[0072] AS1: Starting from the current time, obtain the historical damage and defect location records of the printed surface of the printed products in the previous 30 days, and mark this period of time as the recording period;

[0073] AS2: Count the total number of times each of several sub-areas is marked as a damage defect during the recording period and record it as ,in, , i represents the i-th sub-region, is the total number of sub-regions;

[0074] AS3: Then pass the formula , determine the number of times each of several sub-areas was marked as a broken defect each day during the recording period , and Make a judgment:

[0075] like , then the i-th sub-region is classified as a key defect region;

[0076] like , then the i-th sub-region is classified as a non-key defect region, where It is represented as a preset threshold value, which is determined by professional staff;

[0077] Step 3: Capture images of subsequent printed products, and simultaneously obtain a printing standard image, compare the captured image with the printing standard image, determine the image abnormality area of ​​the captured image, and mark the printed product corresponding to the image abnormality area to obtain the abnormal area of ​​the defective printed product;

[0078] The content of determining the abnormal area of ​​the image includes:

[0079] An image that is inconsistent with a printing standard image is identified and marked as an abnormal printing image, and the printed product is marked as a defective print. The image inconsistency area between the abnormal printing image and the printing standard image is then determined and marked as an image abnormality area, and the image abnormality area is determined in the entire printing surface area of ​​the defective print.

[0080] It should be noted that the printing standard image is determined by professional staff, and the printing standard image is represented as an image without any defects or abnormalities;

[0081] Step 4: based on the key defect sub-regions and the non-key defect sub-regions, determining in turn whether several sub-regions of the defective printed product are marked as key defect identification sub-regions or non-key defect identification sub-regions, and determining abnormal points for each sub-region marked as key defect identification sub-region or non-key defect identification sub-region;

[0082] The specific contents of sequentially judging whether a plurality of sub-regions of the defective printed product are marked as defect-focused identification sub-regions or defect-non-focused identification sub-regions, and determining abnormal points for each of the sub-regions marked as defect-focused identification sub-regions or defect-non-focused identification sub-regions include:

[0083] BS1: Determine one of the defect key identification sub-areas or defect non-key identification sub-areas, and establish a horizontal reference lower plane on the horizontal plane where the defect key identification sub-area or defect non-key identification sub-area is located, and the height from the horizontal reference lower plane is The position of the parallel reference upper plane is established, where It is the preset standard height;

[0084] BS2: According to the specific location of the defect key identification sub-area or defect non-key identification sub-area, make a perpendicular line segments, where It is determined according to the defect focus identification sub-region or defect non-focus identification sub-region. If the sub-region is recorded as a defect focus identification sub-region, the larger k is, and if the sub-region is recorded as a defect non-focus identification sub-region, the larger k is. The smaller it is, the selection of k will be determined by professional staff;

[0085] BS3: Then count the height of each perpendicular line segment , get the height of each perpendicular line segment With standard height The absolute value of ,in, for Any one of the perpendicular segments, ; Where m = the area of ​​the defect focus identification sub-region or the defect non-focus identification sub-region × ;

[0086] Then the absolute value With the preset threshold For comparison:

[0087] like , mark the perpendicular line segment as an abnormal perpendicular line segment;

[0088] like , no processing is done;

[0089] BS4: Count the total number of abnormal perpendicular segments and record it as , and calculate and Ratio , if the ratio Greater than the preset judgment value When the vertical points of all the vertical lines marked as abnormal foot segments that touch the defect key identification sub-area or the defect non-key identification sub-area are marked as abnormal points;

[0090] BS5: Repeat steps BS1-BS4 to determine each abnormal point marked as a defect focus identification sub-region or a defect non-focus identification sub-region;

[0091] By conducting detailed inspections on each key or non-key sub-region marked as a defect, the abnormal points are further determined using the perpendicular line segment statistical method, significantly improving the accuracy and reliability of abnormality identification. Combined with the classification information of the sub-region, the abnormal characteristics of each sub-region are accurately calculated by constructing horizontal benchmarks and benchmark planes, and the abnormal perpendicular line segments are screened using thresholds to ensure that only the real abnormal regions are focused on. This step can not only reduce misjudgments, but also provide reliable data basis for subsequent in-depth analysis and corrections.

[0092] Step 5: Determine the area of ​​the abnormal region according to the abnormal points of each defect key identification sub-region or defect non-key identification sub-region, and determine whether each defect key identification sub-region or defect non-key identification sub-region is a printing breakage defect region according to the abnormal region area, and determine the specific location of the breakage defect in the printing breakage defect region;

[0093] The specific method of determining the area of ​​the abnormal region according to the abnormal points of each defect key identification sub-region or defect non-key identification sub-region, determining whether each defect key identification sub-region or defect non-key identification sub-region is a printing breakage defect region according to the area of ​​the abnormal region, and determining the specific position of the breakage defect in the printing breakage defect region is as follows:

[0094] CS1: After obtaining the abnormal points in one of the defect key identification sub-areas or defect non-key identification sub-areas, a rectangular coordinate system is established and the distance between two abnormal points is calculated using the following formula :

[0095] ;

[0096] In the formula, and For any two outliers, Expressed as Point horizontal coordinate, Expressed as The vertical coordinate of the point, Expressed as Point horizontal coordinate, Expressed as Point vertical coordinate;

[0097] CS2: Then get the two outliers with the largest distance and mark the distance as , then determine the distance The center point of Draw a circle with the radius in the defect key identification sub-area or defect non-key identification sub-area, and determine the area of ​​the abnormal area by the following formula :

[0098] ;

[0099] In the formula, It is an influencing factor, and its value is different in the defect key identification sub-area and the defect non-key identification sub-area, and is specifically set by professional staff;

[0100] CS3: Get the preset reference threshold area and mark it as , the area of ​​the benchmark threshold region The area of ​​abnormal area For comparison:

[0101] like When , it indicates the area of ​​abnormal region If the defect is too large, the defect key identification sub-region or the defect non-key identification sub-region is determined as a printing broken region; at the same time, the area of ​​the abnormal region is Determine the specific location of the broken defect as the printing broken defect area;

[0102] like When , it indicates the area of ​​abnormal region Within an acceptable range, the defect key identification sub-region or the defect non-key identification sub-region is determined as a normal printing area;

[0103] CS4: repeat steps CS1-CS3 to determine whether each defect key identification sub-region or defect non-key identification sub-region is a printing breakage defect region, and determine the specific location of the breakage defect in the printing breakage defect region;

[0104] By calculating the abnormal area of ​​each defect key identification sub-area or non-key identification sub-area, it is further confirmed whether it belongs to the printing broken defect area and the specific location of the damaged defect is accurately located; using the coordinate system and the distance calculation method of the abnormal point, the area of ​​the abnormal area is accurately defined and compared with the preset benchmark, thereby achieving a quantitative judgment of the severity of the damage.

[0105] Embodiment 2

[0106] See also Figure 2 In the specific implementation process, this embodiment is based on the first embodiment and is different from the first embodiment in that this embodiment further provides a printed product defect detection system based on deep learning, including:

[0107] A region division module is used to obtain the entire printing surface area of ​​the printed matter and divide the entire printing surface area into a plurality of sub-areas;

[0108] A region classification module is used to classify several sub-regions into key defect sub-regions and non-key defect sub-regions according to the historical damage defect position records of the printed surface of the printed product;

[0109] The image comparison module is used to collect images of subsequent printed products and simultaneously obtain a printing standard image, compare the collected image with the printing standard image, determine the image abnormality area of ​​the collected image, and mark the printed product corresponding to the image abnormality area to obtain the abnormal area of ​​the defective printed product;

[0110] An abnormality judgment module is used to judge whether several sub-regions of defective printed products are marked as defect key identification sub-regions or defect non-key identification sub-regions based on the key defect sub-regions and non-key defect sub-regions, and determine abnormal points for each sub-region marked as a defect key identification sub-region or a defect non-key identification sub-region;

[0111] The specific contents of sequentially judging whether a plurality of sub-regions of the defective printed product are marked as defect-focused identification sub-regions or defect-non-focused identification sub-regions, and determining abnormal points for each of the sub-regions marked as defect-focused identification sub-regions or defect-non-focused identification sub-regions include:

[0112] Determine one of the defect key identification sub-regions or defect non-key identification sub-regions, and establish a horizontal reference lower plane on the horizontal plane where the defect key identification sub-region or defect non-key identification sub-region is located, and the height from the horizontal reference lower plane is The position of the parallel reference upper plane is established, where It is the preset standard height;

[0113] According to the specific location of the defect key identification sub-region or the defect non-key identification sub-region, a plane parallel to the reference is made from the plane parallel to the defect key identification sub-region or the defect non-key identification sub-region per area unit. perpendicular line segments, where It is determined according to the defect focus identification sub-region or defect non-focus identification sub-region. If the sub-region is recorded as a defect focus identification sub-region, the larger k is, and if the sub-region is recorded as a defect non-focus identification sub-region, the larger k is. The smaller it is, the selection of k will be determined by professional staff;

[0114] Then count the height of each perpendicular line segment , get the height of each perpendicular line segment With standard height The absolute value of ,in, for Any one of the perpendicular segments, ; Where m = the area of ​​the defect focus identification sub-region or the defect non-focus identification sub-region × ;

[0115] Then the absolute value With the preset threshold For comparison:

[0116] like , mark the perpendicular line segment as an abnormal perpendicular line segment;

[0117] like , no processing is done;

[0118] Count the total number of abnormal perpendicular segments and record it as , and calculate and Ratio , if the ratio Greater than the preset judgment value When the vertical points of all the vertical lines marked as abnormal foot segments that touch the defect key identification sub-area or the defect non-key identification sub-area are marked as abnormal points;

[0119] Repeat steps BS1-BS4 to determine each abnormal point marked as a defect focus identification sub-region or a defect non-focus identification sub-region;

[0120] The defect positioning module is used to determine the area of ​​the abnormal area according to the abnormal points of each defect key identification sub-area or defect non-key identification sub-area, and determine whether each defect key identification sub-area or defect non-key identification sub-area is a printing breakage defect area according to the area of ​​the abnormal area, and determine the specific location of the breakage defect in the printing breakage defect area.

[0121] Embodiment 3

[0122] The specific implementation process of this embodiment includes the entire implementation process of the above two groups of embodiments.

[0123] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0124] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for detecting printed product defects based on deep learning, characterized in that: include: Step 1: obtaining the entire printing surface area of ​​the printed matter, and dividing the entire printing surface area into a plurality of sub-areas; Step 2: According to the historical damage defect position records of the printed surface of the printed product, several sub-areas are classified into key defect sub-areas and non-key defect sub-areas; Step 3: Capture images of subsequent printed products, and simultaneously obtain a printing standard image, compare the captured image with the printing standard image, determine the image abnormality area of ​​the captured image, and mark the printed product corresponding to the image abnormality area to obtain the abnormal area of ​​the defective printed product; Step 4: based on the key defect sub-regions and the non-key defect sub-regions, determining in turn whether several sub-regions of the defective printed product are marked as key defect identification sub-regions or non-key defect identification sub-regions, and determining abnormal points for each sub-region marked as key defect identification sub-region or non-key defect identification sub-region; The specific contents of the abnormal point determination include: BS1: Determine one of the defect key identification sub-areas or defect non-key identification sub-areas, and establish a horizontal reference lower plane on the horizontal plane where the defect key identification sub-area or defect non-key identification sub-area is located, and the height from the horizontal reference lower plane is The position of the parallel reference upper plane is established, where It is the preset standard height; BS2: According to the specific location of the defect key identification sub-area or defect non-key identification sub-area, make a perpendicular line segments, where It is determined according to the defect focus identification sub-region or defect non-focus identification sub-region. If the sub-region is recorded as a defect focus identification sub-region, the larger k is, and if the sub-region is recorded as a defect non-focus identification sub-region, the larger k is. The smaller; BS3: Then count the height of each perpendicular line segment , get the height of each perpendicular line segment With standard height The absolute value of ,in, for Any one of the perpendicular segments, ; Where m = the area of ​​the defect focus identification sub-region or the defect non-focus identification sub-region × ; Then the absolute value With the preset threshold For comparison: like , mark the perpendicular line segment as an abnormal perpendicular line segment; like , no processing is done; BS4: Count the total number of abnormal perpendicular segments and record it as , and calculate and Ratio , if the ratio Greater than the preset judgment value When the vertical points of all the vertical lines marked as abnormal foot segments that touch the defect key identification sub-area or the defect non-key identification sub-area are marked as abnormal points; Step 5: Determine the area of ​​the abnormal region based on the abnormal points of each defect key identification sub-region or defect non-key identification sub-region, and determine whether each defect key identification sub-region or defect non-key identification sub-region is a printing breakage defect region based on the abnormal region area, and determine the specific location of the breakage defect in the printing breakage defect region.

2. The method for detecting printed product defects based on deep learning according to claim 1, characterized in that: In the step 2, the specific method of classifying the plurality of sub-regions into key defect sub-regions and non-key defect sub-regions according to the historical defect position record of the printed surface of the printed product is as follows: AS1: Starting from the current time, obtain the historical damage and defect location records of the printed surface of the printed products in the previous 30 days, and mark this period of time as the recording period; AS2: Count the total number of times each of several sub-areas is marked as a damage defect during the recording period and record it as ,in, , i represents the i-th sub-region, is the total number of sub-regions; AS3: Then pass the formula , determine the number of times each of several sub-areas was marked as a broken defect each day during the recording period , and Make a judgment: like , then the i-th sub-region is classified as a key defect region; like , then the i-th sub-region is classified as a non-key defect region, where Indicates the preset threshold.

3. The method for detecting printed product defects based on deep learning according to claim 2, characterized in that: In the step 3, determining the content of the abnormal area of ​​the image includes: An image that is inconsistent with a printing standard image is identified and marked as an abnormal printing image, and the printed product is marked as a defective print. The image inconsistency area between the abnormal printing image and the printing standard image is then determined and marked as an image abnormality area, and the image abnormality area is determined in the entire printing surface area of ​​the defective print.

4. The method for detecting printed product defects based on deep learning according to claim 1, characterized in that: The step BS4 further includes: BS5: Repeat steps BS1-BS4 to determine each abnormal point marked as a defect focus identification sub-region or a defect non-focus identification sub-region.

5. The method for detecting printed product defects based on deep learning according to claim 4, characterized in that: In the step 5, the abnormal area is determined according to the abnormal points of each defect key identification sub-region or defect non-key identification sub-region, and whether each defect key identification sub-region or defect non-key identification sub-region is a printing breakage defect region is determined according to the abnormal area, and the specific location of the breakage defect in the printing breakage defect region is determined in the following specific manner: CS1: After obtaining the abnormal points in one of the defect key identification sub-areas or defect non-key identification sub-areas, a rectangular coordinate system is established and the distance between two abnormal points is calculated using the following formula : ； In the formula, and For any two outliers, Expressed as Point horizontal coordinate, Expressed as Point vertical coordinate, Expressed as Point horizontal coordinate, Expressed as Point vertical coordinate; CS2: Then get the two outliers with the largest distance and mark the distance as , then determine the distance The center point of Draw a circle with the radius in the defect key identification sub-area or defect non-key identification sub-area, and determine the area of ​​the abnormal area by the following formula : ； In the formula, is an influencing factor, and its value is different in the defect focus identification sub-area and the defect non-focus identification sub-area; CS3: Get the preset reference threshold area and mark it as , the area of ​​the benchmark threshold region The area of ​​abnormal area Comparing and determining whether the defect key identification sub-region or the defect non-key identification sub-region is a printing breakage defect region, and determining the specific location of the breakage defect in the printing breakage defect region; CS4: Repeat steps CS1-CS3 to determine whether each defect key identification sub-region or defect non-key identification sub-region is a printing breakage defect region, and determine the specific location of the breakage defect in the printing breakage defect region.

6. The method for detecting printed product defects based on deep learning according to claim 5, characterized in that: In step CS3, the specific method of determining whether the defect focus identification sub-region or the defect non-focus identification sub-region is a printing breakage defect region, and determining the specific position of the breakage defect in the printing breakage defect region is: like When , it indicates the area of ​​abnormal region If the defect is too large, the defect key identification sub-region or the defect non-key identification sub-region is determined as a printing broken region; at the same time, the area of ​​the abnormal region is Determine the specific location of the broken defect as the printing broken defect area; like When , it indicates the area of ​​abnormal region Within an acceptable range, the defect focus identification sub-region or the defect non-focus identification sub-region is determined as a normal printing area.

7. A printed product defect detection system based on deep learning, applied to a printed product defect detection method based on deep learning as claimed in any one of claims 1 to 6, characterized in that: include: A region division module is used to obtain the entire printing surface area of ​​the printed matter and divide the entire printing surface area into a plurality of sub-areas; A region classification module is used to classify several sub-regions into key defect sub-regions and non-key defect sub-regions according to the historical damage defect position records of the printed surface of the printed product; The image comparison module is used to collect images of subsequent printed products and simultaneously obtain a printing standard image, compare the collected image with the printing standard image, determine the image abnormality area of ​​the collected image, and mark the printed product corresponding to the image abnormality area to obtain the abnormal area of ​​the defective printed product; An abnormality judgment module is used to judge whether several sub-regions of defective printed products are marked as defect key identification sub-regions or defect non-key identification sub-regions based on the key defect sub-regions and non-key defect sub-regions, and determine abnormal points for each sub-region marked as a defect key identification sub-region or a defect non-key identification sub-region; The specific contents of the abnormal point determination include: Determine one of the defect key identification sub-regions or defect non-key identification sub-regions, and establish a horizontal reference lower plane on the horizontal plane where the defect key identification sub-region or defect non-key identification sub-region is located, and the height from the horizontal reference lower plane is The position of the parallel reference upper plane is established, where It is the preset standard height; According to the specific location of the defect key identification sub-region or the defect non-key identification sub-region, a plane parallel to the reference is made from the plane parallel to the defect key identification sub-region or the defect non-key identification sub-region per area unit. perpendicular line segments, where It is determined according to the defect focus identification sub-region or defect non-focus identification sub-region. If the sub-region is recorded as a defect focus identification sub-region, the larger k is, and if the sub-region is recorded as a defect non-focus identification sub-region, the larger k is. The smaller; Then count the height of each perpendicular line segment , get the height of each perpendicular line segment With standard height The absolute value of ,in, for Any one of the perpendicular segments, ; Where m = the area of ​​the defect focus identification sub-region or the defect non-focus identification sub-region × ; Then the absolute value With the preset threshold For comparison: like , mark the perpendicular line segment as an abnormal perpendicular line segment; like , no processing is done; Count the total number of abnormal perpendicular segments and record it as , and calculate and Ratio , if the ratio Greater than the preset judgment value When the vertical points of all the vertical lines marked as abnormal foot segments that touch the defect key identification sub-area or the defect non-key identification sub-area are marked as abnormal points; The defect positioning module is used to determine the area of ​​the abnormal area according to the abnormal points of each defect key identification sub-area or defect non-key identification sub-area, and determine whether each defect key identification sub-area or defect non-key identification sub-area is a printing breakage defect area according to the area of ​​the abnormal area, and determine the specific location of the breakage defect in the printing breakage defect area.

Images