CN117314829A

CN117314829A - Industrial part quality inspection method and system based on computer vision

Info

Publication number: CN117314829A
Application number: CN202311110344.3A
Authority: CN
Inventors: 余丹; 兰雨晴; 李其蓁; 邢智涣
Original assignee: China Standard Intelligent Security Technology Co Ltd
Current assignee: China Standard Intelligent Security Technology Co Ltd
Priority date: 2023-08-30
Filing date: 2023-08-30
Publication date: 2023-12-29

Abstract

The invention provides an industrial part quality inspection method and system based on computer vision. The industrial part quality inspection method based on computer vision comprises the following steps: acquiring image data of an industrial part by using a camera, and preprocessing the image data to obtain preprocessed image data; extracting features of the preprocessed image data to obtain feature information of the industrial part; analyzing the characteristic information by using a convolutional neural network model to obtain a defect detection result of the industrial part; and generating an industrial part quality inspection report according to the defect detection result. The system comprises modules corresponding to the method steps.

Description

Industrial part quality inspection method and system based on computer vision

Technical Field

The invention provides an industrial part quality inspection method and system based on computer vision, and belongs to the technical field of industrial part quality inspection.

Background

Conventional detection methods relying on manual visual inspection, manual measurement, ultrasonic detection, and the like, while playing a key role in many fields, are accompanied by a series of challenges and limitations. The problems of human error, subjectivity and low efficiency exist, and the ultrasonic detection and optical detection need professional personnel to operate and interpret results; specific defects are as follows:

human error and subjectivity: the manual visual inspection and measurement process is susceptible to subjective factors and individual differences of the operator. Different operators may have different observations and evaluations of the same situation, which may lead to inconsistent results. In addition, fatigue, distraction, or undertraining may also lead to the occurrence of human error.

The efficiency is low: manual inspection and measurement typically requires a significant amount of time and effort. For mass production or in the case where high frequency detection is required, this may not only increase the cost but also reduce the production efficiency. Furthermore, long repetitive tasks may lead to operator fatigue, thereby affecting accuracy.

Professional knowledge requirements: advanced techniques such as ultrasonic testing and optical testing typically require a specially trained operator to perform. This increases the cost of recruitment and training and may limit the widespread use of these techniques because of the need for specific expertise and skills.

Not adapted to the specific environment: some environmental conditions may pose challenges for manual detection, such as high temperature, low temperature, high radiation, or hazardous gas environments. In these cases, personnel safety and health may be compromised and accurate measurements may be difficult to obtain.

Disclosure of Invention

The invention provides a quality inspection method and a quality inspection system for industrial parts based on computer vision, which are used for solving the technical problems in the prior art, and the adopted technical scheme is as follows:

a computer vision-based industrial part quality inspection method, the computer vision-based industrial part quality inspection method comprising:

acquiring image data of an industrial part by using a camera, and preprocessing the image data to obtain preprocessed image data;

extracting features of the preprocessed image data to obtain feature information of the industrial part;

analyzing the characteristic information by using a convolutional neural network model to obtain a defect detection result of the industrial part;

and generating an industrial part quality inspection report according to the defect detection result.

Further, collecting image data of the industrial part by using a camera, preprocessing the image data to obtain preprocessed image data, including:

the method comprises the steps of utilizing a camera to acquire images of industrial parts appearing in a shooting visual field, and obtaining image data corresponding to the industrial parts;

preprocessing the image data corresponding to the industrial part to obtain preprocessed image data, wherein the preprocessing comprises noise reduction processing, contrast enhancement processing and brightness adjustment processing.

Further, feature extraction is performed on the preprocessed image data to obtain feature information of the industrial part, including:

extracting image data of each industrial part;

performing visual identification processing on the image data to obtain characteristic information of the industrial part contained in the image data;

wherein the characteristic information of the industrial part includes shape, color and texture.

Further, analyzing the characteristic information by using a convolutional neural network model to obtain a defect detection result of the industrial part, including:

inputting the characteristic information into a convolutional neural network model;

after the neural network model receives the characteristic information, carrying out characteristic identification on the characteristic information to judge whether the industrial part has defects or not;

and when the industrial part has defects, extracting defect distribution conditions of the industrial part, and marking the defects on image data corresponding to the industrial part.

Further, generating an industrial part quality inspection report according to the defect detection result, including:

after the defect detection of each batch of industrial parts is finished, extracting image data with defect marks in the image data corresponding to each batch of industrial parts;

counting the image data with the defect marks to generate an industrial part quality inspection report corresponding to each batch of industrial parts; wherein, the quality inspection report comprises the qualification and defect distribution conditions of the industrial parts.

A computer vision-based industrial part quality inspection system, the computer vision-based industrial part quality inspection system comprising:

the acquisition and processing module is used for acquiring image data of the industrial part by using the camera and preprocessing the image data to obtain preprocessed image data;

the characteristic information acquisition module is used for carrying out characteristic extraction on the preprocessed image data to obtain characteristic information of the industrial part;

the defect detection result acquisition module is used for analyzing the characteristic information by using a convolutional neural network model to obtain a defect detection result of the industrial part;

and the report generation module is used for generating an industrial part quality inspection report according to the defect detection result.

Further, the acquisition and processing module includes:

the image acquisition module is used for acquiring images of the industrial parts appearing in the shooting visual field by using the camera to obtain image data corresponding to the industrial parts;

the preprocessing module is used for preprocessing the image data corresponding to the industrial part to obtain preprocessed image data, wherein the preprocessing comprises noise reduction processing, contrast enhancement processing and brightness adjustment processing.

Further, the feature information acquisition module includes:

the image data extraction module is used for extracting the image data of each industrial part;

the visual identification processing module is used for performing visual identification processing on the image data to acquire characteristic information of the industrial part contained in the image data;

Further, the defect detection result obtaining module includes:

the information input module is used for inputting the characteristic information into the convolutional neural network model;

the defect judging module is used for carrying out characteristic identification on the characteristic information after the neural network model receives the characteristic information to judge whether the industrial part has defects or not;

and the defect marking module is used for extracting the defect distribution condition of the industrial part when the industrial part has defects, and marking the defects on the image data corresponding to the industrial part.

Further, the report generating module includes:

the mark extraction module is used for extracting image data with defect marks in the image data corresponding to each batch of industrial parts after the defect detection of each batch of industrial parts is finished;

the statistics report module is used for carrying out statistics on the image data with the defect marks to generate an industrial part quality inspection report corresponding to each batch of industrial parts; wherein, the quality inspection report comprises the qualification and defect distribution conditions of the industrial parts.

The invention has the beneficial effects that:

the invention provides a quality inspection method and system for industrial parts based on computer vision, which are characterized in that an industrial part image is acquired through a camera or external equipment, the defect detection and classification are carried out on the part image, the defect is marked, and the quality inspection result is fed back in real time; after finishing quality inspection of each batch of parts, generating a quality inspection report which comprises information such as qualification rate of the parts, defect distribution and the like, helping to know occurrence frequency of various defects and improving production process. Accuracy and consistency are improved, efficiency is improved, and real-time monitoring and quality control are realized.

Drawings

FIG. 1 is a flow chart of the method of the present invention;

fig. 2 is a system block diagram of the system of the present invention.

Detailed Description

The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.

The embodiment of the invention provides an industrial part quality inspection method based on computer vision, which is shown in fig. 1 and comprises the following steps:

s1, acquiring image data of an industrial part by using a camera, and preprocessing the image data to obtain preprocessed image data;

s2, extracting features of the preprocessed image data to obtain feature information of the industrial part;

s3, analyzing the characteristic information by using a convolutional neural network model to obtain a defect detection result of the industrial part;

s4, generating an industrial part quality inspection report according to the defect detection result.

The working principle of the technical scheme is as follows: image acquisition and preprocessing (S1): image data of the industrial parts are acquired by a camera and then preprocessed. The preprocessing process comprises operations such as denoising, enhancement, image segmentation and the like, and aims to improve the accuracy of subsequent feature extraction and analysis.

Feature extraction (S2): and extracting characteristic information of the industrial part from the preprocessed image data. These features may include shape, texture, color, etc., which are critical to identifying defects in industrial parts.

Convolutional neural network analysis (S3): the extracted characteristic information is input into a Convolutional Neural Network (CNN) model for analysis. CNN is a deep learning model that can effectively learn and identify complex patterns in images. The model analyzes the characteristic information to detect whether the industrial part has a defect. This process typically requires a large amount of labeled training data to train the CNN model.

Generating a quality inspection report (S4): and generating a quality inspection report of the industrial part according to the output result of the CNN model. Reports typically include information about the status of industrial parts (normal or defective), the location and type of defects, and the like. These reports may be used in the fields of decision making, quality management and maintenance.

The technical scheme has the effects that: automation and efficiency: the method can realize automatic quality inspection of industrial parts, reduces manual operation and improves efficiency. It can process a large number of part images in a short time.

High precision and reliability: the convolutional neural network model has strong image recognition capability, can detect defects in industrial parts with high precision, and reduces false alarm and missing detection.

Real-time performance: due to the rapid processing capability of the computer vision and deep learning model, the method can perform quality inspection in real time and is suitable for application on continuous production lines.

Data recording and traceability: the generated quality inspection report can be used for recording the quality state of each industrial part, so that quality tracing and statistical analysis are realized, and the quality of products is improved.

In a word, the technical scheme of the embodiment realizes high-efficiency, accurate and automatic quality inspection of industrial parts by combining the image processing and the deep learning technology, and improves the level of quality control.

In one embodiment of the present invention, capturing image data of an industrial part using a camera and preprocessing the image data to obtain preprocessed image data, comprising:

s101, acquiring an image of an industrial part appearing in a shooting visual field by using a camera to obtain image data corresponding to the industrial part;

s102, preprocessing the image data corresponding to the industrial part to obtain preprocessed image data, wherein the preprocessing comprises noise reduction processing, contrast enhancement processing and brightness adjustment processing.

The working principle of the technical scheme is as follows: image acquisition (S101): first, an image of an industrial part present in its field of view is acquired using a camera. This means that the camera takes a picture or image of the industrial part, converting it into image data in digital form. These image data will be used for subsequent quality control analysis.

Image preprocessing (S102): the acquired image data often contains various noise and imperfections that may affect the accuracy of the quality inspection. Therefore, the image needs to be preprocessed before further analysis. The step of pre-treatment generally comprises:

and (3) noise reduction treatment: noise in the image is removed to reduce false positive or false negative results due to noise.

Contrast enhancement treatment: the contrast of the images is adjusted to highlight features of the industrial parts making them easier to analyze and identify.

Brightness adjustment processing: the brightness of the image is adjusted to ensure that the brightness level of the image is suitable for subsequent analysis requirements.

Through these preprocessing steps, improved, clean, well-contrasted image data is obtained, which facilitates subsequent feature extraction and quality inspection analysis.

The technical scheme has the effects that: improving the image quality: the preprocessing step helps to improve the quality of the image, removes noise and unnecessary interference, and makes the industrial parts more clearly visible.

Enhancement features: enhanced contrast and brightness adjustment can highlight features of industrial parts making them easier to identify and analyze by computer vision algorithms.

And (3) misjudgment is reduced: the noise reduction processing is helpful for reducing misjudgment caused by image noise, and improves the accuracy of quality inspection.

The adaptability: through the preprocessing step, the system can adapt to different illumination conditions and image quality, so that quality inspection is performed in various environments.

In a word, the image acquisition and preprocessing steps in the technical scheme of the embodiment are beneficial to improving the accuracy and reliability of quality inspection of industrial parts, so that the level of product quality control is improved.

In one embodiment of the present invention, feature extraction is performed on the preprocessed image data to obtain feature information of an industrial part, including:

s201, extracting image data of each industrial part;

s202, performing visual identification processing on the image data to acquire characteristic information of industrial parts contained in the image data;

The working principle of the technical scheme is as follows: extracting image data (S201): in this step, the image data of each industrial part is extracted from the preprocessed image data. This can be achieved by calibration or segmentation techniques to ensure that the image of each industrial part can be extracted independently for subsequent feature analysis.

Visual recognition processing (S202): next, visual recognition processing is performed on the extracted image data. This may include analyzing feature information of the industrial part contained in the image data using computer vision algorithms and techniques, such as Convolutional Neural Networks (CNNs), image segmentation, feature extraction, and the like.

Extracting characteristic information: during the visual recognition process, the system analyzes the image data and extracts characteristic information of the industrial part. Such characteristic information may include, but is not limited to:

shape: profile features of industrial parts such as contours, boundaries, etc.

Color: color information of industrial parts can be used to distinguish between different parts or anomalies.

Texture: texture features of industrial part surfaces, such as smoothness, roughness, etc.

The technical scheme has the effects that: automated feature extraction: through visual recognition processing, the system can automatically extract characteristic information from image data, so that the requirement of manual intervention is reduced, and the quality inspection efficiency is improved.

Multidimensional information: the extracted feature information includes aspects of shape, color, and texture, thereby providing a more comprehensive and accurate description of the features of the industrial part.

Accuracy: the feature extraction based on the computer vision algorithm can keep consistent accuracy in different scenes, and errors possibly introduced by artificial subjective judgment are avoided.

The adaptability: the method can be suitable for different types of industrial parts, and algorithm parameters and models only need to be adjusted according to actual needs.

In summary, the above technical solution of the present embodiment is helpful for converting image data into characteristic information about industrial parts, and provides a basis for subsequent defect detection and quality inspection result generation.

In one embodiment of the present invention, analyzing the characteristic information with a convolutional neural network model to obtain a defect detection result of an industrial part includes:

s301, inputting the characteristic information into a convolutional neural network model;

s302, after the neural network model receives the characteristic information, carrying out characteristic identification on the characteristic information to judge whether the industrial part has defects or not;

and S303, when the industrial part has defects, extracting defect distribution conditions of the industrial part, and marking the defects on image data corresponding to the industrial part.

The working principle of the technical scheme is as follows: inputting feature information (S301): first, the feature information extracted in the previous step is input into the convolutional neural network model. Such characteristic information includes characteristics of the shape, color, texture, etc. of the industrial part.

Feature recognition judgment (S302): after the convolutional neural network model receives the characteristic information, characteristic recognition and defect detection are started. The CNN model learns and extracts abstract representation of characteristic information through multi-layer convolution and pooling operation, and then classifies or regresses through a full connection layer to judge whether industrial parts have defects. The determination result may be binary (defect/no defect) or multi-category (different types of defects).

Defect distribution extraction and marking (S303): if the CNN model judges that the industrial part has defects, the defect distribution situation is extracted. This may include determining information of the location, size, shape, etc. of the defect. Then, defect marking is performed on the image data corresponding to the industrial part, typically by drawing a bounding box or marking defective pixels.

The technical scheme has the effects that: automated defect detection: by using the convolutional neural network model, the system can automatically detect defects, reduce the dependence on manual operation and improve the efficiency and consistency.

High accuracy: the CNN model can learn complex characteristic representation under a large amount of training data, so that highly accurate defect detection can be realized, and the probability of missing detection and false detection is reduced.

Real-time performance: the convolutional neural network can perform image processing and defect detection at real-time or nearly real-time speed, and is suitable for production environments requiring rapid feedback.

Marking defects: when a defect is detected, the system may automatically mark the defect, helping the operator to more easily identify and address the problem.

In a word, the technical scheme of the embodiment realizes the defect detection of the industrial parts through the deep learning technology of the convolutional neural network, and improves the efficiency and accuracy of quality inspection.

According to one embodiment of the invention, the quality inspection report of the industrial part is generated according to the defect detection result, and the quality inspection report comprises the following components:

s401, after the defect detection of each batch of industrial parts is finished, extracting image data with defect marks in image data corresponding to each batch of industrial parts;

s402, counting the image data with the defect marks to generate an industrial part quality inspection report corresponding to each batch of industrial parts; wherein, the quality inspection report comprises the qualification and defect distribution conditions of the industrial parts.

The working principle of the technical scheme is as follows: extracting image data with defect marks (S401): after each batch of industrial parts is finished with defect detection, the system extracts image data with defect marks from the image data corresponding to the batch of industrial parts. These image data contain the identified defects and their location information.

Statistical defect information (S402): next, the system performs a statistical analysis on the image data with the defect markers. This includes counting the total number, the number of pass, and the number of fail for each batch of industrial parts. Meanwhile, the distribution situation of the defects, including the information of the types, the number, the positions and the like of the defects, is analyzed.

Generating a quality inspection report: based on the results of the statistical analysis, the system generates a quality inspection report of the industrial part. This report typically includes the following:

qualified number: the report will indicate how many of the industrial parts in the batch are acceptable, i.e., the number of industrial parts for which defects have not been identified.

Number of rejects: the report may indicate how many industrial parts were identified as failed, i.e., the number of defective industrial parts.

Defect distribution conditions: the report will detail the defect distribution of each failed industrial part, including information on the type, location, size, etc. of the defect.

Summary and advice: based on the analysis results, the report may provide summary and improvement suggestions to improve the quality of production.

The technical scheme has the effects that: automatic report generation: according to the technical scheme, the quality inspection report is generated through an automatic process, so that the time and the workload for compiling the manual report are reduced.

Accuracy: reports are generated based on actual defect detection results, so that the reports have high accuracy, and production management staff can be helped to accurately know the quality condition of each batch of industrial parts.

Real-time performance: reports can be generated immediately after each batch of industrial parts is inspected, providing timely feedback, and facilitating early discovery and resolution of quality problems.

Improved production: by analyzing the defect distribution situation and providing improvement suggestions, the technical scheme is beneficial to the improvement of the production process, and the overall quality level of the industrial parts is improved.

In a word, the technical scheme of the embodiment helps production management staff to know the quality condition of industrial parts by automatically generating quality inspection reports, and provides targeted improvement suggestions, so that the efficiency and quality of a production flow are improved.

The embodiment of the invention provides an industrial part quality inspection system based on computer vision, as shown in fig. 2, which comprises:

The working principle of the technical scheme is as follows: image acquisition and preprocessing: image data of the industrial parts are acquired by a camera and then preprocessed. The preprocessing process comprises operations such as denoising, enhancement, image segmentation and the like, and aims to improve the accuracy of subsequent feature extraction and analysis.

Feature extraction: and extracting characteristic information of the industrial part from the preprocessed image data. These features may include shape, texture, color, etc., which are critical to identifying defects in industrial parts.

Convolutional neural network analysis: the extracted characteristic information is input into a Convolutional Neural Network (CNN) model for analysis. CNN is a deep learning model that can effectively learn and identify complex patterns in images. The model analyzes the characteristic information to detect whether the industrial part has a defect. This process typically requires a large amount of labeled training data to train the CNN model.

Generating a quality inspection report: and generating a quality inspection report of the industrial part according to the output result of the CNN model. Reports typically include information about the status of industrial parts (normal or defective), the location and type of defects, and the like. These reports may be used in the fields of decision making, quality management and maintenance.

In one embodiment of the present invention, the acquisition and processing module includes:

The working principle of the technical scheme is as follows: and (3) image acquisition: first, an image of an industrial part present in its field of view is acquired using a camera. This means that the camera takes a picture or image of the industrial part, converting it into image data in digital form. These image data will be used for subsequent quality control analysis.

Image preprocessing: the acquired image data often contains various noise and imperfections that may affect the accuracy of the quality inspection. Therefore, the image needs to be preprocessed before further analysis. The step of pre-treatment generally comprises:

In one embodiment of the present invention, the feature information obtaining module includes:

The working principle of the technical scheme is as follows: extracting image data: in this step, the image data of each industrial part is extracted from the preprocessed image data. This can be achieved by calibration or segmentation techniques to ensure that the image of each industrial part can be extracted independently for subsequent feature analysis.

Visual recognition processing: next, visual recognition processing is performed on the extracted image data. This may include analyzing feature information of the industrial part contained in the image data using computer vision algorithms and techniques, such as Convolutional Neural Networks (CNNs), image segmentation, feature extraction, and the like.

shape: profile features of industrial parts such as contours, boundaries, etc.

In one embodiment of the present invention, the defect detection result obtaining module includes:

The working principle of the technical scheme is as follows: inputting characteristic information: first, the feature information extracted in the previous step is input into the convolutional neural network model. Such characteristic information includes characteristics of the shape, color, texture, etc. of the industrial part.

And (3) feature identification judgment: after the convolutional neural network model receives the characteristic information, characteristic recognition and defect detection are started. The CNN model learns and extracts abstract representation of characteristic information through multi-layer convolution and pooling operation, and then classifies or regresses through a full connection layer to judge whether industrial parts have defects. The determination result may be binary (defect/no defect) or multi-category (different types of defects).

Defect distribution extraction and marking: if the CNN model judges that the industrial part has defects, the defect distribution situation is extracted. This may include determining information of the location, size, shape, etc. of the defect. Then, defect marking is performed on the image data corresponding to the industrial part, typically by drawing a bounding box or marking defective pixels.

In one embodiment of the invention, the report generating module includes:

The working principle of the technical scheme is as follows: extracting image data with defect marks: after each batch of industrial parts is finished with defect detection, the system extracts image data with defect marks from the image data corresponding to the batch of industrial parts. These image data contain the identified defects and their location information.

Counting defect information: next, the system performs a statistical analysis on the image data with the defect markers. This includes counting the total number, the number of pass, and the number of fail for each batch of industrial parts. Meanwhile, the distribution situation of the defects, including the information of the types, the number, the positions and the like of the defects, is analyzed.

It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims

1. The industrial part quality inspection method based on computer vision is characterized by comprising the following steps of:

2. The reliability multicast method based on receiver confirmation maximum reliability frame according to claim 1, wherein capturing image data of an industrial part by a camera and preprocessing the image data to obtain preprocessed image data, comprising:

3. The reliability multicast method based on the receiver confirmation maximum reliability frame according to claim 1, wherein the feature extraction is performed on the preprocessed image data to obtain feature information of an industrial part, comprising:

extracting image data of each industrial part;

4. The reliability multicast method based on the receiver confirmation maximum reliability frame according to claim 1, wherein analyzing the characteristic information with a convolutional neural network model to obtain a defect detection result of an industrial part comprises:

5. The reliability multicast method based on receiver confirmation maximum reliability frame according to claim 1, wherein generating an industrial part quality inspection report according to a defect detection result comprises:

6. A computer vision-based industrial part quality inspection system, the computer vision-based industrial part quality inspection system comprising:

7. The receiver acknowledgement maximum reliability multicast system according to claim 6, wherein the acquisition and processing module comprises:

8. The receiver acknowledgement maximum reliable frame based reliability multicast system according to claim 6, wherein the characteristic information obtaining module comprises:

9. The reliability multicast system based on receiver-acknowledged maximum reliability frames according to claim 6, wherein said defect detection result acquisition module comprises:

10. The reliability multicast system based on receiver acknowledgement maximum reliability frames according to claim 6, wherein said report generating module comprises: