Automobile injection molding part surface roughness detection method and system

Abstract

The present application belongs to the field of image data processing technology of computer information science, and relates to a kind of automobile injection molding piece surface roughness detection method and system, its steps include: using industrial vision sensor to collect automobile injection molding piece image and execute standardization preprocessing, obtain automobile injection molding piece gray image;Based on automobile injection molding piece gray image, the gradient energy distribution of each pixel point in full circumference direction is obtained, and anisotropy characteristic value is calculated;Based on the average gray value of the current moment, the anisotropy characteristic value is compensated adaptively by illumination, and local roughness is obtained;The intensity value in the detection window is statistically integrated, and the quality evaluation index is evaluated in combination with the coefficient of variation;Based on the index, the cylinder action is driven in real time to execute and adjust the injection molding machine process parameters;It is simple and easy in process, and the application environment is friendly, reduces the disturbance of environmental illumination fluctuation and background texture on feature extraction, improves the robustness and precision of surface quality detection.

Description

Technical Field

[0001] This invention belongs to the field of image data processing technology in computer information science, and relates to a method and system for detecting the surface roughness of automotive injection molded parts, which is used in the automotive industry to scientifically detect and analyze the surface smoothness of injection molded parts. Background Technology

[0002] As a component that accounts for a large proportion of modern vehicle manufacturing, the physical morphology of automotive injection molded parts not only directly determines the visual quality of the vehicle, but is also closely related to its weather resistance and anti-aging performance during long-term service. In the complex injection molding process, molten polymer materials undergo high-speed filling and rapid cooling in the mold cavity. Due to the non-uniform distribution of thermal resistance on the mold cavity surface and the non-stationary fluctuations of the mold flow velocity in local areas, the cooling orientation and packing density of the polymer long chains will exhibit significant microscopic differences, which will lead to unexpected microscopic fluctuations on the surface of automotive injection molded parts, namely, abnormal fluctuations in surface roughness. At present, online monitoring of the surface quality of automotive injection molded parts generally adopts industrial vision sensing technology. By capturing optical images of the injection molded part surface, and using operators such as fast Fourier transform or gray-level co-occurrence matrix to extract the frequency domain energy distribution or second-order statistical features of the image, it is hoped to achieve the evaluation of surface roughness indicators.

[0003] In the existing technology, several patents have involved visual inspection schemes for the surface of injection molded parts. For example, Chinese invention patent application (publication number CN119090870A) discloses a method and system for detecting defects in automotive interior injection molded parts based on computer vision. It uses an initial threshold to perform threshold segmentation on the acquired injection molded part image to obtain possible defect areas, and uses the contrast between the possible defect areas and non-defect areas to obtain an adjusted threshold for secondary judgment. Chinese invention patent (authorization announcement number CN118334019B) discloses a method and system for detecting the injection molding quality of injection molded parts. This invention acquires grayscale images of injection molded parts, extracts texture morphology features and grayscale distribution features, and uses the morphology and grayscale value features of grayscale areas to filter out edge and spot abnormal areas. In addition, Chinese invention patent application (publication number CN119795519A) discloses a visual inspection method for burrs in automotive injection molded parts, which uses industrial vision methods to extract and calculate the edge features of injection molded parts to identify surface structure defects in the production process.

[0004] However, in actual industrial production line environments, such methods have significant technical limitations. On the one hand, to enhance the interior texture, the surfaces of automotive injection molded parts are often pre-molded with macroscopic decorative textures or imitation leather features in specific directions. These structured features have strong directional guidance in the spatial domain, and their corresponding gradient energy density is much higher than that of random roughness signals at the microscale. This makes it difficult for traditional frequency domain analysis methods to effectively decouple microscopic features from strong background interference, and they are prone to misjudging normal macroscopic texture fluctuations as roughness defects. On the other hand, the ambient illumination in injection molding production workshops is extremely complex and variable. Non-uniform local illumination will produce highly nonlinear reflection distributions on the surface of injection molded parts with complex geometric configurations. Existing algorithms generally lack the ability to adaptively model the physical reflection process of materials. When facing local highlight or shadow areas, because the algorithm cannot dynamically adjust the gain coefficient according to the real-time changes in ambient illumination, the extracted feature parameters often become severely distorted, leading to a large number of false alarms or missed detections, which makes it difficult to meet the needs of high-precision surface quality inspection of automotive injection molded parts. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and to provide a method and apparatus system for detecting the surface roughness of automotive injection molded parts. This invention addresses the technical problems of traditional techniques, such as difficulty in decoupling macroscopic texture interference and lack of adaptive compensation capability for environmental illumination fluctuations, which leads to low accuracy and poor robustness in surface roughness assessment.

[0006] To achieve the above-mentioned objectives, this invention provides a method for detecting the surface roughness of automotive injection molded parts, comprising: acquiring images of automotive injection molded parts using an industrial vision sensor at a preset sampling rate and performing standardized preprocessing to obtain a grayscale image of the automotive injection molded parts; obtaining the gradient energy distribution of each pixel in the full circumferential direction based on the grayscale image of the automotive injection molded parts, and obtaining anisotropic feature values ​​based on the gradient energy distribution; obtaining local roughness based on the anisotropic feature values ​​and the average grayscale value at the current moment; obtaining a quality evaluation index based on the local roughness within the total duration of the detection window; and controlling the execution of cylinder actions based on the quality evaluation index to realize the surface roughness detection of automotive injection molded parts.

[0007] This invention utilizes industrial vision sensors to collect signals and perform standardized preprocessing. Based on gradient energy distribution, it obtains anisotropic feature values ​​and performs adaptive illumination compensation. Finally, it controls the cylinder action according to the evaluated surface roughness index. By utilizing the difference between macroscopic texture and microscopic roughness in gradient energy distribution, it reduces the interference of inherent decorative textures on the surface of automotive injection molded parts and fluctuations in workshop ambient illumination on feature extraction, thereby improving the robustness of surface roughness detection.

[0008] The standardized preprocessing described in this invention includes: using a standard bilateral filtering algorithm to denoise the automotive injection molded part image, and converting the processed result into a two-dimensional arrangement of automotive injection molded part grayscale images.

[0009] This invention employs a standard bilateral filtering algorithm to denoise automotive injection molded part images. While suppressing random discrete noise in industrial settings, it retains edge information reflecting surface morphology, providing a relatively clean data foundation for subsequent feature analysis of automotive injection molded part grayscale images.

[0010] The anisotropic eigenvalues ​​described in this invention satisfy the expression: In the formula, Indicates time Anisotropic eigenvalues; Indicates the time sampling index; Indicates the spatial scanning direction angle; Indicates time Spatial scanning direction angle Gradient magnitude at the location; Represents pi; Indicates the scanning direction angle in all spaces. The maximum value of the gradient magnitude; Indicates time The average gradient vector; Indicates time The average gradient vector; Indicates the material attenuation coefficient; Indicates the zero constant; This indicates the modulo operation.

[0011] This invention obtains anisotropic feature values ​​based on gradient energy distribution. By utilizing the ratio of the gradient integral over the entire circumference to the maximum gradient modulus, it distinguishes between macroscopic textures with strong directional orientation and micro-roughness that tends towards isotropy, thereby reducing misjudgments of micro-roughness assessment by the structured texture features inherent on the surface of automotive injection molded parts.

[0012] The material attenuation coefficient described in this invention is obtained by irradiating the surface of the automotive injection molded part with a Gaussian beam and collecting the reflected light spot. The ratio of the diffusion diameter of the reflected light spot to the diameter of the incident beam is calculated and used as the material attenuation coefficient.

[0013] This invention utilizes a Gaussian beam to illuminate the surface of an automotive injection-molded part and calculates the ratio of the reflected light spot diffusion diameter to the incident beam diameter as a material attenuation coefficient. This adjusts the model's sensitivity to automotive injection-molded part materials with different transmittances and reduces feature extraction deviations caused by differences in the physical reflection properties of injection-molded materials.

[0014] The local roughness described in this invention satisfies the expression: In the formula, Indicates time Local roughness; Indicates time Anisotropic eigenvalues; Indicates time The local grayscale mean; Indicates the reference ambient brightness value; This represents the sensing sensitivity coefficient; Indicates time The maximum grayscale value in a grayscale image of an automotive injection molded part; Indicates the zero constant; This represents an exponential function with the natural constant as its base.

[0015] This invention obtains local roughness based on the difference between anisotropic eigenvalues ​​and local grayscale mean and reference ambient brightness value. It dynamically adjusts the feature gain using an exponential mapping relationship, reducing signal drift caused by fluctuations in ambient illumination at the production site and ensuring the consistency of roughness evaluation results under different brightness conditions.

[0016] The method for obtaining the sensing sensitivity coefficient described in this invention is as follows: collecting standard flatness automotive injection molded parts under different ambient illumination and calculating anisotropic characteristic values, statistically analyzing the linear regression slope of the anisotropic characteristic values ​​as a function of the average gray value, and using the reciprocal of this slope as the sensing sensitivity coefficient.

[0017] This invention obtains the perception sensitivity coefficient by statistically analyzing the linear regression slope of anisotropic feature values ​​as a function of average gray value under different illumination levels. Based on the physical reflectivity of the material, the compensation algorithm is calibrated to suppress the depth of local highlights, thereby reducing the nonlinear interference of local highlights or shadows caused by non-uniform illumination on feature extraction.

[0018] The quality evaluation index described in this invention satisfies the following expression: In the formula, Indicates a quality evaluation index; Indicates time Local roughness; Indicates the total duration of the detection window; This represents the standard deviation of local roughness over the total duration of the detection window. This represents the average local roughness over the total duration of the detection window; Indicates a dynamic penalty factor; This represents the zero constant.

[0019] This invention statistically integrates the intensity values ​​within the detection window and combines them with the coefficient of variation to evaluate the quality assessment index, smoothing out the instantaneous numerical jitter that may exist in a single frame signal. Through the statistical characteristics of time-series data, it more accurately reflects the overall state of the surface quality of automotive injection molded parts.

[0020] The dynamic penalty factor described in this invention is obtained by performing sliding variance analysis on an image sequence of automotive injection molded parts known to have local defects, calculating the variance ratio between the defective region and the normal region, and using this ratio as the dynamic penalty factor.

[0021] This invention utilizes the ratio of the sliding variance of the defective region to the normal region as a dynamic penalty factor, which enhances the model's ability to respond to local sudden roughness anomalies when calculating the quality evaluation index, and reduces the risk of local minor defects being masked by the time-series averaging algorithm.

[0022] The control of the actuator cylinder action described in this invention includes: when the quality evaluation index is detected to exceed a preset alarm threshold in real time, sending a deflection signal to the actuator cylinder to remove unqualified automotive injection molded parts, and simultaneously adjusting the injection molding machine's holding pressure compensation device.

[0023] This invention controls the execution cylinder to reject defective products and simultaneously adjusts the pressure compensation device of the injection molding machine when the index is detected to exceed the alarm threshold in real time. This realizes a closed-loop feedback from defect detection to process parameter correction, which helps to maintain the quality stability of automotive injection molded parts production process.

[0024] This invention also provides a surface roughness detection system for automotive injection molded parts, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the aforementioned method for detecting the surface roughness of automotive injection molded parts. By adopting the above technical solution, a computer program for detecting the surface roughness of automotive injection molded parts is generated and stored in the memory for loading and execution by the processor. This allows for the creation of a terminal device based on the memory and processor, facilitating its use.

[0025] Compared with existing technologies, the advantages of this invention are as follows: First, by analyzing the gradient energy distribution of pixels along the entire circumference and calculating anisotropic feature values, the invention decouples features by utilizing the directionality of macroscopic textures and the isotropic difference of microscopic roughness, reducing the interference of pre-set decorative textures on the surface of automotive injection molded parts on microscopic roughness detection and improving the accuracy of complex surface morphology recognition. Second, it introduces an adaptive compensation mechanism based on the deviation between the average gray value and the reference ambient brightness value, using an exponential function to nonlinearly scale the feature intensity, reducing the impact of non-uniform lighting and illuminance fluctuations in the production workshop on the detection results and enhancing the algorithm's adaptability to complex optical environments. Third, it evaluates the surface roughness index based on the time-series data of the detection window and drives cylinder action and adjusts injection molding process parameters accordingly. Statistical integration smooths out instantaneous noise interference, achieving automatic rejection of defective products and dynamic correction of production parameters, thus improving the quality control level of automated production lines for automotive injection molded parts. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a method for detecting the surface roughness of automotive injection molded parts according to the present invention.

[0027] Figure 2 This is a schematic diagram illustrating the effect of adaptive lighting compensation.

[0028] Figure 3 This is a schematic diagram illustrating the changes in surface defect detection and quality evaluation index. Detailed Implementation

[0029] The technical solution of the present invention will now be clearly and completely described in conjunction with the embodiments and accompanying drawings.

[0030] Example 1:

[0031] This embodiment discloses a method for detecting the surface roughness of automotive injection molded parts, referring to... Figure 1 This includes steps S1-S4:

[0032] S1. Use an industrial vision sensor to acquire images of automotive injection molded parts at a preset sampling rate and perform standardized preprocessing to obtain grayscale images of automotive injection molded parts.

[0033] To eliminate dark current noise generated by industrial vision sensors during data acquisition at the automotive injection molding production site and to unify the calculation benchmark for subsequent feature analysis, this embodiment must perform standardized preprocessing on the acquired raw surface signals. The use of spatial domain adaptive filtering technology can effectively suppress random discrete noise while fully preserving the edge information reflecting the surface morphology, providing a clean data carrier for the subsequent extraction of surface roughness features of automotive injection molding parts.

[0034] This embodiment utilizes an industrial vision sensor to acquire images of automotive injection molded parts at a preset sampling rate. The present invention employs a standard bilateral filtering algorithm to denoise the automotive injection molded part images and converts the processed results into two-dimensional grayscale images of automotive injection molded parts.

[0035] S2. Obtain the gradient energy distribution of each pixel in the full circumference direction based on the grayscale image of the automotive injection molded part, and obtain the anisotropic feature value based on the gradient energy distribution.

[0036] It should be noted that the macroscopic texture of the surface of automotive injection molded parts has a strong directional orientation in spatial distribution, while the abnormal surface roughness of automotive injection molded parts exhibits isotropic random speckle characteristics at the microscopic level. If the gradient modulus of the entire field is directly calculated, the energy intensity of the macroscopic texture will dominate, causing the microscopic fluctuation signal to be masked. Therefore, this invention constructs an anisotropic evaluation method based on the gradient direction angle energy distribution, and utilizes the difference in directional energy density between macroscopic and microscopic signals to achieve enhanced extraction of surface roughness information of automotive injection molded parts.

[0037] This embodiment performs multi-directional derivative calculations on the grayscale image of the automotive injection molded part to obtain the gradient energy distribution of each pixel in the grayscale image of the automotive injection molded part in the full circumferential direction, and calculates the anisotropic feature value.

[0038] The anisotropic eigenvalues ​​satisfy the expression:

[0039]

[0040] In the formula, Indicates time Anisotropic eigenvalues; Indicates the time sampling index; Indicates the spatial scanning direction angle; Indicates time Spatial scanning direction angle Gradient magnitude at the location; Represents pi; Indicates the scanning direction angle in all spaces. The maximum value of the gradient magnitude; Indicates time The average gradient vector; Indicates time The average gradient vector; Indicates the material attenuation coefficient; Represents the zero-prevention constant, an empirical value. ; This indicates the modulo operation.

[0041] In the formula, when the integral value of the square of the gradient magnitude in all directions is... It increases, and relative to the square of the maximum gradient. When the growth rate is faster, the value of the first term in the formula will increase; at the same time, when the magnitude of the difference between the average gradient vectors at adjacent times increases... Increases, in the material attenuation coefficient Under the regulation of anisotropic eigenvalues, the anisotropic eigenvalues ​​are ultimately driven. The calculated result becomes larger; anisotropic eigenvalues The larger the value, the more random the surface undulations of the automotive injection molded parts are at the microscale, that is, the more the surface micromorphology tends to be an isotropic rough state.

[0042] In this embodiment, the material attenuation coefficient is responsible for adjusting the model's sensitivity to automotive injection molded parts with different light transmittance. The material attenuation coefficient is obtained by irradiating the surface of the automotive injection molded part with a Gaussian beam and collecting the reflected light spot. The ratio of the diffusion diameter of the reflected light spot to the diameter of the incident beam is calculated and used as the material attenuation coefficient.

[0043] S3. Obtain local roughness based on anisotropic feature values ​​and the average gray value at the current time.

[0044] It should be noted that fluctuations in ambient illumination at the injection molding production site directly alter the average energy value of the grayscale image of automotive injection molded parts; with a fixed detection threshold, ambient illumination can lead to anisotropic eigenvalues. Synchronization offsets can occur, leading to misjudgments. To address this, this invention introduces an adaptive gain calculation method based on exponential mapping. This method uses the deviation between the mean local grayscale value and the baseline ambient brightness value as feedback to nonlinearly scale the feature intensity, ensuring the accuracy of the surface roughness assessment results for automotive injection molded parts under different physical environments.

[0045] This embodiment corrects the anisotropic feature values ​​based on the average gray level at the current moment to obtain the local roughness.

[0046] Local roughness satisfies the expression:

[0047]

[0048] In the formula, Indicates time Local roughness; Indicates time Anisotropic eigenvalues; Indicates time The local grayscale mean; Indicates the reference ambient brightness value; This represents the sensing sensitivity coefficient; Indicates time The maximum grayscale value in a grayscale image of an automotive injection molded part; This represents the zero constant.

[0049] In the formula, the local gray mean Compared with the reference ambient brightness value When the difference increases, it will cause the numerator to increase and cause... The calculation result within the function increases; simultaneously, the maximum gray value in the denominator term... The increase in this factor will have a suppressive effect on the sensitivity coefficient. Under the adjustment, the local roughness is ultimately reduced. In anisotropic eigenvalues Based on changes in ambient illuminance, dynamic scaling is achieved; local roughness The larger the value, the higher the degree of microscopic undulation of the physical surface corresponding to the pixel area after excluding the interference of ambient illumination.

[0050] Sensing sensitivity coefficient in this embodiment Used to control the depth of suppression of local highlights by the compensation algorithm; perception sensitivity coefficient The method of obtaining the data is as follows: collect standard flatness automotive injection molded parts under different ambient illumination conditions and calculate anisotropic characteristic values. Calculate the linear regression slope of the anisotropic characteristic values ​​as a function of the average gray value, and use the reciprocal of the slope as the perception sensitivity coefficient.

[0051] For example, Figure 2 This is a schematic diagram illustrating the effect of adaptive illumination compensation. The figure shows the changes in uncompensated anisotropic eigenvalues ​​and compensated local roughness over time when ambient illumination fluctuates. Due to the instability of workshop illumination, the anisotropic eigenvalues ​​exhibit significant drift related to illumination intensity, failing to accurately reflect the surface condition. However, this embodiment introduces an exponential mapping mechanism based on the deviation between average grayscale and reference brightness, ensuring that the calculated local roughness remains highly stable throughout the process. This demonstrates that the method effectively reduces the interference of ambient illumination fluctuations on feature extraction and guarantees the consistency of detection results.

[0052] S4. Obtain the quality evaluation index based on the local roughness within the total detection window duration; control the cylinder action based on the quality evaluation index to realize the surface roughness detection of automotive injection molded parts.

[0053] It should be noted that the feature output of a single frame may be subject to numerical jitter caused by transient interference. In order to achieve robust quality judgment, this embodiment must perform statistical evaluation on the features within the time series to obtain the final evaluation parameters. By calculating the combination of the first moment and the coefficient of variation of the feature field, the quality evaluation index is evaluated and used as the control criterion for the production line execution cylinder.

[0054] This embodiment integrates data within a preset testing period to calculate the final quality evaluation index.

[0055] The quality evaluation index satisfies the expression:

[0056]

[0057] In the formula, Indicates a quality evaluation index; Indicates time Local roughness; Indicates the total duration of the detection window; This represents the standard deviation of local roughness over the total duration of the detection window. This represents the average local roughness over the total duration of the detection window; Indicates a dynamic penalty factor; This represents the zero constant.

[0058] In the formula, the local roughness Total detection window duration The sum of the contents When the value increases, the value of the first term in the formula increases; simultaneously, the standard deviation of the local roughness... Compared with the average The ratio increases with the dynamic penalty factor. Under the adjustment of [the system / mechanism], the final driving force of the quality evaluation index is [the following]. The value increases; the quality evaluation index The larger the value, the worse the overall quality of the automotive injection molded part surface during the total inspection window, and the more its micro-roughness exceeds the preset process standard.

[0059] The dynamic penalty factor in this embodiment Used to enhance the model's sensitivity to localized sudden roughness defects; dynamic penalty factor The method for obtaining the intensity value is as follows: the surface roughness intensity field of the automotive injection molded part in the sample image with known local defects is binarized and segmented using the Otsu method. The set of pixels with intensity values ​​higher than the segmentation threshold is defined as the defect region, and the set of pixels with intensity values ​​lower than or equal to the segmentation threshold is defined as the normal region. The spatial distribution variance of the intensity values ​​of each point in the defect region and the spatial distribution variance of the intensity values ​​of each point in the normal region are calculated respectively, and the ratio of the two is used as the parameter, which has an empirical range of 0.1 to 0.3.

[0060] In this embodiment, the calculated quality evaluation index is used to send a deflection signal to the actuator cylinder when the value exceeds the preset alarm threshold in real time. The actuator cylinder then removes the defective automotive injection molded parts to the waste collection tank according to the deflection signal, and simultaneously adjusts the pressure compensation device of the injection molding machine to correct the production process parameters, thus realizing closed-loop management of the production quality of automotive injection molded parts.

[0061] For example, Figure 3 This diagram illustrates the relationship between surface defect detection and the quality evaluation index. It shows the temporal response of local roughness and the quality evaluation index when a local defect appears on the surface of an automotive injection-molded part. During the detection process, the local roughness undergoes a sudden change in the defect area. The quality evaluation index, by integrating the mean and coefficient of variation within the detection window, responds significantly to this change, rapidly rising and exceeding a preset alarm threshold. This triggers a cylinder action to remove defective products, demonstrating that this method can capture local defects and achieve quality control.

[0062] This embodiment also discloses a surface roughness detection system for automotive injection molded parts, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a surface roughness detection method for automotive injection molded parts according to the present invention.

[0063] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for detecting the surface roughness of automotive injection molded parts, characterized in that, The process includes the following steps: The image of the automotive injection molded part is obtained by using an industrial vision sensor to acquire images of automotive injection molded parts at a preset sampling rate and performing standardized preprocessing. The gradient energy distribution of each pixel along the full circumference is obtained from the grayscale image of the automotive injection molded part. Anisotropic feature values ​​are then obtained based on the gradient energy distribution, satisfying the expression: ； In the formula, Indicates time Anisotropic eigenvalues; Indicates the time sampling index; Indicates the spatial scanning direction angle; Indicates time Spatial scanning direction angle Gradient magnitude at the location; Represents pi; Indicates the scanning direction angle in all spaces. The maximum value of the gradient magnitude; Indicates time The average gradient vector; Indicates time The average gradient vector; Indicates the material attenuation coefficient; Indicates the zero constant; This represents the modulo operation; Local roughness is obtained based on anisotropic eigenvalues ​​and the average gray value at the current moment, satisfying the expression: ； In the formula, Indicates time Local roughness; Indicates time Anisotropic eigenvalues; Indicates time The local grayscale mean; Indicates the reference ambient brightness value; This represents the sensing sensitivity coefficient; Indicates time The maximum grayscale value in a grayscale image of an automotive injection molded part; Indicates the zero constant; Represents an exponential function with the natural constant as its base; The quality evaluation index is obtained based on the local roughness within the total detection window duration; the cylinder action is controlled based on the quality evaluation index to realize the surface roughness detection of automotive injection molded parts.

2. The method for detecting the surface roughness of automotive injection molded parts according to claim 1, characterized in that, The standardized preprocessing includes: The standard bilateral filtering algorithm is used to denoise the automotive injection molded part image, and the processed result is converted into a two-dimensional grayscale image of the automotive injection molded part.

3. The method for detecting the surface roughness of automotive injection molded parts according to claim 1, characterized in that, The method for obtaining the material attenuation coefficient is as follows: A Gaussian beam is used to illuminate the surface of an automotive injection molded part and the reflected light spot is collected. The ratio of the diffusion diameter of the reflected light spot to the diameter of the incident beam is calculated and used as the material attenuation coefficient.

4. The method for detecting the surface roughness of automotive injection molded parts according to claim 1, characterized in that, The sensing sensitivity coefficient is obtained as follows: Automotive injection molded parts with standard flatness were collected under different ambient illuminance and anisotropic characteristic values ​​were calculated. The slope of the linear regression of the anisotropic characteristic values ​​with the change of the average gray value was statistically analyzed, and the reciprocal of the slope was used as the sensing sensitivity coefficient.

5. The method for detecting the surface roughness of automotive injection molded parts according to claim 1, characterized in that, The quality evaluation index satisfies the expression: ； In the formula, Indicates a quality evaluation index; Indicates time Local roughness; Indicates the total duration of the detection window; This represents the standard deviation of local roughness over the total duration of the detection window. This represents the average local roughness over the total duration of the detection window; Indicates a dynamic penalty factor; This represents the zero constant.

6. The method for detecting the surface roughness of automotive injection molded parts according to claim 5, characterized in that, The dynamic penalty factor is obtained as follows: A sliding variance analysis was performed on an image sequence of automotive injection molded parts with known local defects to calculate the variance ratio between the defective region and the normal region, and this ratio was used as a dynamic penalty factor.

7. The method for detecting the surface roughness of automotive injection molded parts according to claim 1, characterized in that, The control of the cylinder action includes: When the quality evaluation index is detected to exceed the preset alarm threshold in real time, a deflection signal is sent to the actuator cylinder to remove unqualified automotive injection molded parts, and the pressure compensation device of the injection molding machine is adjusted simultaneously.

8. A surface roughness detection system for automotive injection molded parts, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a method for detecting the surface roughness of automotive injection molded parts according to any one of claims 1-7.

Images