Part machining error detection system and method based on machine vision

Abstract

The invention discloses a part machining error detection method based on machine vision, and relates to the technical field of part machining detection. According to the method, the part machining process is subjected to stage division, and the time data of each stage is acquired, so that each stage is ensured to be completed within a reasonable time range, and the quality problem caused by too long time is avoided. Meanwhile, the method also combines a plurality of steps of image acquisition, image preprocessing, feature extraction, error detection and the like so as to comprehensively evaluate the processing quality of the parts. In the aspects of feature extraction and error detection, the method adopts an advanced algorithm model, can accurately identify different types of errors such as size errors, deformation errors and surface flaws of parts, and calculates the severity and comprehensive evaluation coefficients of the errors. Finally, the method also provides effective quality control and fault early warning means for manufacturing enterprises through the steps of result output, troubleshooting, unqualified part analysis and the like.

Description

Technical Field

[0001] The present invention belongs to the technical field of parts processing detection, and in particular relates to a parts processing error detection system and method based on machine vision. Background Art

[0002] In the manufacturing industry, the processing quality of parts directly affects the performance and reliability of the final product. Traditional parts processing error detection methods mainly rely on manual detection, which is not only time-consuming and labor-intensive, but also easily affected by human factors, making it difficult to ensure the accuracy and consistency of the detection results. With the rapid development of machine vision technology, its high precision and high efficiency make it show great application potential in parts processing error detection.

[0003] However, there are still some problems with the existing machine vision-based component processing error detection methods. On the one hand, these methods often only focus on the final detection results, while ignoring the various stages of the component processing process. Since there may be various uncertainties in the processing process, such as equipment failure, material changes, etc., these factors may affect the processing quality of the components. Therefore, only testing the final results is not enough to fully evaluate the processing quality of the components.

[0004] On the other hand, existing detection methods also have limitations in feature extraction and error detection. For example, some methods may only focus on the dimensional error of parts, while ignoring other aspects such as deformation error and surface defects. In addition, even for the same type of error, existing detection methods may have difficulty in accurately identifying its severity and type due to the lack of effective algorithm models. Summary of the invention

[0005] In order to overcome the shortcomings and deficiencies of the above-mentioned prior art, the first purpose of the present invention is to provide a component processing error detection system based on machine vision; the second purpose of the present invention is to provide a component processing error detection method based on machine vision.

[0006] The first object of the present invention adopts the following technical solution:

[0007] A machine vision-based component processing error detection system includes a stage management and time monitoring module, an image acquisition and preprocessing module, a feature extraction and error detection module, a result output and troubleshooting module, and a non-conforming component analysis module;

[0008] Stage management and time monitoring module: divides the parts processing process into multiple stages, marks these stages, analyzes time data, determines whether to generate preliminary stage switching instructions, and ensures that each stage is completed within a reasonable time by comparing with the preset time threshold;

[0009] Image acquisition and preprocessing module: Use image capture equipment to capture multi-view images of parts, preprocess the acquired images, and collect production process data of each part, including production links, production machine or container numbers, and production time;

[0010] Feature extraction and error detection module: extract the features of parts, detect dimensional errors, deformation errors and surface defect errors, calculate the comprehensive evaluation coefficient, and determine whether the parts are qualified according to the preset defect threshold;

[0011] Result output and troubleshooting module: output the information of qualified parts and generate destruction marks and related information for unqualified parts; for unqualified parts, obtain their production process data, analyze the failure of production equipment or containers, and determine whether troubleshooting or quality inspection is needed based on the preset production failure threshold and production quantity difference threshold, and perform troubleshooting or quality inspection on production equipment or containers in the order of numbers;

[0012] Defective parts analysis module: Analyze the number of defective parts, calculate the anomaly ratio and standard deviation, determine whether the abnormal type of the defective parts is sporadic or long-term, generate fault warning instructions, compare the statistical number of defective parts with the preset threshold, and mark the normal period or abnormal period.

[0013] The second object of the present invention adopts the following technical solution:

[0014] A component processing error detection method based on machine vision is used to implement a component processing error detection system based on machine vision. The method flow is as follows:

[0015] Step 1: Divide the parts processing process into multiple stages and mark them as different stages, including image acquisition stage, image preprocessing stage, feature extraction stage, error detection stage and result output stage;

[0016] Step 2: Collect the time data of each stage, analyze the time data, and determine whether to generate a preliminary stage switching instruction;

[0017] Step 3: If a preliminary stage switching instruction is generated, the image acquisition stage is entered, and a multi-view image of the component is captured using an image capture device and pre-processed;

[0018] Step 4: Perform feature extraction and error detection, including dimensional error detection, deformation error detection and surface defect detection. By calculating the dimensional error coefficient, deformation error coefficient and surface defect coefficient, as well as the comprehensive evaluation coefficient, determine whether the parts are qualified;

[0019] Step 5: Output the information of qualified or unqualified parts according to the test results. For unqualified parts, generate a destruction mark and obtain its production process data for analysis to determine whether troubleshooting or quality inspection is needed;

[0020] Step 6: Analyze the number of unqualified parts, calculate the anomaly ratio and standard deviation, determine the abnormal type of unqualified parts, and generate corresponding fault warning instructions.

[0021] Preferably, the image preprocessing in step 1 includes grayscale conversion and Gaussian filtering to reduce noise and improve image quality.

[0022] Preferably, the determination of whether to generate a preliminary stage switching instruction in step 2 is specifically as follows:

[0023] The number of statistical stages is n1, and n2 time thresholds are preset, n1=n2;

[0024] The collected time data is compared with the corresponding time threshold; if the time data is greater than or equal to the time threshold, a preliminary stage switching instruction is generated; otherwise, no preliminary stage switching instruction is generated.

[0025] Preferably, the multi-view image in step three includes a top view, a side view and a bottom view.

[0026] Preferably, the dimensional error detection in step 4 is achieved by comparing the image of the key part of the component with the standard image; the deformation error detection is achieved by intercepting the key shape of the component from different views and comparing it with the standard shape; the surface defect detection is achieved by training a multi-target recognition neural network model to identify surface defects and their severity; specifically as follows:

[0027] S301, dimensional error detection;

[0028] Cut out the key parts of the parts from the image, including length, width, and diameter, compare the cut out image with the standard image, and calculate the size error coefficient S p , S p The calculation formula is as follows: Among them, A actual is the actual size; A standard It is a standard size;

[0029] S302, deformation error detection;

[0030] Cut out the key shapes of parts from different views, such as the shape of the bottle mouth in the top view and the shape of the bottle body in the side view, compare the cut shapes with the standard shapes, and calculate the deformation error coefficient S x , S x The calculation formula is as follows: Among them, P actual，i is the coordinate of the ith point of the actual shape; P standard，i is the coordinate of the i-th point of the standard shape, and n is the number of points;

[0031] S303, surface defect detection;

[0032] Collect images of parts with surface defects and parts without surface defects in advance and annotate them; train a multi-target recognition neural network model to identify surface defects and their severity; calculate the surface defect coefficient S f , S f The calculation formula is as follows: Among them, ω i is the weight of the i-th defect point; I i is the severity of the i-th defect point; m is the number of defect points;

[0033] S304, calculate the comprehensive evaluation coefficient S total , S total The calculation formula is as follows: total =αS p +βS x +γS f ; where α, β and γ are S p , S x and S f Preset scaling factor;

[0034] According to the preset defect threshold X, determine whether the component is qualified. If S total If S is less than or equal to X, the component is considered qualified; total If it is greater than X, the component is deemed unqualified and a destruction mark is generated.

[0035] As a preference, step five is specifically as follows:

[0036] Qualified parts: Output the information of qualified parts;

[0037] Unqualified parts: Generate destruction marks and output information of unqualified parts;

[0038] For the parts that generate destruction marks, obtain their production process data, including the production link, production machine or production container number and production time;

[0039] Analyze the failure of production equipment or production containers, calculate the proportion of the number of occurrences of each production equipment or production container number, sort them from large to small according to the proportion, and save them in the set V s middle;

[0040] Count the number of parts V marked for destructions and the total number of parts T s , according to the preset production failure threshold Q s and production quantity difference threshold Q d , determine whether troubleshooting or quality inspection is needed;

[0041] Troubleshooting: If Perform troubleshooting on production equipment or production containers in the order of numbers in set Vs;

[0042] Quality inspection: If and For production equipment or production containers according to set V s Quality inspection is carried out in the order of numbers in the

[0043] As a preference, step six is ​​specifically as follows:

[0044] Analyze the number of defective parts, calculate the anomaly ratio and standard deviation, and determine the abnormal type of defective parts;

[0045] Get the number of unqualified parts in period t+1, where period t+1 is the next period after time t;

[0046] The number of unqualified parts counted is compared with the preset threshold of unqualified parts. If the number of unqualified parts is less than the threshold of unqualified parts, it is marked as a normal period; if the number of unqualified parts is greater than or equal to the threshold of unqualified parts, it is marked as an abnormal period;

[0047] The ratio of the number of marks in the abnormal period to the number of unqualified parts in the t+1 period is marked as the abnormal ratio; the standard deviation of the data set of all unqualified parts in the t+1 period is calculated;

[0048] If the anomaly ratio is less than or equal to the anomaly threshold, and the standard deviation is less than or equal to the standard deviation threshold, it is marked as an occasional anomaly; otherwise, it is marked as a long-term anomaly and a fault warning instruction is generated.

[0049] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0050] 1. The present invention divides the parts processing process into multiple clear stages, sets a time threshold for each stage, collects and analyzes the time data of each stage in real time, ensures that each stage is completed within a reasonable time range, and effectively prevents quality problems caused by timeout. This staged time management and monitoring mechanism is not common in traditional machine vision inspection, and improves the controllability and efficiency of the production process.

[0051] 2. The present invention not only focuses on the dimensional error and deformation error of parts, but also introduces surface defect detection, using a multi-target recognition neural network model to identify surface defects and their severity. By calculating the comprehensive evaluation coefficient, the impact of various error types on the quality of parts is comprehensively considered, achieving a comprehensive evaluation of the quality of parts.

[0052] 3. The present invention intelligently determines whether troubleshooting or quality inspection is needed based on the production process data and quantity statistics of unqualified parts. This intelligent decision-making mechanism can timely discover potential production problems, reduce the production of unqualified products, and improve production efficiency and product quality; by calculating the anomaly ratio and standard deviation, the method can determine the abnormal type of unqualified parts (occasional abnormality or long-term abnormality) and generate fault warning instructions. This abnormal type determination mechanism helps to quickly locate the root cause of component error problems and take targeted measures to improve them. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] 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.

[0054] Figure 1 A module diagram of a component processing error detection system based on machine vision of the present invention is shown;

[0055] Figure 2 A flow chart of a method for detecting component processing errors based on machine vision is shown in the present invention;

[0056] Figure 3 A flow chart of feature extraction and error detection of the present invention is shown. DETAILED DESCRIPTION

[0057] 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.

[0058] In addition, the described features, structures or characteristics may be combined in one or more example embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the example embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, steps, etc. may be adopted. In other cases, well-known structures, methods, implementations or operations are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0059] Embodiment 1:

[0060] See also Figure 1 As shown, a machine vision-based component processing error detection system of this embodiment includes a stage management and time monitoring module, an image acquisition and preprocessing module, a feature extraction and error detection module, a result output and troubleshooting module, and an unqualified component analysis module.

[0061] Stage management and time monitoring module: Divide the parts processing process into multiple stages, mark these stages, analyze time data, determine whether to generate preliminary stage switching instructions, and ensure that each stage is completed within a reasonable time by comparing with the preset time threshold.

[0062] Image acquisition and preprocessing module: Use image capture equipment to capture multi-view images of parts (top view, side view, bottom view), preprocess the acquired images, including grayscale conversion, Gaussian filtering, etc., to reduce noise and improve image quality, and collect production process data for each part, such as production link, production machine or container number, and production time.

[0063] Feature extraction and error detection module: extracts the features of parts and detects size, deformation and surface defect errors, calculates the comprehensive evaluation coefficient, and determines whether the parts are qualified based on the preset defect threshold.

[0064] Result output and troubleshooting module: output the information of qualified parts and generate destruction marks and related information for unqualified parts; for unqualified parts, obtain their production process data, analyze the failure of production equipment or containers, and determine whether troubleshooting or quality inspection is needed based on the preset production failure threshold and production quantity difference threshold, and perform troubleshooting or quality inspection on production equipment or containers in order of numbering.

[0065] Defective parts analysis module: Analyze the number of defective parts, calculate the anomaly ratio and standard deviation, determine the abnormal type of the defective parts (occasional anomaly or long-term anomaly), and generate fault warning instructions, compare the statistical number of defective parts with the preset threshold, and mark the normal period or abnormal period.

[0066] The beneficial effects of this embodiment include: achieving refined management of the processing process and improving image quality; accurately detecting errors and ensuring product quality; effectively outputting detection results and guiding troubleshooting; optimizing the production process, reducing the failure rate, and improving overall production efficiency and product quality management level.

[0067] Embodiment 2:

[0068] See also Figure 2 As shown, a method for detecting component processing errors based on machine vision in this embodiment has the following process:

[0069] Step 1: Divide the parts processing process into stages and mark them into different stages, including image acquisition stage, image preprocessing stage, feature extraction stage, error detection stage and result output stage.

[0070] The time data of each stage is collected, and each time data corresponds to one stage.

[0071] The time data is analyzed to determine whether a preliminary stage switching instruction is generated.

[0072] The judgment method is as follows: count the number of stages n1, preset n2 time thresholds, n1 = n2;

[0073] The collected time data is compared with the corresponding time threshold; if the time data is greater than or equal to the time threshold, a preliminary stage switching instruction is generated; otherwise, no preliminary stage switching instruction is generated.

[0074] Preliminary stage switching instructions ensure that each stage is completed within a reasonable time frame to avoid quality problems caused by excessive time.

[0075] Step 2: If a preliminary stage switching instruction is generated, the image acquisition stage is entered.

[0076] Use image capture equipment to capture multi-view images of parts, including top view, side view and bottom view, and pre-process the acquired images, including grayscale conversion, Gaussian filtering, etc., to reduce noise and improve image quality.

[0077] Before testing, collect the production process data of each component, including the production link, production machine or production container number and production time.

[0078] Step 3: Feature extraction and error detection.

[0079] See also Figure 3 As shown, feature extraction and error detection are as follows:

[0080] S301, dimensional error detection.

[0081] Cut out the key parts of the parts from the image, such as length, width, diameter, etc., compare the cut out image with the standard image, and calculate the size error coefficient S p , S p The calculation formula is as follows:

[0082] Among them, A actual is the actual size; A standard It is a standard size.

[0083] S302, deformation error detection:

[0084] Cut out the key shapes of parts from different views, such as the shape of the bottle mouth in the top view and the shape of the bottle body in the side view, compare the cut shapes with the standard shapes, and calculate the deformation error coefficient S x , S x The calculation formula is as follows:

[0085] Among them, P actual，i is the coordinate of the ith point of the actual shape; P standard，i are the coordinates of the i-th point of the standard shape, and n is the number of points.

[0086] S303: Surface defect detection.

[0087] Images of parts with surface defects and images of parts without surface defects are collected in advance and annotated.

[0088] Train a multi-object recognition neural network model to identify surface defects and their severity.

[0089] Calculate the surface defect coefficient S f , S f The calculation formula is as follows:

[0090]

[0091] Among them, ω i is the weight of the i-th defect point; I i is the severity of the i-th defect point; m is the number of defect points.

[0092] Calculate the comprehensive evaluation coefficient S total , S total The calculation formula is as follows:

[0093] S total =αS p +βS x +γS f ;

[0094] Among them, α, β and γ are S p , S x and S f Preset scaling factor.

[0095] According to the preset defect threshold X, determine whether the parts are qualified.

[0096] If S total If it is less than or equal to X, the component is considered qualified;

[0097] If S total If it is greater than X, the component is deemed unqualified and a destruction mark is generated.

[0098] Step 4: Result output and troubleshooting.

[0099] Qualified Parts: Output the information of qualified parts.

[0100] Defective parts: Generate destruction marks and output information of defective parts.

[0101] For the parts that generate destruction marks, obtain their production process data, including the production link, production machine or production container number and production time.

[0102] Analyze the failure of production equipment or production containers, calculate the proportion of the number of occurrences of each production equipment or production container number, sort them from large to small according to the proportion, and save them in the set V s middle;

[0103] Count the number of parts V marked for destruction s and the total number of parts T s , according to the preset production failure threshold Q s and production quantity difference threshold Q d , determine whether troubleshooting or quality inspection is needed.

[0104] Troubleshooting: If Troubleshoot the production equipment or production containers in the order of their numbers in set Vs.

[0105] Quality inspection: If and For production equipment or production containers according to set V s Quality inspection is carried out in the order of numbers in the

[0106] Step 5: Analysis of unqualified parts.

[0107] Analyze the number of defective parts, calculate the anomaly ratio and standard deviation, and determine the abnormal type of defective parts.

[0108] Quantity statistics: Get the number of unqualified parts in time period t+1, where time period t+1 is the next time period after time t.

[0109] The number of unqualified parts counted is compared with the preset threshold of unqualified parts. If the number of unqualified parts is less than the threshold of unqualified parts, it is marked as a normal period; if the number of unqualified parts is greater than or equal to the threshold of unqualified parts, it is marked as an abnormal period.

[0110] The ratio of the number of marks in the abnormal period to the number of unqualified parts in the t+1 period is marked as the abnormal ratio; the standard deviation of the data set of all the unqualified parts in the t+1 period is calculated.

[0111] If the anomaly ratio is less than or equal to the anomaly threshold, and the standard deviation is less than or equal to the standard deviation threshold, it is marked as an occasional anomaly; otherwise, it is marked as a long-term anomaly and a fault warning instruction is generated.

[0112] The beneficial effects of this embodiment are: realizing refined management and error detection of the processing process, effectively improving the quality and efficiency of parts processing. Through stage time monitoring, image preprocessing, feature extraction and error detection, result output and troubleshooting and analysis of unqualified parts, the product qualification rate is ensured and the failure rate is reduced.

[0113] The weights of the present invention are used to measure the degree of influence of different factors or variables on a certain result or decision. The definition of weight refers to the value assigned to each factor when comparing and evaluating multiple factors to reflect its importance or priority. These weights can be determined according to specific circumstances and needs, and are usually jointly formulated and confirmed by professionals or relevant stakeholders. By setting weights reasonably, it can help programs or systems make decisions or predictions more accurately.

[0114] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

[0115] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A component processing error detection system based on machine vision, characterized in that: The system includes a stage management and time monitoring module, an image acquisition and preprocessing module, a feature extraction and error detection module, a result output and troubleshooting module, and a non-conforming parts analysis module; Stage management and time monitoring module: divides the parts processing process into multiple stages, marks these stages, analyzes time data, determines whether to generate preliminary stage switching instructions, and ensures that each stage is completed within a reasonable time by comparing with the preset time threshold; Image acquisition and preprocessing module: Use image capture equipment to capture multi-view images of parts, preprocess the acquired images, and collect production process data of each part, including production links, production machine or container numbers, and production time; Feature extraction and error detection module: extract the features of parts, detect dimensional errors, deformation errors and surface defect errors, calculate the comprehensive evaluation coefficient, and determine whether the parts are qualified according to the preset defect threshold; Result output and troubleshooting module: outputs the information of qualified parts and generates destruction marks and related information for unqualified parts; For unqualified parts, obtain their production process data and analyze the failure of production equipment or containers. According to the preset production failure threshold and production quantity difference threshold, determine whether troubleshooting or quality inspection is needed, and perform troubleshooting or quality inspection on production equipment or containers in the order of numbers; Defective parts analysis module: Analyze the number of defective parts, calculate the anomaly ratio and standard deviation, determine whether the abnormal type of the defective parts is sporadic or long-term, generate fault warning instructions, compare the statistical number of defective parts with the preset threshold, and mark the normal period or abnormal period.

2. A method for detecting component processing errors based on machine vision, used to implement a component processing error detection system based on machine vision as claimed in claim 1, characterized in that: The method flow is as follows: Step 1: Divide the parts processing process into multiple stages and mark them as different stages, including image acquisition stage, image preprocessing stage, feature extraction stage, error detection stage and result output stage; Step 2: Collect the time data of each stage, analyze the time data, and determine whether to generate a preliminary stage switching instruction; Step 3: If a preliminary stage switching instruction is generated, the image acquisition stage is entered, and a multi-view image of the component is captured using an image capture device and pre-processed; Step 4: Perform feature extraction and error detection, including dimensional error detection, deformation error detection and surface defect detection. By calculating the dimensional error coefficient, deformation error coefficient and surface defect coefficient, as well as the comprehensive evaluation coefficient, determine whether the parts are qualified; Step 5: Output the information of qualified or unqualified parts according to the test results. For unqualified parts, generate a destruction mark and obtain its production process data for analysis to determine whether troubleshooting or quality inspection is needed; Step 6: Analyze the number of unqualified parts, calculate the anomaly ratio and standard deviation, determine the abnormal type of unqualified parts, and generate corresponding fault warning instructions.

3. The method for detecting component processing errors based on machine vision according to claim 2, characterized in that: The image preprocessing in step 1 includes grayscale conversion and Gaussian filtering to reduce noise and improve image quality.

4. The method for detecting component processing errors based on machine vision according to claim 2, characterized in that: The determination of whether to generate a preliminary stage switching instruction in step 2 is specifically as follows: The number of statistical stages is n1, and n2 time thresholds are preset, n1=n2; Compare the collected time data with the corresponding time threshold; if the time data is greater than or equal to the time threshold, generate a preliminary stage switching instruction; Otherwise, no preliminary phase switching instruction is generated.

5. The method for detecting component processing errors based on machine vision according to claim 2, characterized in that: The multi-view images in step three include a top view, a side view and a bottom view.

6. The method for detecting component processing errors based on machine vision according to claim 2, characterized in that: The size error detection in step 4 is achieved by comparing the image of the key part of the component with the standard image; the deformation error detection is achieved by intercepting the key shape of the component from different views and comparing it with the standard shape; the surface defect detection is achieved by training a multi-target recognition neural network model to identify surface defects and their severity; the details are as follows: S301, dimensional error detection; Cut out the key parts of the parts from the image, including length, width, and diameter, compare the cut out image with the standard image, and calculate the size error coefficient S p , S p The calculation formula is as follows: Among them, A actual is the actual size; A standard It is a standard size; S302, deformation error detection; Capture the key shapes of the parts from different views, including the top shape of the parts in the top view and the bottom shape of the parts in the side view. Compare the captured shapes with the standard shapes and calculate the deformation error coefficient S. x , S x The calculation formula is as follows: Among them, P actual,i is the coordinate of the ith point of the actual shape; P standard,i is the coordinate of the i-th point of the standard shape, and n is the number of points; S303, surface defect detection; Collect images of parts with surface defects and parts without surface defects in advance and annotate them; train a multi-target recognition neural network model to identify surface defects and their severity; calculate the surface defect coefficient S f , S f The calculation formula is as follows: Among them, ω i is the weight of the i-th defect point; I i is the severity of the i-th defect point; m is the number of defect points; S304, calculate the comprehensive evaluation coefficient S total , S total The calculation formula is as follows: total =αS p +βS x +γS f ; where α, β and γ are S p , S x and S f Preset scaling factor; According to the preset defect threshold X, determine whether the component is qualified. If S total If S is less than or equal to X, the component is considered qualified; total If it is greater than X, the component is deemed unqualified and a destruction mark is generated.

7. The method for detecting component processing errors based on machine vision according to claim 2, characterized in that: The step five is as follows: Qualified parts: Output the information of qualified parts; Unqualified parts: Generate destruction marks and output information of unqualified parts; For the parts that generate destruction marks, obtain their production process data, including the production link, production machine or production container number and production time; Analyze the failure of production equipment or production containers, calculate the proportion of the number of occurrences of each production equipment or production container number, sort them from large to small according to the proportion, and save them in the set V s middle; Count the number of parts V marked for destruction s and the total number of parts T s , according to the preset production failure threshold Q s and production quantity difference threshold Q d , determine whether troubleshooting or quality inspection is needed; Troubleshooting: If Perform troubleshooting on production equipment or production containers in the order of numbers in set Vs; Quality inspection: If and For production equipment or production containers according to set V s Quality inspection is carried out in the order of numbers in the 8. The method for detecting component processing errors based on machine vision according to claim 2, characterized in that: The step six is ​​as follows: Analyze the number of defective parts, calculate the anomaly ratio and standard deviation, and determine the abnormal type of defective parts; Get the number of unqualified parts in period t+1, where period t+1 is the next period after time t; The counted number of unqualified parts is compared with the preset threshold of the number of unqualified parts. If the number of unqualified parts is less than the threshold of the number of unqualified parts, it is marked as a normal period. If the number of defective parts is greater than or equal to the threshold number of defective parts, it is marked as an abnormal period; The ratio of the number of marks in the abnormal period to the number of unqualified parts in the t+1 period is marked as the abnormal ratio; the standard deviation of the data set of all unqualified parts in the t+1 period is calculated; If the anomaly ratio is less than or equal to the anomaly threshold, and the standard deviation is less than or equal to the standard deviation threshold, it is marked as an occasional anomaly; otherwise, it is marked as a long-term anomaly and a fault warning instruction is generated.

Images