Multi-azimuth machine vision detection system and method for automobile parts

Abstract

The invention relates to a multidirectional machine vision detection system and method for automobile parts, and the system comprises an image sensor module, a self-adaptive light source control module, a mechanical arm system, and a data transmission and analysis module. The self-adaptive light source control module adjusts the intensity and angle of a light source in real time according to the detection environment and the gloss characteristic of the surface of the accessory, interference of reflection and shadow is reduced, the mechanical arm system drives the accessory to move at multiple angles through an inverse kinematics algorithm, and it is ensured that the surface of the accessory is fully covered; and the data transmission and analysis module is used for processing the collected image data, identifying potential defects of the accessories through a deep learning algorithm and a high-dimensional feature analysis technology, and generating a detection report. The method is suitable for quality detection of automobile parts with complex shapes and high glossiness.

Description

Technical Field

[0001] The invention relates to a multi-directional machine vision detection system and method for automobile parts. Background Art

[0002] With the rapid development of the automobile industry, the quality management of auto parts has become increasingly important. Traditional manual inspection methods face problems such as low efficiency, poor accuracy and high labor intensity. These problems are particularly prominent in large-scale production environments. Manual inspection is not only easily affected by human factors, but also difficult to ensure the consistency and reliability of inspection when facing auto parts with complex geometric shapes and diverse materials.

[0003] In recent years, machine vision technology has gradually become an important solution in the field of industrial inspection due to its high precision and high efficiency. Machine vision systems capture object images through high-resolution cameras and use image processing and analysis algorithms to automatically inspect objects. They can quickly identify defects, measure dimensions, and evaluate quality. However, existing machine vision inspection systems still have some limitations in practical applications:

[0004] Many existing systems can only inspect from a fixed perspective, which makes it easy to overlook certain hidden parts of accessories, resulting in potential defects going undetected. For accessories with high surface gloss or complex materials, traditional image processing algorithms are often affected by light reflection and shadows, resulting in inaccurate inspection results.

[0005] Therefore, technical personnel in this field need a multi-directional machine vision inspection system for automotive parts that integrates multi-directional visual inspection, adaptive light source control, and intelligent data analysis functions to achieve more efficient and reliable quality management. Summary of the invention

[0006] The purpose of the present invention is to solve the above deficiencies in the prior art and provide a multi-directional machine vision inspection system and method for automobile parts, aiming to solve the limitations of automobile parts inspection in the prior art, especially the problem of missed detection of potential defects caused by fixed viewing angle, light interference and low inspection efficiency. The following is a specific solution:

[0007] In one aspect, a multi-directional machine vision inspection system for automobile parts includes an image sensor module, an adaptive light source control module, a mechanical arm system and a data transmission and analysis module, each image sensor module is configured to capture multiple viewing angle images of automobile parts; the adaptive light source control module is used to adjust the light source intensity and direction in the inspection environment in real time to reduce the interference of the reflection of the accessory surface on the image capture, and the adaptive light source control module calculates the gloss characteristics of the accessory surface and adjusts the angle and brightness of the light source according to the data captured by the image sensor; the mechanical arm system is used to drive the automobile parts to be inspected to move or rotate at multiple angles to ensure that the sensor can capture the entire surface of the accessory from all angles, and the mechanical arm system can automatically adjust its motion path according to the shape, size and specific requirements of the accessory and the inspection task. By using the inverse kinematics algorithm, the movement of the mechanical arm can be precisely controlled, and the motion path can be dynamically adjusted according to the sensor feedback; the data transmission and analysis module is used to receive and process the image data obtained from each image sensor in real time, and analyze the inspection results in combination with the defect recognition algorithm, and the data transmission module can collect the data captured by the multi-directional sensor to the central processing unit, integrate the data, and generate an inspection report.

[0008] During the operation of the system, multiple image sensor modules work together to capture automotive parts from different angles. The images captured by each sensor are optimized through the image preprocessing module to ensure that the images have consistent brightness and clarity under a variety of lighting conditions. Next, the deep learning analysis module analyzes the processed images to automatically identify and mark potential defects in the images. The module makes a comprehensive judgment by combining images obtained by different sensors to avoid misjudgments due to angle limitations of a single sensor or undesirable lighting conditions.

[0009] Based on the data fed back by the sensor, the adaptive light source control module adjusts the direction and brightness of the light source in real time, so that the detection light can be evenly distributed on the surface of the accessory, reducing the impact of high light reflection or shadow on the detection results. For example, when it is detected that the glossiness of the metal accessory surface is too high, the system will automatically dim the light source and adjust the incident angle of the light to reduce the impact of surface reflection. Through the collaborative work of the multi-axis robotic arm, the system can rotate the accessory at multiple angles to ensure that the detection system captures the complete accessory image from all angles, especially on the recessed or complex geometric structure of the accessory, to ensure that no surface area is missed.

[0010] Finally, the data from all sensors are transmitted in real time through the data transmission and analysis module and analyzed uniformly in the central processing unit. This module is capable of processing large amounts of image data, performing real-time defect judgment and classification, and generating complete inspection reports.

[0011] Furthermore, the image sensor module includes a plurality of high-precision imaging units, and the imaging units are configured to capture images using different viewing angles and focal lengths, and the imaging model thereof can be expressed as: i =f(X, Y, Z, θ i ,φ i ), where Z i Imaging unit s i The captured image data, (X, Y, Z) is the three-dimensional space coordinate, θ i and φ i are the corresponding pitch angles and azimuth angles; it represents how each imaging unit in the image sensor module captures multi-view images by defining multiple angles, such as θ i and φ i The system can capture images of the same accessory from different spatial positions. This multi-perspective design significantly improves the shortcomings of existing systems that only rely on fixed perspectives, ensuring that hidden parts of the accessories can also be fully exposed, reducing the risk of missed detection.

[0012] Furthermore, the adaptive light source control module includes a light source adjustment algorithm module, which calculates the reflected light intensity based on the real-time captured image data, and the light source compensation algorithm is based on the following formula: L′=L target -k·R(θ, φ), where L′ is the adjusted light source intensity, L target is the target light intensity, k is the adjustment coefficient, and R(θ, φ) represents the function of the current reflected light intensity; it represents the characteristic of real-time adjustment of the light source intensity. When facing high-gloss surfaces, traditional systems are easily disturbed by reflections, which affects the image quality. By calculating the current reflected light intensity R(θ, φ), the adaptive light source control module can dynamically adjust the intensity L′ of the light source, thereby reducing the interference of high-light reflections on the image. This improvement ensures the quality and clarity of the image even under complex lighting conditions.

[0013] Furthermore, the motion path of the manipulator system is controlled by an inverse kinematics model, wherein the inverse kinematics model is based on the joint angle vector q=[q1, q2, ..., q m ] T , and solve the accessory position P by the following equation:

[0014]

[0015] , where l iis the length of the i-th joint of the robot arm, and h is the height from the base of the robot arm to the target plane. The motion path of the robot arm is accurately controlled by the inverse kinematics model, so that the robot arm can automatically adjust its motion according to the shape and size of the accessory, enabling it to be detected from all angles. This flexibility allows the system to perform comprehensive multi-angle detection of accessories, avoiding hidden defects that cannot be detected in traditional methods due to fixed viewing angles.

[0016] Furthermore, the robotic arm system is equipped with a motion compensation mechanism based on a dynamic model, which performs path adjustment through a real-time motion trajectory optimization algorithm and the following optimization objective function: Among them, J is the optimization target, d i Represents the error of the i-th position point, Δt i is the time to reach the i-th point, and λ is the time weight factor; it represents the motion compensation mechanism of the robot arm, which aims to optimize the motion path of the robot arm. By introducing the dynamic model, the system can adjust the motion trajectory in real time to adapt to the actual detection environment. This improvement enables the robot arm to respond quickly to changes in the detection task and ensure the full exposure of accessories at multiple angles, especially to maintain the accuracy of detection in complex scenes.

[0017] Furthermore, the data transmission and analysis module adopts a defect recognition algorithm based on high-dimensional features, and through cluster analysis of multi-dimensional image data, the following formula represents the construction of the feature matrix F:

[0018]

[0019] , where f j (X j , Y j ) is a feature extraction function for each image, combined with a clustering algorithm to identify defects on the surface of the accessory and form a final inspection report; unlike traditional algorithms, this system can simultaneously process image data from multiple perspectives to form a comprehensive feature matrix, thereby improving the accuracy of detection. This multi-dimensional feature analysis method overcomes the shortcomings of existing systems in dealing with high gloss and complex surfaces, allowing even smaller defects to be identified in a timely manner.

[0020] Furthermore, the data transmission and analysis module also includes a deep learning-based anomaly detection network, which is trained by the following loss function, and the following is an optimized detection model:

[0021] Among them, L is the total loss, N is the number of samples, and y i is the actual value, is the predicted value, λ is the regularization parameter, θ jThe model parameters are used to ensure the accuracy and robustness of the model in complex detection tasks. Through this model, the training of the deep learning model is optimized and the prediction error is reduced. Combined with the functions of multi-view and light source adjustment, the model can effectively learn on high-gloss surfaces and adapt to various complex detection conditions.

[0022] On the other hand, a multi-directional machine vision inspection method for automobile parts includes the following steps: a) using multiple image sensor modules to simultaneously capture images of automobile parts to be inspected from different perspectives to ensure that all visible surfaces of the parts are covered; b) real-time monitoring of image data, evaluating the impact of ambient lighting on image quality, automatically adjusting the intensity and angle of the light source, and eliminating the interference of reflection and shadow on image capture; c) driving the robotic arm to move the parts at multiple angles so that the image sensor can shoot from all directions; d) transmitting all captured image data to a central processing unit for image preprocessing and feature extraction to generate feature vectors; e) analyzing the processed image data to identify potential defects and generate a defect candidate set; f) further analyzing the identified defects, verifying the results in combination with a deep learning model, and determining the specific location and severity of the defects;

[0023] Furthermore, step f) further includes step g) automatically generating a detailed inspection report based on the inspection and confirmation results, including the defect type, location and impact degree, and feeding back the results to the relevant production control system.

[0024] Furthermore, step g) further includes h) collecting detection data, analyzing detection performance and results, and optimizing detection algorithms and system settings based on historical data.

[0025] Beneficial effects: Through multiple image sensors and precise robotic arm control, the present invention can conduct comprehensive inspections of automotive parts from multiple directions, greatly reducing the problem of missed detection of hidden defects due to viewing angle limitations; the adaptive light source control module effectively solves the problem of surface reflection and shadow interference on high-gloss accessories, ensuring image clarity and accuracy, thereby improving the accuracy of defect detection.

[0026] By controlling the motion path of the robotic arm through the inverse kinematics algorithm, the present invention realizes the intelligent movement of the robotic arm, can flexibly respond to the inspection requirements of different accessories, and improves the inspection efficiency; through the combination of high-dimensional feature extraction and deep learning models, the present invention can accurately identify the defects of automotive accessories in complex inspection tasks, ensure the reliability and stability of the inspection results, and generate detailed inspection reports to support quality control in the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a structural diagram of a multi-directional machine vision inspection system for automotive parts;

[0028] Figure 2 It is a schematic diagram of a multi-directional machine vision inspection method for automotive parts. DETAILED DESCRIPTION

[0029] In order to deepen the understanding of the present invention, the present invention will be further described in detail below in conjunction with embodiments and drawings. The embodiments are only used to explain the present invention and do not constitute a limitation on the protection scope of the present invention.

[0030] Please see Figure 1-2 A multi-directional machine vision inspection system for automobile parts includes an image sensor module, an adaptive light source control module, a mechanical arm system and a data transmission and analysis module, each image sensor module is configured to capture multiple viewing angle images of automobile parts; the adaptive light source control module is used to adjust the light source intensity and direction in the inspection environment in real time to reduce the interference of the reflection of the accessory surface on the image capture, and the adaptive light source control module calculates the gloss characteristics of the accessory surface and adjusts the angle and brightness of the light source according to the data captured by the image sensor; the mechanical arm system is used to drive the automobile parts to be inspected to move or rotate at multiple angles to ensure that the sensor can capture the entire surface of the accessory from all angles, and the mechanical arm system can automatically adjust its motion path according to the shape, size and specific requirements of the accessory and the inspection task. By using the inverse kinematics algorithm, the movement of the mechanical arm can be precisely controlled, and the motion path can be dynamically adjusted according to the sensor feedback; the data transmission and analysis module is used to receive and process the image data obtained from each image sensor in real time, and analyze the inspection results in combination with the defect recognition algorithm, and the data transmission module can collect the data captured by the multi-directional sensor to the central processing unit, integrate the data, and generate an inspection report.

[0031] The image sensor module contains multiple high-precision imaging units, which are configured to capture images with different viewing angles and focal lengths. Its imaging model can be expressed as: i =f(X, Y, Z, θ i ,φ i ), where Z i The imaging unit S i The captured image data, (X, Y, Z) is the three-dimensional space coordinate, θ i and φ i are the corresponding elevation and azimuth angles.

[0032] The adaptive light source control module includes a light source adjustment algorithm module, which calculates the reflected light intensity based on the real-time captured image data. The light source compensation algorithm is based on the following formula: L′=Lt argeet -k·R(θ, φ), where L′ is the adjusted light source intensity, L target为The target light intensity, k is the adjustment coefficient, and R(θ, φ) represents the function of the current reflected light intensity.

[0033] The motion path of the robot system is controlled by the inverse kinematics model, which uses the joint angle vector q = [q1, q2, ..., q m ] T , and solve the accessory position P by the following equation:

[0034]

[0035] , where l i is the length of the i-th joint of the robot, and h is the height from the base of the robot to the target plane.

[0036] The robotic arm system is equipped with a motion compensation mechanism based on a dynamic model. The mechanism uses a real-time motion trajectory optimization algorithm to adjust the path through the following optimization objective function:

[0037] Among them, J is the optimization target, d i Represents the error of the i-th position point, Δt i is the time to reach the i-th point, and λ is the time weight factor.

[0038] The data transmission and analysis module adopts a defect recognition algorithm based on high-dimensional features. Through cluster analysis of multi-dimensional image data, the following formula represents the construction of the feature matrix F:

[0039]

[0040] , where f j (X j , Y j ) is used as the feature extraction function for each image, combined with the clustering algorithm to identify the defects on the surface of the accessories and form the final inspection report.

[0041] The data transmission and analysis module also includes a deep learning-based anomaly detection network, which is trained using the following loss function to optimize the detection model:

[0042] Among them, L is the total loss, N is the number of samples, and y i is the actual value, is the predicted value, λ is the regularization parameter, θ j are model parameters to ensure the accuracy and robustness of the model in complex detection tasks.

[0043] A multi-directional machine vision inspection method for automobile parts comprises the following steps: a) using multiple image sensor modules to simultaneously capture images of automobile parts to be inspected from different viewing angles to ensure that all visible surfaces of the parts are covered; b) monitoring image data in real time, evaluating the impact of ambient lighting on image quality, automatically adjusting the intensity and angle of the light source, and eliminating the interference of reflection and shadow on image capture; c) driving a mechanical arm to move the parts at multiple angles so that the image sensor can shoot from all directions; d) transmitting all captured image data to a central processing unit for image preprocessing and feature extraction to generate feature vectors; e) analyzing the processed image data to identify potential defects and generate a defect candidate set; f) further analyzing the identified defects, verifying the results in combination with a deep learning model, and determining the specific location and severity of the defects; g) automatically generating a detailed inspection report based on the inspection and confirmation results, including the defect type, location and impact, and feeding the results back to the relevant production control system; h) collecting inspection data, analyzing the inspection performance and results, and optimizing the inspection algorithm and system settings based on historical data.

[0044] For the purpose of greater clarity, the invention is described in detail through the following embodiments.

[0045] In a specific embodiment, the image sensor module includes four 5-megapixel CMOS image sensors, which are installed at different positions of the system and arranged at vertical viewing angles of 90 degrees, 45 degrees, 0 degrees and -45 degrees to ensure that various surfaces of automotive accessories can be captured from different directions. The focal length of each sensor is 50 mm and is equipped with an optical filter to reduce ambient light interference. The frame rate of image acquisition is 30 frames per second.

[0046] In a specific embodiment, the adaptive light source control module is equipped with an adjustable LED ring light source with a light source intensity range of 0 to 3000 lumens. The angle can be adjusted by 360 degrees rotation through a stepper motor. The adaptive light source control system is adjusted based on the real-time light reflection data collected by the sensor, and the reflection intensity is monitored by a light intensity sensor with a feedback frequency of 1000 Hz.

[0047] In a specific embodiment, the robotic arm system uses a 6-degree-of-freedom robotic arm with an arm length of 1.2 meters and a maximum load-bearing weight of 5 kilograms. The robotic arm's motion range covers 360-degree rotation and ±90-degree pitch angles with an accuracy of 0.1 mm. The robotic arm is based on inverse kinematics control with a movement speed of 0.5 meters per second and can automatically adjust the path according to detection requirements.

[0048] The following is the experimental data:

[0049] Six common auto parts were selected, including gears, bearings, cams, door handles, etc., with sizes ranging from 50mm to 200mm and surface gloss ranging from 10GU to 90GU; the experimental ambient light was 5000K white light with a light intensity of 2000Lux. In the experiment, the brightness and angle of the LED light source were adjusted through the adaptive light source module to ensure that the surface reflection of the accessories was controlled within an acceptable range. The experimental results show that in terms of detection accuracy, the average defect recognition rate of the system under different gloss and angle conditions is 98.5%, which greatly improves the detection accuracy of the traditional detection system. At the same time, through the adjustment of the adaptive light source, the problem of reflective interference has been significantly improved. The motion accuracy of the robotic arm system ensures multi-angle detection on accessories with complex geometric shapes, achieving coverage without dead angles.

[0050] The specific implementation of the present invention combines multi-sensor, multi-angle acquisition technology, adaptive light source control and intelligent defect analysis method, effectively solves the problem of traditional system's inability to fully detect and light reflection interference, and significantly improves the detection accuracy and efficiency of automotive parts.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A multi-directional machine vision inspection system for automobile parts, characterized in that: The system comprises an image sensor module, an adaptive light source control module, a mechanical arm system and a data transmission and analysis module, wherein each image sensor module is configured to capture images of automobile accessories from multiple viewing angles; the adaptive light source control module is used to adjust the intensity and direction of the light source in the detection environment in real time to reduce the interference of the reflection of the accessory surface on the image capture; the adaptive light source control module calculates the gloss characteristics of the accessory surface and adjusts the angle and brightness of the light source according to the data captured by the image sensor; the mechanical arm system is used to drive the automobile accessories to be detected to move or rotate at multiple angles to ensure that the sensor can capture the entire surface of the accessories from all angles; the mechanical arm system can automatically adjust its motion path according to the shape, size and specific requirements of the accessory and the detection task; the mechanical arm system uses an inverse kinematics algorithm to accurately control the movement of the mechanical arm and dynamically adjust the motion path according to sensor feedback; the data transmission and analysis module is used to receive and process the image data obtained from each image sensor in real time, and analyze the detection results in combination with the defect recognition algorithm; the data transmission module can collect the data captured by the multi-directional sensor to the central processing unit, integrate the data, and generate a detection report.

2. The multi-directional machine vision inspection system for automobile parts according to claim 1 is characterized in that: The image sensor module includes a plurality of high-precision imaging units, which are configured to capture images using different viewing angles and focal lengths. The imaging model can be expressed as: i =f(X, Y, Z, θ i ,φ i ), where Z i The imaging unit S i The captured image data, (X, Y, Z) is the three-dimensional space coordinate, θ i and φ i are the corresponding elevation and azimuth angles.

3. The multi-directional machine vision inspection system for automobile parts according to claim 1 is characterized in that: The adaptive light source control module includes a light source adjustment algorithm module, which calculates the reflected light intensity based on the real-time captured image data. The light source compensation algorithm is based on the following formula: L′=L target -k·R(θ, φ), where L′ is the adjusted light source intensity, L target is the target light intensity, k is the adjustment coefficient, and R(θ, φ) represents the function of the current reflected light intensity.

4. The multi-directional machine vision inspection system for automobile parts according to claim 1, characterized in that: The motion path of the manipulator system is controlled by an inverse kinematics model, wherein the inverse kinematics model is based on the joint angle vector q=[q1, q2, ..., q m ] T , and solve the accessory position P by the following equation: Among them, l i is the length of the i-th joint of the robot, and h is the height from the base of the robot to the target plane.

5. The multi-directional machine vision inspection system for automobile parts according to claim 4, characterized in that: The robot system is equipped with a motion compensation mechanism based on a dynamic model, which uses a real-time motion trajectory optimization algorithm to adjust the path through the following optimization objective function: Among them, J is the optimization target, d i Represents the error of the i-th position point, Δt i is the time to reach the i-th point, and λ is the time weight factor.

6. The multi-directional machine vision inspection system for automobile parts according to claim 1, characterized in that: The data transmission and analysis module adopts a defect recognition algorithm based on high-dimensional features. Through cluster analysis of multi-dimensional image data, the following formula represents the construction of the feature matrix F: ， Among them, f j (X j , Y j ) is used as the feature extraction function for each image, combined with the clustering algorithm to identify the defects on the surface of the accessories and form the final inspection report.

7. The multi-directional machine vision inspection system for automobile parts according to claim 6, characterized in that: The data transmission and analysis module also includes a deep learning-based anomaly detection network, which is trained using the following loss function, and the following is an optimized detection model: Among them, L is the total loss, N is the number of samples, and y i is the actual value, is the predicted value, λ is the regularization parameter, θ j are model parameters to ensure the accuracy and robustness of the model in complex detection tasks.

8. A multi-directional machine vision inspection method for automobile parts, used to implement the multi-directional machine vision inspection system for automobile parts according to any one of claims 1 to 7, characterized in that: The following steps are involved: a) Use multiple image sensor modules to simultaneously capture images of the automotive parts to be inspected from different perspectives to ensure that all visible surfaces of the parts are covered; b) Monitor image data in real time, evaluate the impact of ambient lighting on image quality, automatically adjust the intensity and angle of the light source, and eliminate the interference of reflections and shadows on image capture; c) Drive the robotic arm to move the parts at multiple angles so that the image sensor can shoot from all directions; d) Transmit all captured image data to the central processing unit for image preprocessing and feature extraction to generate feature vectors; e) Analyze the processed image data, identify potential defects and generate defect candidate sets; f) Further analyze the identified defects, verify the results with the deep learning model, and determine the specific location and severity of the defects.

9. The multi-directional machine vision inspection method for automobile parts according to claim 8, characterized in that: Step f) also includes step g) automatically generating a detailed inspection report based on the inspection and confirmation results, including the defect type, location and impact degree, and feeding back the results to the relevant production control system.

10. The multi-directional machine vision inspection method for automobile parts according to claim 9, characterized in that: Step g) also includes h) collecting detection data, analyzing detection performance and results, and optimizing detection algorithms and system settings based on historical data.

Images