Parts identification and detection system and method
By designing a part recognition and detection system, and using image recognition technology to automatically detect and analyze sensor parts, the problem of low manual detection reliability is solved, efficient and reliable part recognition and repair is achieved, and the yield and quality of sensor production is improved.
Patent Information
- Application Number
- CN202411296378.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-18
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-09-18
AI Technical Summary
In the prior art, manual detection parts are low in reliability during sensor production, especially in multi-configured sensor parts, which can easily lead to identification fatigue and detection errors, which will lead to sensor configuration errors and after-sales problems.
A part recognition and detection system is designed, including a detection and identification module, an error type judgment module, a repair and allocation module and a cause troubleshooting module. The system realizes automated part recognition and repair by establishing a standard image library, real-time detection of the difference between part images and standard images, marks misalignment points, and judges the error type, and outputs corresponding repair processes.
It improves the reliability and efficiency of part inspection, reduces manual errors, realizes automatic identification and classification and rework processing of unqualified parts, avoids resource waste, and reduces computing storage space.
Smart Images

Figure CN118823014B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of parts identification, and in particular to a parts identification detection system and method. Background Art
[0002] Sensor technology plays a very important role in developing the economy and promoting social progress. With the rapid development of the sensor industry, there are more and more types of sensors, and the assembly parts of different types of sensors are also different. In order to avoid the occurrence of wrong or missing parts during sensor production, the general assembly workshop used to strengthen the pre-delivery personnel inspection to solve this problem. However, from the actual operation effect, the reliability of manual inspection is low, especially for sensor parts involving multiple configurations. Manual recognition fatigue causes detection errors, resulting in errors in the configuration parts of the sensor, which in turn causes a series of after-sales problems. Summary of the invention
[0003] The object of the present invention is to provide a parts identification and detection system and method to solve the problems raised in the above background technology.
[0004] In order to solve the above technical problems, the present invention provides the following technical solutions: a part identification and detection system, comprising a detection and identification module, an error type judgment module, a repair allocation module and a cause investigation module; the output end of the detection and identification module is connected to the input end of the error type judgment module, and the output end of the error type judgment module is connected to the repair allocation module and the input end of the repair allocation module;
[0005] A detection and recognition module, which is used to establish a standard library, obtain images of standard parts as reference images, collect real-time images of parts to be detected, compare the real-time images with the reference images, mark the differences between the two images as misalignment points, and mark parts with misalignment points as unqualified parts;
[0006] An error type judgment module, the error type judgment module is used to judge the error type of the misalignment point; when the number of misalignment points on the same part is one, it can be recorded as an error operation and the error type is output; when the number of misalignment points on the same part is multiple, the error type corresponding to each misalignment point is obtained, and it is judged whether there is an interchange relationship between each two misalignment points. If there is an interchange relationship and there is no other misalignment type, it is marked as an interchange type; if it is not an interchange relationship, the error type of each misalignment point is output;
[0007] A repair allocation module, which is used to obtain the error type of each dislocation point, output the repair process corresponding to each error type, and arrange the repair processes in the processing order according to the order of parts processing, and classify the unqualified products according to the order of the repair processes, and transport them to the corresponding maintenance storage device;
[0008] The cause investigation module is used to calculate the occurrence frequency of the corresponding error type. When the occurrence frequency is greater than the frequency threshold, all occurrence events of this type of error will be extracted, and the process cause that caused the error event will be output, and the process cause will be rectified.
[0009] The detection and identification module includes a picture library establishment unit and a dislocation point extraction unit;
[0010] The library building unit is used to obtain a 3D modeling image of a part, use the 3D modeling image as a primary standard image of the part, collect images of qualified standard parts, use the images of the standard parts as secondary standard images, extract features in the primary standard images, compare them with features in the secondary standard images, and obtain actual image error data; collect images of standard parts at time intervals T, store the images of the standard parts in a standard library, replace the original secondary standard images, and update the standard library;
[0011] The misalignment point extraction unit is used to collect a horizontal image of the part to be inspected, compare it with a first-level reference image in a standard library, mark the differences between the horizontal image and the first-level reference image, and determine whether they are within the actual image error data range. If not, the differences are marked as misalignment points, and edge lines of the misalignment points are extracted. The horizontal image is segmented, and the edge lines on both sides of the misalignment points are used as segmentation lines, which are extended to the ends of the horizontal image. At this time, a vertical image perpendicular to the horizontal image is collected on the product to be inspected, and the segmentation line in the horizontal image is extended to the vertical image. The vertical image is segmented according to the segmentation line, and the image within the segmentation line range is extracted for three-dimensional modeling. The area at the misalignment point forms a misalignment area model, and the part is marked as an unqualified part.
[0012] The error type judgment module repairs the process corresponding unit and the output unit;
[0013] The repair process corresponding unit is used to extract the historical error types in the historical database, and record the correction operations as repair processes, and correspond the error types to the repair processes one by one to form a repair set, which is input into the deep learning model for model training;
[0014] The output unit is used to obtain the error type of each misalignment point, extract the erroneous part information in the misalignment point, and simultaneously obtain the correct part information corresponding to the misalignment point. If the erroneous part information of one misalignment point is the same as the correct part information of another misalignment point, it is determined that the two misalignment points are in an interchangeable relationship. If multiple misalignment points are in an interchangeable relationship and there are no other error types, the part is marked as an interchangeable type. If the part is not an interchangeable type, the error type corresponding to the misalignment point on the part is output.
[0015] A parts identification and detection method is applied to a parts identification and detection system, and the detection method comprises the following steps:
[0016] Step S1, establish a standard library, obtain images of standard parts as reference images, collect real-time images of parts to be tested, compare the real-time images with the reference images, mark the differences between the two images as misalignment points, and mark parts with misalignment points as unqualified parts;
[0017] Step S2: When the number of misaligned points on the same part is one, it can be recorded as an erroneous operation and the error type can be output;
[0018] When there are multiple misalignment points on the same part, obtain the error type corresponding to each misalignment point, and determine whether there is an interchange relationship between every two misalignment points. If there is an interchange relationship and there is no other misalignment type, mark it as an interchange type; if there is no interchange relationship, output the error type of each misalignment point;
[0019] Step S3, after obtaining the error type of each dislocation point, output the repair process corresponding to each error type, and arrange the repair processes in the processing order according to the order of parts processing, and classify the unqualified products according to the order of the repair processes, and transport them to the maintenance storage device accordingly;
[0020] Step S4, calculate the occurrence frequency of the corresponding error type. When the occurrence frequency is greater than the frequency threshold, extract all occurrence events of this type of error, output the process cause that causes the error event, and rectify the process cause.
[0021] The step S1 comprises:
[0022] Step S101, obtaining a 3D modeling image of a part, using the 3D modeling image as a primary standard image of the part, collecting an image of a qualified standard part, using the image of the standard part as a secondary standard image, extracting features in the primary standard image, and comparing them with features in the secondary standard image to obtain actual image error data;
[0023] Step S102, setting a time interval T, collecting an image of a standard part at the time interval T, and storing the image of the standard part in a standard library, replacing the original secondary standard image, and updating the standard library;
[0024] Step S103, collect a horizontal image of the part to be inspected, compare it with the first-level reference image in the standard library, mark the differences between the horizontal image and the first-level reference image, and determine whether they are within the actual image error data range. If not, mark the differences as misaligned points, extract the edge lines of the misaligned points, segment the horizontal image, use the edge lines on both sides of the misaligned points as segmentation lines, and extend them to the end of the horizontal image. At this time, collect a vertical image of the product to be inspected that is perpendicular to the horizontal image, extend the segmentation line in the horizontal image to the vertical image, segment the vertical image according to the segmentation line, extract the image within the segmentation line range, perform three-dimensional modeling, form a misaligned area model for the area at the misaligned point, and mark the part as an unqualified part.
[0025] The step S2 comprises:
[0026] Step S201, extract the historical error types in the historical database, and record the correction operations as repair processes, correspond the error types to the repair processes one by one, form a repair set, input it into the deep learning model, and perform model training;
[0027] Step S202, extract all the misalignment points on the unqualified parts. If the number of misalignment points is one, determine the error type of the misalignment point according to the constructed misalignment area model, input the error type into the deep learning model, and output the repair process corresponding to the misalignment type; the misalignment types include diameter size error, length size error, installation position error, installation angle error and appearance error;
[0028] Step S203: if there are multiple misalignment points on the unqualified part, obtain the error type of each misalignment point, extract the error part information in the misalignment point, and simultaneously obtain the correct part information corresponding to the misalignment point; if the error part information of one misalignment point is the same as the correct part information of another misalignment point, then it is determined that the two misalignment points are in an interchangeable relationship; if multiple misalignment points are in an interchangeable relationship and there is no other error type, then the part is marked as an interchangeable type;
[0029] Step S203: If the part is not of an interchangeable type, the error type corresponding to the misalignment point on the part is output.
[0030] The step S3 comprises:
[0031] Output the repair process corresponding to the error type, and match the repair process with the processing process one by one. Arrange the repair processes in sequence according to the time sequence of the parts' processing processes, and classify the unqualified products belonging to the same repair process into the same area, and output them to the designated maintenance storage device accordingly.
[0032] The step S4 comprises:
[0033] The error type classification includes {error type 1, error type 2, error type 3...error type n}, and the occurrence frequency of each error type classification is recorded as {(error type 1, P1), (error type 2, P2), (error type 3, P3),......., (error type n, P n )}, calculate the frequency P of each error type n , the calculation formula is P n =m / M, where M represents the total number of all error types that occur, and m represents the number of times this error type occurs. Set the frequency threshold to P 阈 , if P n >P 阈 , then output the repair process corresponding to the error type, obtain the processing process corresponding to the repair process, take the processing process as the process cause, and output the result of maintenance and detection of the processing process corresponding to the process cause.
[0034] Compared with the prior art, the beneficial effects achieved by the present invention are:
[0035] The present invention utilizes visual recognition and image recognition technology to detect and analyze the parts of the product, determine whether there are installation errors and whether the number of parts is normal, so as to ensure the yield of the product after production. At the same time, for the position with incorrect installation, the error type can be further determined, and the parts that are installed upside down with two misaligned points can be separately identified. The misalignment can be repaired manually without having to be put back on the production line, thereby achieving classified rework of the parts and avoiding waste of resources caused by repeated processing. A three-dimensional model of the specified error area can also be generated for the misaligned point, which is convenient for determining the error type and also reduces the space for computing and storage. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0037] Figure 1 It is a flow chart of a parts identification and detection method of the present invention. DETAILED DESCRIPTION
[0038] 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.
[0039] Embodiment 1, the present invention provides a technical solution: a part identification and detection system, comprising a detection and identification module, an error type judgment module, a repair allocation module and a cause investigation module; the output end of the detection and identification module is connected to the input end of the error type judgment module, the output end of the error type judgment module is connected to the repair allocation module and the input end of the repair allocation module;
[0040] A detection and recognition module, which is used to establish a standard library, obtain images of standard parts as reference images, collect real-time images of parts to be detected, compare the real-time images with the reference images, mark the differences between the two images as misalignment points, and mark parts with misalignment points as unqualified parts;
[0041] An error type judgment module, the error type judgment module is used to judge the error type of the misalignment point; when the number of misalignment points on the same part is one, it can be recorded as an error operation and the error type is output; when the number of misalignment points on the same part is multiple, the error type corresponding to each misalignment point is obtained, and it is judged whether there is an interchange relationship between each two misalignment points. If there is an interchange relationship and there is no other misalignment type, it is marked as an interchange type; if it is not an interchange relationship, the error type of each misalignment point is output;
[0042] A repair allocation module, which is used to obtain the error type of each dislocation point, output the repair process corresponding to each error type, and arrange the repair processes in the processing order according to the order of parts processing, and classify the unqualified products according to the order of the repair processes, and transport them to the corresponding maintenance storage device;
[0043] The cause investigation module is used to calculate the occurrence frequency of the corresponding error type. When the occurrence frequency is greater than the frequency threshold, all occurrence events of this type of error will be extracted, and the process cause that caused the error event will be output, and the process cause will be rectified.
[0044] The detection and identification module includes a picture library establishment unit and a dislocation point extraction unit;
[0045] The library building unit is used to obtain a 3D modeling image of a part, use the 3D modeling image as a primary standard image of the part, collect images of qualified standard parts, use the images of the standard parts as secondary standard images, extract features in the primary standard images, compare them with features in the secondary standard images, and obtain actual image error data; collect images of standard parts at time intervals T, store the images of the standard parts in a standard library, replace the original secondary standard images, and update the standard library;
[0046] The misalignment point extraction unit is used to collect a horizontal image of the part to be inspected, compare it with a first-level reference image in a standard library, mark the differences between the horizontal image and the first-level reference image, and determine whether they are within the actual image error data range. If not, the differences are marked as misalignment points, and edge lines of the misalignment points are extracted. The horizontal image is segmented, and the edge lines on both sides of the misalignment points are used as segmentation lines, which are extended to the ends of the horizontal image. At this time, a vertical image perpendicular to the horizontal image is collected on the product to be inspected, and the segmentation line in the horizontal image is extended to the vertical image. The vertical image is segmented according to the segmentation line, and the image within the segmentation line range is extracted for three-dimensional modeling. The area at the misalignment point forms a misalignment area model, and the part is marked as an unqualified part.
[0047] The error type judgment module repairs the process corresponding unit and the output unit;
[0048] The repair process corresponding unit is used to extract the historical error types in the historical database, and record the correction operations as repair processes, and correspond the error types to the repair processes one by one to form a repair set, which is input into the deep learning model for model training;
[0049] The output unit is used to obtain the error type of each misalignment point, extract the erroneous part information in the misalignment point, and simultaneously obtain the correct part information corresponding to the misalignment point. If the erroneous part information of one misalignment point is the same as the correct part information of another misalignment point, it is determined that the two misalignment points are in an interchangeable relationship. If multiple misalignment points are in an interchangeable relationship and there are no other error types, the part is marked as an interchangeable type. If the part is not an interchangeable type, the error type corresponding to the misalignment point on the part is output.
[0050] For example 2, please refer to Figure 1 A part recognition and detection method is applied to a part recognition and detection system, and the detection method comprises the following steps:
[0051] Step S1, establish a standard library, obtain images of standard parts as reference images, collect real-time images of parts to be tested, compare the real-time images with the reference images, mark the differences between the two images as misalignment points, and mark parts with misalignment points as unqualified parts;
[0052] Step S2: When the number of misaligned points on the same part is one, it can be recorded as an erroneous operation and the error type can be output;
[0053] When there are multiple misalignment points on the same part, obtain the error type corresponding to each misalignment point, and determine whether there is an interchange relationship between every two misalignment points. If there is an interchange relationship and there is no other misalignment type, mark it as an interchange type; if there is no interchange relationship, output the error type of each misalignment point;
[0054] Step S3, after obtaining the error type of each dislocation point, output the repair process corresponding to each error type, and arrange the repair processes in the processing order according to the order of parts processing, and classify the unqualified products according to the order of the repair processes, and transport them to the maintenance storage device accordingly;
[0055] Step S4, calculate the occurrence frequency of the corresponding error type. When the occurrence frequency is greater than the frequency threshold, extract all occurrence events of this type of error, output the process cause that causes the error event, and rectify the process cause.
[0056] The step S1 comprises:
[0057] Step S101, obtaining a 3D modeling image of a part, using the 3D modeling image as a primary standard image of the part, collecting an image of a qualified standard part, using the image of the standard part as a secondary standard image, extracting features in the primary standard image, and comparing them with features in the secondary standard image to obtain actual image error data;
[0058] Step S102, setting a time interval T, collecting an image of a standard part at the time interval T, and storing the image of the standard part in a standard library, replacing the original secondary standard image, and updating the standard library;
[0059] Step S103, collect a horizontal image of the part to be inspected, compare it with the first-level reference image in the standard library, mark the differences between the horizontal image and the first-level reference image, and determine whether they are within the actual image error data range. If not, mark the differences as misaligned points, extract the edge lines of the misaligned points, segment the horizontal image, use the edge lines on both sides of the misaligned points as segmentation lines, and extend them to the end of the horizontal image. At this time, collect a vertical image of the product to be inspected that is perpendicular to the horizontal image, extend the segmentation line in the horizontal image to the vertical image, segment the vertical image according to the segmentation line, extract the image within the segmentation line range, perform three-dimensional modeling, form a misaligned area model for the area at the misaligned point, and mark the part as an unqualified part.
[0060] For example, first take a photo of the part from the front as a horizontal image. If a misalignment point appears in the horizontal image, the type of the misalignment point is further determined. Since the horizontal image can only determine two-dimensional problems such as the end diameter, horizontal position, and color, it is necessary to establish a three-dimensional image to further determine the specific misalignment type. At this time, the image of the part can be collected again from the vertical direction, such as the top, side, and bottom of the part, as a vertical image. First, the horizontal image and the vertical image are constructed in a three-dimensional coordinate system, and then the edge line of the misalignment point is used as the dividing line to divide the horizontal image and the vertical image, retaining only the part within the dividing line covering the misalignment point, and establishing a misalignment module for the misalignment area corresponding to the misalignment point.
[0061] The step S2 comprises:
[0062] Step S201, extract the historical error types in the historical database, and record the correction operations as repair processes, correspond the error types to the repair processes one by one, form a repair set, input it into the deep learning model, and perform model training;
[0063] Step S202, extract all the misalignment points on the unqualified parts. If the number of misalignment points is one, determine the error type of the misalignment point according to the constructed misalignment area model, input the error type into the deep learning model, and output the repair process corresponding to the misalignment type; the misalignment types include diameter size error, length size error, installation position error, installation angle error and appearance error;
[0064] Step S203: if there are multiple misalignment points on the unqualified part, obtain the error type of each misalignment point, extract the error part information in the misalignment point, and simultaneously obtain the correct part information corresponding to the misalignment point; if the error part information of one misalignment point is the same as the correct part information of another misalignment point, then it is determined that the two misalignment points are in an interchangeable relationship; if multiple misalignment points are in an interchangeable relationship and there is no other error type, then the part is marked as an interchangeable type;
[0065] Step S203: If the part is not of an interchangeable type, the error type corresponding to the misalignment point on the part is output.
[0066] The step S3 comprises:
[0067] Output the repair process corresponding to the error type, and match the repair process with the processing process one by one. Arrange the repair processes in sequence according to the time sequence of the parts' processing processes, and classify the unqualified products belonging to the same repair process into the same area, and output them to the designated maintenance storage device accordingly.
[0068] The step S4 comprises:
[0069] The error type classification includes {error type 1, error type 2, error type 3...error type n}, and the occurrence frequency of each error type classification is recorded as {(error type 1, P1), (error type 2, P2), (error type 3, P3),......., (error type n, P n )}, calculate the frequency P of each error type n , the calculation formula is P n =m / M, where M represents the total number of all error types that occur, and m represents the number of times this error type occurs. Set the frequency threshold to P 阈 , if P n >P 阈 , then output the repair process corresponding to the error type, obtain the processing process corresponding to the repair process, take the processing process as the process cause, and output the result of maintenance and detection of the processing process corresponding to the process cause.
[0070] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A parts identification and detection method, characterized in that: The detection method comprises the following steps: Step S1, establish a standard library, obtain images of standard parts as reference images, collect real-time images of parts to be tested, compare the real-time images with the reference images, mark the differences between the two images as misalignment points, and mark parts with misalignment points as unqualified parts; Step S2: When the number of misaligned points on the same part is one, it can be recorded as an erroneous operation and the error type can be output; When there are multiple misalignment points on the same part, obtain the error type corresponding to each misalignment point, and determine whether there is an interchange relationship between every two misalignment points. If there is an interchange relationship and there is no other misalignment type, mark it as an interchange type; if there is no interchange relationship, output the error type of each misalignment point; Step S3, after obtaining the error type of each dislocation point, output the repair process corresponding to each error type, and arrange the repair processes in the processing order according to the order of parts processing, and classify the unqualified products according to the order of the repair processes, and transport them to the maintenance storage device accordingly; Step S4, calculating the occurrence frequency of the corresponding error type. When the occurrence frequency is greater than the frequency threshold, all occurrence events of the error type are extracted, the process cause leading to the error event is output, and the process cause is rectified; The step S1 comprises: Step S101, obtaining a 3D modeling image of a part, using the 3D modeling image as a primary standard image of the part, collecting an image of a qualified standard part, using the image of the standard part as a secondary standard image, extracting features in the primary standard image, and comparing them with features in the secondary standard image to obtain actual image error data; Step S102, setting a time interval T, collecting an image of a standard part at the time interval T, and storing the image of the standard part in a standard library, replacing the original secondary standard image, and updating the standard library; Step S103, collect a horizontal image of the part to be inspected, compare it with the first-level standard image in the standard library, mark the differences between the horizontal image and the first-level standard image, and determine whether they are within the actual image error data range. If not, mark the differences as misaligned points, extract edge lines of the misaligned points, segment the horizontal image, use the edge lines on both sides of the misaligned points as segmentation lines, and extend them to the ends of the horizontal image. At this time, collect a vertical image of the product to be inspected that is perpendicular to the horizontal image, extend the segmentation lines in the horizontal image to the vertical image, segment the vertical image according to the segmentation lines, extract the image within the segmentation line range, perform three-dimensional modeling, form a misaligned area model for the area at the misaligned point, and mark the part as an unqualified part; The step S2 comprises: Step S201, extract the historical error types in the historical database, and record the correction operations as repair processes, correspond the error types to the repair processes one by one, form a repair set, input it into the deep learning model, and perform model training; Step S202, extract all the misalignment points on the unqualified parts. If the number of misalignment points is one, determine the error type of the misalignment point according to the constructed misalignment area model, input the error type into the deep learning model, and output the repair process corresponding to the misalignment type; the misalignment types include diameter size error, length size error, installation position error, installation angle error and appearance error; Step S203: if there are multiple misalignment points on the unqualified part, obtain the error type of each misalignment point, extract the error part information in the misalignment point, and simultaneously obtain the correct part information corresponding to the misalignment point; if the error part information of one misalignment point is the same as the correct part information of another misalignment point, then it is determined that the two misalignment points are in an interchangeable relationship; if multiple misalignment points are in an interchangeable relationship and there is no other error type, then the part is marked as an interchangeable type; Step S203: If the part is not of an interchangeable type, the error type corresponding to the misalignment point on the part is output.
2. A parts identification and detection method according to claim 1, characterized in that: The step S3 comprises: Output the repair process corresponding to the error type, and match the repair process with the processing process one by one. Arrange the repair processes in sequence according to the time sequence of the parts' processing processes, and classify the unqualified products belonging to the same repair process into the same area, and output them to the designated maintenance storage device accordingly.
3. A parts identification and detection method according to claim 2, characterized in that: The step S4 comprises: The error type classification includes {error type 1, error type 2, error type 3...error type n}, and the occurrence frequency of each error type classification is recorded as {(error type 1, P1), (error type 2, P2), (error type 3, P3),......., (error type n, P n )}, calculate the frequency P of each error type n , the calculation formula is P n =m / M, where M represents the total number of all error types that occur, and m represents the number of times this error type occurs. Set the frequency threshold to P 阈 , if P n >P 阈 , then output the repair process corresponding to the error type, obtain the processing process corresponding to the repair process, take the processing process as the process cause, and output the result of maintenance and detection of the processing process corresponding to the process cause.
4. A parts identification and detection system, applied to a parts identification and detection method according to any one of claims 1 to 3, characterized in that: It includes a detection and identification module, an error type judgment module, a repair allocation module and a cause investigation module; the output end of the detection and identification module is connected to the input end of the error type judgment module, and the output end of the error type judgment module is connected to the repair allocation module and the input end of the repair allocation module; A detection and recognition module, which is used to establish a standard library, obtain images of standard parts as reference images, collect real-time images of parts to be detected, compare the real-time images with the reference images, mark the differences between the two images as misalignment points, and mark parts with misalignment points as unqualified parts; An error type judgment module, the error type judgment module is used to judge the error type of the misalignment point; when the number of misalignment points on the same part is one, it can be recorded as an error operation and the error type can be output; When there are multiple misalignment points on the same part, obtain the error type corresponding to each misalignment point, and determine whether there is an interchange relationship between every two misalignment points. If there is an interchange relationship and there is no other misalignment type, mark it as an interchange type. If it is not an interchange relationship, then output the error type of each misalignment point; A repair allocation module, which is used to obtain the error type of each dislocation point, output the repair process corresponding to each error type, and arrange the repair processes in the processing order according to the order of parts processing, and classify the unqualified products according to the order of the repair processes, and transport them to the corresponding maintenance storage device; The cause investigation module is used to calculate the occurrence frequency of the corresponding error type. When the occurrence frequency is greater than the frequency threshold, all occurrence events of this type of error will be extracted, and the process cause that caused the error event will be output, and the process cause will be rectified.
5. A parts identification and detection system according to claim 4, characterized in that: The detection and identification module includes a picture library establishment unit and a dislocation point extraction unit; The library building unit is used to obtain a 3D modeling image of a part, use the 3D modeling image as a primary standard image of the part, collect images of qualified standard parts, use the images of the standard parts as secondary standard images, extract features in the primary standard images, compare them with features in the secondary standard images, and obtain actual image error data; collect images of standard parts at time intervals T, store the images of the standard parts in a standard library, replace the original secondary standard images, and update the standard library; The misalignment point extraction unit is used to collect a horizontal image of the part to be inspected, compare it with the first-level standard image in the standard library, mark the differences between the horizontal image and the first-level standard image, and determine whether they are within the actual image error data range. If not, the differences are marked as misalignment points, and the edge lines of the misalignment points are extracted. The horizontal image is segmented, and the edge lines on both sides of the misalignment points are used as segmentation lines, which are extended to the ends of the horizontal image. At this time, a vertical image perpendicular to the horizontal image is collected on the product to be inspected, and the segmentation line in the horizontal image is extended to the vertical image. The vertical image is segmented according to the segmentation line, and the image within the segmentation line range is extracted for three-dimensional modeling. The area at the misalignment point forms a misalignment area model, and the part is marked as an unqualified part.
6. A parts identification and detection system according to claim 5, characterized in that: The error type judgment module repairs the process corresponding unit and the output unit; The repair process corresponding unit is used to extract the historical error types in the historical database, and record the correction operations as repair processes, and correspond the error types to the repair processes one by one to form a repair set, which is input into the deep learning model for model training; The output unit is used to obtain the error type of each misalignment point, extract the erroneous part information in the misalignment point, and simultaneously obtain the correct part information corresponding to the misalignment point. If the erroneous part information of one misalignment point is the same as the correct part information of another misalignment point, it is determined that the two misalignment points are in an interchangeable relationship. If multiple misalignment points are in an interchangeable relationship and there are no other error types, the part is marked as an interchangeable type. If the part is not an interchangeable type, the error type corresponding to the misalignment point on the part is output.
Citation Information
Patent Citations
Defective product analysis method and system for intelligent manufacturing
CN114219799A
Product quality detection equipment and detection method based on visual detection
CN118464914A
