Deep learning-based industrial part surface defect detection method

Abstract

The application discloses a deep learning-based industrial part surface defect detection method, and relates to the technical field of data processing, and comprises the following implementation steps: configuring an industrial part surface defect detection control end server, entering a uniform illumination monitoring end, judging illumination abnormalities in real time by combining dynamic changes in the environment, and capturing illumination shadows in real time; entering a trailing shadow early warning end, early warning whether industrial part surface defect detection is abnormal by combining vibration parameters; and entering a multi-angle irradiation adjustment end, detecting whether the presentation angle of the industrial part deviates in real time during dynamic adjustment. The deep learning-based industrial part surface defect detection method can timely discover whether uneven illumination occurs due to dynamic changes in the imaging environment, can timely discover whether imaging trailing shadows caused by motion blurring occur during industrial part surface defect detection in a dynamic state, and improves the accuracy and detection effect of industrial part surface defect detection.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for detecting surface defects in industrial parts based on deep learning. Background Technology

[0002] Deep learning-based industrial part surface defect detection is an intelligent technology that uses deep neural networks to automatically identify, locate, and classify surface anomalies of parts, including cracks, scratches, pits, and rust. By learning abstract features of defects from a large amount of image data, it replaces traditional detection methods that rely on manually designed features, significantly improving detection efficiency and accuracy, and becoming a key breakthrough in the field of industrial quality inspection.

[0003] Currently, there are some shortcomings in the detection of surface defects in industrial parts: 1. Imaging conditions are easily affected during the detection of surface defects in industrial parts. Due to the dynamic changes in the environment, uneven lighting can occur, and shadows and reflective areas can easily obscure defects, affecting the accuracy of defect detection. 2. During the detection of defects in industrial parts, motion blur is prone to occur in dynamic states, resulting in image ghosting, which further reduces the accuracy of defect contour extraction. 3. At the same time, the angle of the part is dynamically adjusted during defect detection. During dynamic adjustment, it is impossible to detect whether the viewing angle of the industrial part has shifted in real time, which further affects the accuracy of surface defect detection and cannot guarantee the detection accuracy of surface defects of industrial parts of different sizes.

[0004] Therefore, a deep learning-based method for detecting surface defects in industrial parts is proposed to address the aforementioned problems. Summary of the Invention

[0005] The main objective of this invention is to provide a deep learning-based method for detecting surface defects in industrial parts, in order to solve the problems mentioned in the background above.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is: a deep learning-based method for detecting surface defects in industrial parts, the method comprising the following implementation steps:

[0007] Step 1: Configure the industrial parts surface defect detection control server, enter the light uniformity monitoring terminal, collect the imaging environment of industrial parts surface defect detection in real time, and judge whether the imaging conditions are affected by light interference in real time. Combine the dynamic changes of the environment to judge the light anomaly in real time, capture the light and shadow in real time, and locate the reflective area in real time.

[0008] Step 2: Enter the image trailing warning terminal, collect the imaging parameters during industrial part defect detection in real time, capture the movement of industrial part defect detection in real time under dynamic conditions, judge whether the defect detection image shows image trailing in real time, and combine vibration parameters to give early warning of whether the surface defect detection of industrial parts is abnormal.

[0009] Step 3: Enter the multi-angle illumination adjustment end. During defect detection, the part angle is dynamically adjusted in real time, and the presenting angle of the industrial part is detected in real time during the dynamic adjustment. The presenting angle abnormality is corrected in real time in combination with the workpiece shape and height to ensure the accuracy of surface defect detection of industrial parts. The multi-angle real-time fusion of multiple light sources solves the reflection problem and ensures the surface defect detection accuracy of industrial parts of different sizes.

[0010] The uniformity of illumination monitoring terminal includes an imaging environment acquisition unit, an interference judgment unit, and an illumination and shadow capture unit.

[0011] The imaging environment acquisition unit is used to acquire the imaging environment for detecting surface defects of industrial parts in real time through a data acquisition instrument. The imaging environment includes temperature parameters, humidity parameters, illumination parameters, brightness parameters, resolution parameters, and vibration parameters, and sets the surface defect detection area of ​​industrial parts.

[0012] The interference judgment unit is used to determine in real time whether the imaging conditions are affected by illumination interference, taking into account the imaging environment. The judgment method is as follows:

[0013] Average brightness deviation in computational imaging environment The calculation formula is as follows:

[0014] ;

[0015] in, This represents the average grayscale value of the current frame image. This represents the brightness parameter under standard imaging conditions, and a deviation threshold is set. If the percentage is greater than 20%, an alarm signal will be triggered.

[0016] Computational imaging environment illumination uniformity The calculation formula is as follows:

[0017] ;

[0018] In calculating illumination uniformity, the image is divided into multiple grids. This represents the average brightness of each grid area. Indicates the maximum brightness of the grid area. Indicates the minimum brightness of the grid area;

[0019] Calculate illumination correction values ​​in dynamic imaging environments The calculation formula is as follows:

[0020] ;

[0021] in, This represents the illumination correction value under imaging conditions. Indicates the ambient light level during calibration. Current ambient light level;

[0022] like If the value is 0, it indicates that the imaging conditions are not affected by illumination. If the value is not equal to 0, it is determined that the imaging conditions are affected by lighting interference, and the reporting system issues a voice alarm.

[0023] The light and shadow capture unit includes a unit for locating reflective areas and a multi-source illumination unit;

[0024] The positioning reflective area unit is used to scan the maximum brightness of the grid area in real time by a scanner and to perform reflective positioning in real time by a positioning device.

[0025] The multi-source illumination unit is used to illuminate the defect detection area in real time by setting up multiple light source illumination devices.

[0026] The motion blur monitoring and early warning terminal includes an imaging parameter acquisition unit, a dynamic capture operation unit, an imaging motion blur judgment unit, and a combined vibration early warning unit.

[0027] The imaging parameter acquisition unit is used to acquire imaging parameters in real time during defect detection of industrial parts through a data acquisition instrument. The imaging parameters include light source parameters, wavelength matching parameters, brightness, uniformity, resolution, lens parameters, environmental parameters, mechanical vibration parameters, temperature parameters, humidity parameters, part contour parameters, workpiece angle parameters, and spatial filtering parameters.

[0028] The dynamic capture operation unit is used to capture the real-time motion of industrial parts for defect detection in a dynamic state. The real-time motion includes imaging parameters.

[0029] The imaging motion detection unit is used to determine in real time whether imaging motion occurs in the defect detection image, and to calculate the imaging motion value that occurs during the detection of surface defects on the part at the current moment. The calculation formula is as follows:

[0030] ;

[0031] in, Indicates the amplitude of vibration. Indicates the exposure time. Indicates the vibration frequency. Indicates pixel size, if If the value is 0, it is determined that the defect detection image has imaging ghosting; otherwise, it is determined that the defect detection image does not have imaging ghosting.

[0032] The vibration early warning unit is used to provide early warning of whether the surface defect detection of industrial parts is abnormal by combining vibration parameters generated by real-time motion. The early warning method is as follows:

[0033] Calculate the abnormal warning value of the vibration parameters generated by the motion at the current moment for the detection of surface defects of industrial parts. The calculation formula is as follows:

[0034] ;

[0035] Among them, if the abnormal warning value A value greater than or equal to 1 indicates an abnormality in the surface defect detection of industrial parts, meaning that motion blur caused by image trailing occurred during the detection process. The system will be notified of this anomaly and an alert will be issued. Conversely, a value of 1 indicates that the surface defect detection of industrial parts is proceeding normally, meaning that motion blur caused by image trailing did not occur during the detection process. This represents the image trailing value that appears during the detection of surface defects on the part at the current moment. Indicates the maximum permissible image motion blur. This represents the decrease in the vibration signal-to-noise ratio caused by vibration. This indicates the maximum permissible vibration signal-to-noise loss. The root mean square value of the vibration acceleration. Indicates the safe vibration acceleration threshold. , and This represents the weight coefficient of each sub-item, and is adjusted in real time according to the detection scenario.

[0036] The multi-angle illumination adjustment end includes a part angle adjustment unit, an angle offset calculation unit, an angle anomaly correction unit, and a multi-source fusion unit;

[0037] The part angle adjustment unit is used to adjust the detection angle of surface defects of the part in real time through the driving device, and to monitor the adjustment angle in real time through the camera.

[0038] The angle offset calculation unit is used to calculate in real time the offset value of the surface defect detection angle of industrial parts during dynamic adjustment. The calculation formula is as follows:

[0039] ;

[0040] in, This indicates the angle of refraction under the adjusted angle. Indicates the angle of refraction at the detection angle. This indicates the angle between the detection angle and the adjustment angle. If the deviation is greater than 1°, it is determined that the detection angle of surface defects of industrial parts has deviated; otherwise, it is determined that the detection angle of surface defects of industrial parts has not deviated.

[0041] The angle anomaly correction unit is used to correct the surface defect detection angle in the industrial field in real time according to the offset value, until the offset value is less than 1°.

[0042] The multi-light source fusion unit is used to achieve multi-light source fusion of multiple light source illumination devices, and the fusion formula is as follows:

[0043] ;

[0044] in, Indicates from the first Subtract the static background from the current frame of each light source, and retain the illumination area where the target changes. Indicates the target mask to be detected. Indicates the first The detection area under illumination by a single light source Indicates the first The original input image under a single light source This refers to the clean illumination area obtained after the light sources from multiple illumination devices are combined.

[0045] The present invention has the following beneficial effects:

[0046] 1. In this invention, by setting up a uniform illumination monitoring end, when detecting surface defects of industrial parts based on deep learning, the imaging conditions are judged in real time to see if they are interfered with by illumination, and whether there are illumination shadows during the illumination process. This prevents the imaging conditions from being interfered with during the detection of surface defects of industrial parts, and timely detects whether the dynamically changing imaging environment causes uneven illumination. This reduces the shadows in the detection of surface defects of workpiece parts, prevents reflective areas from covering defects during detection, and improves the accuracy of defect detection.

[0047] 2. In this invention, by setting up a motion blur monitoring and early warning terminal, when detecting surface defects of industrial parts based on deep learning, the real-time motion of the industrial part defect detection is captured in real time, and the vibration parameters generated by the real-time motion are combined to give an early warning of whether the industrial part surface defect detection is abnormal. This enables timely detection of whether motion blur occurs during the detection of surface defects of industrial parts in a dynamic state, thereby further improving the accuracy of contour extraction of surface defects of industrial parts.

[0048] 3. In this invention, by setting up a multi-angle illumination adjustment end, when detecting surface defects of industrial parts based on deep learning, the presenting angle of the industrial parts is detected in real time to see if there is a deviation. The presenting angle abnormality is corrected in real time by combining the workpiece shape and height. This ensures the accuracy of surface defect detection of industrial parts. Furthermore, by using multi-angle real-time fusion of multiple light sources to solve reflection, the stability and practicality of the surface defect detection accuracy of industrial parts of different sizes can be guaranteed, further improving the accuracy and detection effect of surface defect detection of industrial parts. Attached Figure Description

[0049] Figure 1 This is an overall flowchart of the deep learning-based method for detecting surface defects in industrial parts according to the present invention.

[0050] Figure 2 This is a schematic diagram of the illumination uniformity monitoring end of the deep learning-based industrial part surface defect detection method of the present invention;

[0051] Figure 3 This is a schematic diagram of the architecture of the trailing image monitoring and early warning terminal of the deep learning-based industrial part surface defect detection method of the present invention;

[0052] Figure 4 This is a schematic diagram of the multi-angle illumination adjustment end of the deep learning-based industrial part surface defect detection method of the present invention. Detailed Implementation

[0053] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0054] Example 1

[0055] Please refer to Figures 1 to 2 The following is an example of a deep learning-based method for detecting surface defects in industrial parts. The method includes the following implementation steps:

[0056] Step 1: Configure the industrial parts surface defect detection control server, enter the light uniformity monitoring terminal, collect the imaging environment of industrial parts surface defect detection in real time, and judge whether the imaging conditions are affected by light interference in real time. Combine the dynamic changes of the environment to judge the light anomaly in real time, capture the light and shadow in real time, and locate the reflective area in real time.

[0057] Step 2: Enter the image trailing warning terminal, collect the imaging parameters during industrial part defect detection in real time, capture the movement of industrial part defect detection in real time under dynamic conditions, judge whether the defect detection image shows image trailing in real time, and combine vibration parameters to give early warning of whether the surface defect detection of industrial parts is abnormal.

[0058] Step 3: Enter the multi-angle illumination adjustment end. During defect detection, the part angle is dynamically adjusted in real time, and the presenting angle of the industrial part is detected in real time during the dynamic adjustment. The presenting angle abnormality is corrected in real time in combination with the workpiece shape and height to ensure the accuracy of surface defect detection of industrial parts. The multi-angle real-time fusion of multiple light sources solves the reflection problem and ensures the surface defect detection accuracy of industrial parts of different sizes.

[0059] The illumination uniformity monitoring unit includes an imaging environment acquisition unit, an interference judgment unit, and an illumination and shadow capture unit;

[0060] The imaging environment acquisition unit is used to acquire the imaging environment for surface defect detection of industrial parts in real time through a data acquisition instrument. The imaging environment includes temperature parameters, humidity parameters, illumination parameters, brightness parameters, resolution parameters, and vibration parameters, and sets the surface defect detection area of ​​industrial parts.

[0061] The interference detection unit is used to determine in real time whether the imaging conditions are affected by illumination interference, taking into account the imaging environment. The detection method is as follows:

[0062] Average brightness deviation in computational imaging environment The calculation formula is as follows:

[0063] ;

[0064] in, This represents the average grayscale value of the current frame image. This represents the brightness parameter under standard imaging conditions, and a deviation threshold is set. If the percentage is greater than 20%, an alarm signal will be triggered.

[0065] Computational imaging environment illumination uniformity The calculation formula is as follows:

[0066] ;

[0067] In calculating illumination uniformity, the image is divided into multiple grids. This represents the average brightness of each grid area. Indicates the maximum brightness of the grid area. Indicates the minimum brightness of the grid area;

[0068] Calculate illumination correction values ​​in dynamic imaging environments The calculation formula is as follows:

[0069] ;

[0070] in, This represents the illumination correction value under imaging conditions. Indicates the ambient light level during calibration. Current ambient light level;

[0071] like If the value is 0, it indicates that the imaging conditions are not affected by illumination. If the value is not equal to 0, it is determined that the imaging conditions are affected by lighting interference, and the reporting system issues a voice alarm.

[0072] The lighting and shadow capture unit includes a unit for locating reflective areas and a multi-source illumination unit;

[0073] The positioning reflective area unit is used to scan the maximum brightness of the grid area in real time by a scanner and to perform reflective positioning in real time by a positioning device.

[0074] The multi-source illumination unit is used to illuminate the defect detection area in real time by setting multiple light source illumination devices. By capturing whether light and shadow appear during the illumination process and locating the reflective area in real time, it can prevent the imaging conditions from being interfered with when detecting surface defects of industrial parts, and promptly detect whether the dynamically changing imaging environment causes uneven illumination, thereby reducing the shadows in the detection of surface defects of workpiece parts.

[0075] Example 2

[0076] Please refer to Figure 3 As shown: Based on Embodiment 1, the image trailing monitoring and early warning terminal includes an imaging parameter acquisition unit, a dynamic capture operation unit, an imaging trailing judgment unit, and a combined vibration early warning unit;

[0077] The imaging parameter acquisition unit is used to acquire imaging parameters in real time during defect detection of industrial parts through a data acquisition instrument. The imaging parameters include light source parameters, wavelength matching parameters, brightness, uniformity, resolution, lens parameters, environmental parameters, mechanical vibration parameters, temperature parameters, humidity parameters, part contour parameters, workpiece angle parameters, and spatial filtering parameters.

[0078] The dynamic capture unit is used to capture the real-time motion of industrial parts for defect detection in a dynamic state. The real-time motion includes imaging parameters.

[0079] The image trailing detection unit is used to determine in real time whether image trailing occurs in the defect detection image and calculates the image trailing value that occurs during the detection of surface defects on the part at the current moment. The calculation formula is as follows:

[0080] ;

[0081] in, Indicates the amplitude of vibration. Indicates the exposure time. Indicates the vibration frequency. Indicates pixel size, if If the value is 0, it is determined that the defect detection image has imaging ghosting; otherwise, it is determined that the defect detection image does not have imaging ghosting.

[0082] A vibration early warning unit is used to provide early warnings of abnormalities in the detection of surface defects on industrial parts by combining vibration parameters generated by real-time motion. The early warning method is as follows:

[0083] Calculate the abnormal warning value of the vibration parameters generated by the motion at the current moment for the detection of surface defects of industrial parts. The calculation formula is as follows:

[0084] ;

[0085] Among them, if the abnormal warning value A value greater than or equal to 1 indicates an abnormality in the surface defect detection of industrial parts, meaning that motion blur caused by image trailing occurred during the detection process. The system will be notified of this anomaly and an alert will be issued. Conversely, a value of 1 indicates that the surface defect detection of industrial parts is proceeding normally, meaning that motion blur caused by image trailing did not occur during the detection process. This represents the image trailing value that appears during the detection of surface defects on the part at the current moment. Indicates the maximum permissible image motion blur. This represents the decrease in the vibration signal-to-noise ratio caused by vibration. This indicates the maximum permissible vibration signal-to-noise loss. The root mean square value of the vibration acceleration. Indicates the safe vibration acceleration threshold. , and It represents the weight coefficient of each sub-item and adjusts it in real time according to the detection scenario. Combined with the vibration parameters generated by real-time motion, it provides early warning of whether there are abnormalities in the detection of surface defects of industrial parts. This enables timely detection of whether motion blur causes imaging ghosting when detecting surface defects of industrial parts in a dynamic state.

[0086] Example 3

[0087] Please refer to Figure 4 As shown: Based on Embodiment 1, the multi-angle illumination adjustment end includes a part angle adjustment unit, an angle offset calculation unit, an angle anomaly correction unit, and a multi-light source fusion unit;

[0088] The part angle adjustment unit is used to adjust the detection angle of surface defects of the part in real time through the drive device, and the adjustment angle is monitored in real time by the camera.

[0089] The angle offset calculation unit is used to calculate in real time the offset value of the surface defect detection angle of industrial parts during dynamic adjustment. The calculation formula is as follows:

[0090] ;

[0091] in, This indicates the angle of refraction under the adjusted angle. Indicates the angle of refraction at the detection angle. This indicates the angle between the detection angle and the adjustment angle. If the deviation is greater than 1°, it is determined that the detection angle of surface defects of industrial parts has deviated; otherwise, it is determined that the detection angle of surface defects of industrial parts has not deviated.

[0092] The angle anomaly correction unit is used to correct the surface defect detection angle in the industrial field in real time according to the offset value until the offset value is less than 1°.

[0093] The multi-light source fusion unit is used to achieve multi-light source fusion of multiple illumination devices. The fusion formula is as follows:

[0094] ;

[0095] in, Indicates from the first Subtract the static background from the current frame of each light source, and retain the illumination area where the target changes. This refers to the light source produced by the light source device, and =1, 2, 3, ... ,and It is a positive integer. Indicates the target mask to be detected. Indicates the first The detection area under illumination by a single light source Indicates the first The original input image under a single light source This refers to the clean illumination area obtained after the fusion of light sources from multiple light source irradiation devices. By solving the reflection problem through real-time fusion of multiple light sources from multiple angles, it can ensure the stability and practicality of the surface defect detection accuracy of industrial parts of different sizes. It also allows the industrial angle to be dynamically adjusted, enabling real-time detection of whether the presentation angle of industrial parts has shifted during dynamic adjustment.

[0096] In this invention, a deep learning-based method for detecting surface defects in industrial parts first configures a control terminal for surface defect detection. This terminal enters a lighting uniformity monitoring end, where a data acquisition device collects real-time data on the imaging environment of the industrial part surface defect detection. The system then analyzes the imaging environment to determine if the imaging conditions are affected by lighting interference. During this analysis, it considers dynamic changes in the imaging environment to detect lighting anomalies, captures whether shadows appear during the illumination process, and locates reflective areas in real time. This prevents interference with imaging conditions during surface defect detection and promptly detects whether dynamic changes in the imaging environment cause uneven lighting, reducing shadows on the workpiece surface and preventing reflective areas from obscuring defects, thus improving defect detection accuracy. Next, a motion blur monitoring and early warning end is established. During deep learning-based surface defect detection, the system collects real-time imaging parameters and captures the real-time movement of the industrial part defect detection. Based on this real-time movement, it analyzes whether motion blur appears in the defect detection image. By combining vibration parameters generated by real-time motion, early warnings are provided to detect abnormalities in the surface defect detection of industrial parts. This allows for timely detection of motion blur and image trailing during dynamic surface defect detection, further improving the accuracy of contour extraction for surface defects. In the multi-angle illumination adjustment stage, during deep learning-based surface defect detection, the part's angle is dynamically adjusted in real-time. During this adjustment, any shifts in the presented angle are detected, and real-time corrections are made based on the workpiece's shape and height to ensure accurate surface defect detection. Furthermore, multi-angle real-time fusion of multiple light sources addresses reflection issues, ensuring the stability and practicality of surface defect detection accuracy for industrial parts of different sizes. This allows for real-time detection of angle shifts during dynamic adjustment, further improving the accuracy and effectiveness of surface defect detection. The entire method is simple and efficient.

[0097] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A deep learning-based industrial part surface defect detection method, characterized in that, The method comprises the following implementation steps: Step 1: Configure the industrial part surface defect detection control end server, enter the light uniformity monitoring end, collect the imaging environment of the industrial part surface defect detection in real time, judge whether the imaging condition is disturbed by light in real time, judge the light abnormality in real time combined with the dynamic change of the environment, capture the light shadow in real time, and locate the light reflection area in real time; Step 2: Enter the trailing shadow early warning end, collect the imaging parameters of the industrial part defect detection in real time, capture the industrial part defect detection movement in a dynamic state in real time, judge whether the defect detection image appears imaging trailing shadow in real time, and early warning whether the industrial part surface defect detection is abnormal combined with the vibration parameters; Step 3: Enter the multi-angle irradiation adjustment end, dynamically adjust the part angle during defect detection in real time, detect whether the presented angle of the industrial part deviates during dynamic adjustment in real time, correct the presented angle abnormality in real time combined with the workpiece shape and height, ensure the accuracy of the industrial part surface defect detection, and solve the light reflection by real-time fusion of multi-angle and multi-light source, ensure the precision of the industrial part surface defect detection of different sizes; The light uniformity monitoring end comprises an imaging environment acquisition unit, an interference judgment unit and a light shadow capture unit; The imaging environment acquisition unit is used for collecting the imaging environment of the industrial part surface defect detection in real time through a data acquisition instrument, the imaging environment comprises temperature parameters, humidity parameters, light parameters, brightness parameters, resolution parameters and vibration parameters, and the industrial part surface defect detection area is set; The interference judgment unit is used for judging whether the imaging condition is disturbed by light in real time combined with the imaging environment, and the judgment method is as follows: Average luminance deviation in computed imaging environment The calculation formula is as follows: ； wherein, represents the average gray value of the current frame image, represents the brightness parameter under the standard imaging environment, sets the deviation threshold, if greater than 20%, the alarm signal is triggered; Illumination uniformity in computed imaging environments The formula is as follows: ； wherein the image is divided into a plurality of grids when calculating the illumination uniformity, represents an average brightness of each grid region, represents a maximum brightness of the grid region, represents a minimum brightness of the grid region; Computing illumination correction values in a dynamic imaging environment The formula is as follows: ； wherein, represents a light correction value in an imaging environment, represents an ambient light intensity at calibration, represents a current ambient light intensity; If equal to 0, it is judged that the imaging condition is not disturbed by light, and if not equal to 0, it is judged that the imaging condition is disturbed by light, and a voice alarm is issued to remind the system. The light shadow capture unit comprises a light reflection area positioning unit and a multi-light source irradiation unit; The light reflection area positioning unit is used for scanning the maximum brightness of the grid area in real time through a scanner, and positioning the light reflection in real time through a positioner; The multi-light source irradiation unit is used for irradiating the defect detection area with multiple light sources through multiple light source irradiation devices; The trailing shadow early warning end comprises an imaging parameter acquisition unit, a dynamic capture running unit, an imaging trailing shadow judgment unit and a vibration combined early warning unit; The imaging parameter acquisition unit is used for collecting the imaging parameters of the industrial part defect detection in real time through a data acquisition instrument, the imaging parameters comprise light source parameters, wavelength matching parameters, brightness, uniformity, resolution, lens parameters, environment parameters, mechanical vibration parameters, temperature parameters, humidity parameters, part contour parameters, workpiece angle parameters and spatial filtering parameters; The dynamic capture running unit is used for capturing the real-time movement of the industrial part defect detection in real time in a dynamic state; The imaging lag judgment unit is configured to judge whether imaging lag occurs in the defect detection image in real time according to real-time movement, and calculate an imaging lag value occurring in the part surface defect detection at the current time The calculation formula is as follows: ； wherein, represents a vibration amplitude, represents an exposure time, represents a vibration frequency, represents a pixel size, if if equal to 0, it is judged that the defect detection image appears imaging smear, if not, it is judged that the defect detection image does not appear imaging smear; The vibration combined early warning unit is used for early warning whether the industrial part surface defect detection is abnormal combined with the vibration parameters generated by the real-time movement, and the early warning method is as follows: The abnormal early warning value of surface defect detection of industrial parts caused by vibration parameters generated by current motion is calculated The calculation formula is as follows: ； Wherein, if the abnormal early warning value is greater than or equal to 1, it indicates that the industrial part surface defect detection is abnormal, indicates that imaging trailing caused by motion blur occurs during the industrial part surface defect detection, reports the system, and issues a detection abnormality early warning reminder, if not, it indicates that the industrial part surface defect detection is normal, indicates that imaging trailing caused by motion blur does not occur during the industrial part surface defect detection, indicates the imaging trailing value that occurs during the part surface defect detection at the current time, indicates the maximum allowed imaging trailing, indicates the vibration signal-to-noise ratio drop value caused by vibration, indicates the maximum allowed vibration signal-to-noise loss, indicates the root mean square value of the vibration acceleration, indicates the safety vibration acceleration threshold, , and indicates the weight coefficient of each sub-item, and is adjusted in real time according to the detection scene.

2. The method of claim 1, wherein: The multi-angle irradiation adjustment end comprises a part angle adjustment unit, an angle deviation calculation unit, an angle abnormality correction unit and a multi-light source fusion unit; The part angle adjustment unit is used for adjusting the part surface defect detection angle in real time through a driving device, and monitoring the adjustment angle in real time through a camera.

3. The method of claim 2, wherein: The angle offset calculation unit is configured to calculate the offset value of the detection angle of the surface defects of the industrial part in real time The calculation formula is as follows: ； wherein, represents the refraction angle under the adjustment angle, represents the refraction angle under the detection angle, represents the included angle between the detection angle and the adjustment angle, if is greater than 1°, it is judged that the detection angle of the surface defect of the industrial part deviates, if not, it is judged that the detection angle of the surface defect of the industrial part does not deviate. The angle anomaly correction unit is used for correcting the detection angle of the surface defects of the industrial parts in real time according to the offset value until the offset value is less than 1°.

4. The method of claim 3, wherein: The fusion multi-light source unit is used for realizing multi-light source fusion on the multi-light source irradiation equipment, and the fusion formula is as follows: ； wherein, represents subtracting the static background from the current frame of the light source, retaining the illuminated area of the change of the detection target, represents the detection target mask, represents the detection area under the illumination of the light source, represents the original input image under the light source, represents the clean illuminated area obtained after the light source fusion of the multiple light source illumination devices.

Images