CN114596301A - Coating roller surface defect detection system adopting 3D line laser profile technology - Google Patents

Abstract

The invention provides a coating roller surface defect detection system adopting a 3D line laser profile technology, which comprises a 3D imaging unit and a processing unit; the 3D imaging unit comprises a 3D line laser profile image sensor and is used for periodically scanning the coating roller in line motion at a preset period T, acquiring roller surface image data and sending the roller surface image data to the processing unit; the processing unit is used for: preprocessing a roller surface image; and intelligently analyzing the roll surface image data by using an image processing algorithm so as to identify the roll surface defects and judge the defect grade. In the application, the detection mode is real-time online, continuous, non-contact, highly automatic and intelligent detection and identification, the abnormal state of the coating roller surface can be clearly imaged, meanwhile, the required 3D height information is acquired based on profile data, meanwhile, the data transmission is strong in real-time performance and low in time delay, and the on-site high-precision quality control requirement is met.

Description

Coating roller surface defect detection system adopting 3D line laser profile technology

Technical Field

The invention belongs to the field of defect detection, and particularly relates to a coating roller surface defect detection system adopting a 3D line laser profile technology.

Background

The existing single-set roll coater generally comprises 2 sets of coating rolls, and when the surface of a high-end color coated strip steel is coated, once the defects of meat falling, peeling and the like are generated on the surface of the coating roll, various surface defects such as abnormal coating, coating loss, coating unevenness and the like of the strip steel surface can be caused.

Under the requirement of wisdom manufacturing and labor efficiency promotion, the automation rate of producing the line is improving by a wide margin, and on-the-spot personnel often can't carry out effectual real-time status monitoring to on-the-spot coating roller region because operation centralized control, do not have specific detection device more and do the special item to the coating roller surface and detect. When a manual or post-process detection device finds that the surface of the strip steel has defects caused by coating roller surface abnormity, more batch defects are generated, and slight defects are more in detection missing risk, so that blind areas for judging whether the surface of the coating roller is normal at present are further caused, and great challenges are brought to quality control of users. Therefore, the defects of the strip steel caused by the abnormal surface of the coating roller continuously affect the field product quality control.

The existing defect monitoring technology cannot acquire depth information of defects, and meanwhile, the existing defect detection is easily interfered by the film coating state of the surface of the coating roller, so that the existing detection technology is often limited obviously.

Disclosure of Invention

Technical problem to be solved

The embodiment of the application provides a coating roller surface defect detection system adopting a 3D line laser profile technology, which solves the problem of fluctuation of the coating roller surface state existing on site, eliminates the surface defect of strip steel caused by roller surface abnormality, can avoid the generation of defects from a source end, and improves the quality control level.

(II) technical scheme

In a first aspect, an embodiment of the present application provides a system for detecting defects on a surface of a coating roller by using a 3D line laser profile technology, including: a 3D imaging unit and a processing unit;

the 3D imaging unit comprises a 3D line laser profile image sensor and is used for periodically scanning a coating roller in line motion at a preset period T, acquiring roller surface image data and sending the roller surface image data to the processing unit;

the processing unit is configured to: preprocessing the roller surface image; and intelligently analyzing the roll surface image data by using an image processing algorithm so as to identify the roll surface defects and judge the defect grade.

Wherein the 3D imaging unit is configured to:

scanning a coating roller in online motion at regular intervals with a preset period T by using a 3D line laser profile image sensor to obtain 3D profile data and a 2D gray level image;

the processing unit is configured to:

and preferentially processing and analyzing the 3D contour data to identify the defects, and when the defects are identified, synchronously acquiring and analyzing the 2D gray level image, and performing image segmentation, feature extraction and target detection to identify the defects.

Wherein the processing unit comprises a defect curved surface segmentation module, and the defect curved surface segmentation module is used for: outputting three-dimensional point cloud data in a multi-row outline form finally, detecting the inflection point of each outline, and solving the average value of the inflection point row index values of the corresponding positions detected by the plurality of outlines to obtain a tangent plane for dividing the three-dimensional plane;

the inflection point is determined from the profile data:

for each contour, the horizontal and vertical coordinates of each point on the contour can be obtained, for the point I, the average value of the 5 points before the index I and the 5 points after the index I is obtained as a statistical result, the longitudinal gradient of the point I is calculated, and the first 4 points with the maximum gradient change of the vertical coordinates of each contour are obtained as the inflection points of the contour.

The processing unit comprises a post-segmentation curved surface defect detection module, and the post-segmentation curved surface defect detection module is used for:

fitting each curved surface by a least square method;

the fitted plane is compared to the actual segmentation plane data points to detect defective points.

Wherein, the detection module for the defect of the divided curved surface is used for:

initializing a coefficient matrix and a result matrix;

judging whether the coefficient matrix is a singular matrix;

solving an inverse matrix, and solving A, B, C, D coefficients in the plane equation Ax + By + Cz + D = 0 in sequence after fitting.

Wherein, the detection module for the defect of the divided curved surface is used for:

each block is subjected to block fitting, the divided blocks are subjected to fitting according to 100 × 100 blocks, the area blocks of the sampling points 100 × 100 are subjected to surface fitting, and the sampling point areas less than 100 are subjected to fitting according to the result of the remainder of 100, so that the three-dimensional plane formed by each block is subjected to fitting;

calculating the distance between the fitted plane and the actual sampling point, discarding the first 20% of data points with the largest distance, continuing plane fitting on the basis of the rest data points, finally, enabling the rest data points to be located on the same plane as much as possible through multiple iterations, having an iterative optimization process, finally obtaining the plane fitting equation Ax + By + Cz + D = 0 of the pixel block sampling point, and finally screening out possible defect points By using the fitting equation.

The 3D line laser profile sensor adopts a line laser imaging technology of a triangulation principle, an optimized optical design and blue or red laser.

Wherein, still include alarm unit, be used for: and when the defects of the preset level are detected, sending alarm information, wherein the alarm mode comprises audible and visual alarm or picture alarm.

The processing unit is also used for carrying out preprocessing of image reconstruction, image transformation, image enhancement, restoration and correction on the roller surface image.

The system further comprises a terminal, wherein the terminal is used for displaying production information and latest defect detection data, and various display diagrams including incoming material information, defect grades and 3D information are included.

(III) advantageous effects

The coating roller surface defect detection system adopting the 3D line laser profile technology has the following beneficial effects:

the system of the present application comprises a 3D imaging unit and a processing unit; the 3D imaging unit comprises a 3D line laser profile image sensor and is used for periodically scanning the coating roller in line motion at a preset period T, acquiring roller surface image data and sending the roller surface image data to the processing unit; the processing unit is used for: preprocessing a roller surface image; and intelligently analyzing the roll surface image data by using an image processing algorithm so as to identify the roll surface defects and judge the defect grade. In the application, the detection mode is real-time online, continuous, non-contact, highly automatic and intelligent detection and identification, the abnormal state of the coating roller surface can be clearly imaged, meanwhile, the required 3D height information is acquired based on profile data, meanwhile, the data transmission is strong in real-time performance and low in time delay, and the on-site high-precision quality control requirement is met.

Drawings

FIG. 1 is a schematic structural diagram of a coating roller surface defect detection system using a 3D line laser profile technology according to an embodiment of the present application;

FIG. 2 is a schematic view of inflection points of a contour in the present application;

fig. 3 is a schematic diagram of a block-fitting of a curved surface formed by (280 × 360) sampling points in the present application.

Detailed Description

The present application is further described with reference to the following figures and examples.

In the following description, the terms "first" and "second" are used for descriptive purposes only and are not intended to indicate or imply relative importance. The following description provides embodiments of the invention, which may be combined or substituted for various embodiments, and this application is therefore intended to cover all possible combinations of the same and/or different embodiments described. Thus, if one embodiment includes feature A, B, C and another embodiment includes feature B, D, then this application should also be construed to include embodiments that include A, B, C, D in all other possible combinations, even though such embodiments may not be explicitly recited in the text that follows.

The following description provides examples, and does not limit the scope, applicability, or examples set forth in the claims. Changes may be made in the function and arrangement of elements described without departing from the scope of the disclosure. Various examples may omit, substitute, or add various procedures or components as appropriate. For example, the described methods may be performed in an order different than the order described, and various steps may be added, omitted, or combined. Furthermore, features described with respect to some examples may be combined into other examples.

The existing single-set roll coater generally comprises 2 sets of coating rolls, and when the surface of a high-end color coated strip steel is coated, once the defects of meat falling, peeling and the like are generated on the surface of the coating roll, various surface defects such as abnormal coating, coating loss, coating unevenness and the like of the strip steel surface can be caused. The depth information of the defects can not be acquired by the conventional 2D defect monitoring technology, and meanwhile, the defect detection of the 2D camera is easily interfered due to the film coating state on the surface of the coating roller, so that the 2D defect monitoring technology has obvious limitation.

The 3D vision technique differs from the 2D vision technique as follows: 1. XYZ data are obtained by 3D vision at the same time, and XY data can only be obtained by 2D vision; 2. 3D has no requirement on whether the measured object has enough contrast, while 2D generally needs higher contrast; 3. the 3D vision technology is generally high in integration level and has higher stability and repeatability.

As shown in fig. 1, the system for detecting defects on the surface of a coating roller by using 3D line laser profile technology comprises: a 3D imaging unit 201 and a processing unit 202; the 3D imaging unit 201 includes a 3D line laser profile image sensor, and is configured to periodically scan the coating roller in line motion at a preset period T, acquire roller surface image data, and send the roller surface image data to the processing unit; the processing unit 202 is configured to: preprocessing a roller surface image; and intelligently analyzing the roll surface image data by using an image processing algorithm so as to identify the roll surface defects and judge the defect grade.

In some embodiments, the 3D imaging unit 201 is configured to: scanning a coating roller in online motion at regular intervals with a preset period T by using a 3D line laser profile image sensor to obtain 3D profile data and a 2D gray level image; the processing unit 202 is configured to: and preferentially processing and analyzing the 3D contour data to identify the defects, and when the defects are identified, synchronously acquiring and analyzing the 2D gray level image, and performing image segmentation, feature extraction and target detection to identify the defects.

The processing unit 202 is configured to execute a defect detection algorithm, where the defect detection algorithm in this application is as follows:

1. image processing method

1) Image transformation: (spatial and frequency domain, geometric transformation, chrominance transformation, scale transformation);

2) image enhancement: grayscale transformation enhancement (linear grayscale transformation, piecewise linear grayscale transformation, nonlinear grayscale transformation);

2. histogram enhancement (grey level histogram, histogram equalization);

3. image smoothing/denoising (neighborhood averaging, weighted averaging, median filtering, nonlinear mean filtering, gaussian filtering, bilateral filtering); image (edge) sharpening: gradient sharpening, Roberts operator, Laplace operator, Sobel operator, etc.;

4. image segmentation: the image segmentation is to extract a meaningful characteristic part in the image, wherein the meaningful characteristic is an edge, a region and the like in the image, and the meaningful characteristic is a basis for further image recognition, analysis and understanding.

(1) Threshold segmentation (fixed threshold segmentation, optimal/OTSU threshold segmentation, adaptive threshold segmentation);

(2) based on boundary segmentation (Canny edge detection, contour extraction, boundary tracking);

(3) hough transform (Hough transform straight line detection, Hough transform circle detection);

(4) based on region segmentation (region growing, region merging and splitting, cluster segmentation);

(5) color segmentation;

(6) dividing watershed;

for the sample to be detected, because the brightness information of each pixel point of the sample can be obtained, the 20% pixel points with the lowest brightness can be obtained by a histogram method, the defect area is preliminarily screened in one step, and the effect is general after the operation is actually performed, and the actually used detection algorithm is as follows.

The detection algorithm actually used comprises the following steps:

and dividing the collected sample piece into 5 three-dimensional planes according to the characteristics of the collected sample piece, and respectively carrying out defect detection. The main process comprises two processes of defect curved surface segmentation and surface defect detection after segmentation.

The process of defect surface segmentation is as follows: as shown in fig. 2, each defect curved surface is segmented, that is, three-dimensional point cloud data is finally output in the form of a plurality of lines of contours, the inflection point of each contour is detected, and then the inflection point row index values (for example, the number 1 inflection point in fig. 2) of the corresponding positions detected by the plurality of contours are averaged to obtain a tangent plane (marked by an arrow in fig. 2) for better segmenting the three-dimensional plane.

Method for determining inflection points from profile data: for each contour, the horizontal and vertical coordinates of each point on the contour can be obtained (plus the contour is located on an XOY two-dimensional plane), in order to ensure that the calculated inflection points are relatively accurate, the method of the application is that for a point I, the average value of 5 points before and 5 points after the index I is taken as a statistical result, the longitudinal gradient (the gradient change size) of the point is calculated, the gradient change size of the point can be reflected through the vertical coordinate change because the resolution of the data acquired by a camera in the x or y direction is fixed, and the first 4 points with the maximum gradient change of the vertical coordinate of each contour are obtained through the method, namely the inflection points of the contour.

The detection process of the defects of the curved surfaces of the blocks after the division is as follows: according to the method, a defective workpiece can be divided into a plurality of curved surfaces, then a curved surface defect detection process is mainly introduced, each curved surface is mainly fitted through a least square method, and defect points are detected by comparing a fitting plane with data points of an actual dividing plane, and the method can be divided into the following steps:

1) least square method surface fitting plane

Fitting a plane through discrete points, that is, to find a plane (z = ax + by + c), the "distance" of this plane to each point is made to be the closest, according to the least squares method,

that is, a set of a, b, c is required so that the value of S is minimized for the existing discrete points. Finally, the problem translates into 1. first, the initialA coefficient matrix and a result matrix are quantized. 2. And judging whether the coefficient matrix is a singular matrix. 3. And (5) inverting and calculating a result. By introducing an external library Eigen, the process of solving the inverse matrix can be conveniently realized, and four coefficients of A, B, C and D in a plane equation Ax + By + Cz + D = 0 are solved in sequence after fitting.

2) The result can be obtained by directly fitting the whole plane by using the least square method, but the plane fitting effect is natural and general due to the fact that the data points are too large (the row pixel points are larger than 1900 and the column pixel points are larger than 1200 in the defect area). The optimization method is as follows:

1. each block is subjected to block fitting, 5 divided regions are subjected to fitting according to 100 × 100 regions in specific operation, as shown in fig. 3, the region blocks of sampling points 100 × 100 are subjected to surface fitting, and sampling point regions less than 100 are subjected to fitting according to the result of the remainder of 100, so that the three-dimensional plane formed by each block is respectively subjected to fitting. The numbers in fig. 3 represent the number of pixels/samples corresponding to the side length.

2. Calculating the distance between the fitted plane and the actual sampling point, discarding the first 20% of data points with the largest distance (error), continuing plane fitting on the basis of the rest data points, finally, enabling the rest data points to be positioned on the same plane as much as possible through multiple iterations, having an iterative optimization process, finally obtaining a plane fitting equation Ax + By + Cz + D = 0 of the pixel block sampling point, and finally screening out possible defect points By using the fitting equation. (bumps or pits will be far from the plane and thus this type of defect can be detected very well).

In the application, the system continuously scans the surface of the coating roller through an imaging system consisting of a high-speed 3D sensor and an LED light source to form 3D contour data and a high-definition surface image, the 3D point cloud defect data meeting the depth detection requirement are transmitted to an image processing unit through a special cable, the next step of filtering is carried out by combining a filtering algorithm and a post-processing rule set, and information including 3D defect types, depth, position, size and the like is presented through an HMI (human machine interface).

According to the field environment, a plurality of 3D cameras are arranged, and the three-dimensional data of the coating surface is acquired by synchronously triggering once. The latest 3D sensor with the frequency of more than 10kHz + is selected, the sampling frequency is higher, the resolution is designed with higher precision, and the detection of the defect of interest in each direction of XYZ axes under the imaging parameters is met; 3D defect identification and 2D defect matching defect flaw detection based on point cloud output, and related software optimization design and alarm functions. Meanwhile, specific defect identification and classification processing are carried out based on algorithm development experience of a wire rod and other long-material surface detection system. The optimized equipment electromechanical design, the box body has better tightness, and the push-pull type detection box body is convenient to maintain and has a reliable cooling mode. The system is characterized in that the improvement design of a conveying area is actually carried out according to a client site, the material conveying is stabilized, the detection effect and the environmental protection are ensured, and a communication interface is reserved in the system and used for uploading data such as image data, defect information and grade information to a user quality management platform for subsequent data mining application.

Through the imaging system, the detection device can acquire 3D point cloud data meeting the minimum defect detection and a 2D defect image with high contrast in an optimal resolution, an optical path and an imaging mode, and ensures that a user pays attention to defects, especially imaging data with the same depth and width ratio as shallow and flat pit-shaped defects are sensitive enough. Meanwhile, the method can overcome the diversity and complexity of the surface and achieve stronger identification robustness. The whole imaging system mainly comprises main equipment such as a high-speed 3D laser sensor. The degree of defect detection is determined by the accuracy of imaging design, and the number of cameras and the light path angle which are suitable for different application occasions, defect appearances and specifications of equipment to be detected are designed.

The 3D line laser profile sensor adopts a line laser imaging technology of a triangulation principle, an optimized optical design and blue or red laser, so that the sensor can obtain clearer, high-repeatability and reliable 3D data on a scanning gloss surface and a low-contrast object surface, and is suitable for the surface state of a metal detection object.

The software part of the 3D online detection system consists of a plurality of modules, which are the core of the whole detection system and respectively comprise a DPU module, a detection service module, a data storage module and a terminal HMI, and the whole communication architecture is based on ICE.

The DPU module mainly has the functions of acquiring point cloud data from a 3D sensor, then carrying out defect detection on an image, after the detection is finished, feeding back a detection result to a detection server by the DPU, caching the image in a local memory, waiting for the notification of the detection server to store the image in a hard disk or discard the image, providing a file uploading service by the DPU, transmitting an image file of an external request to a destination address, simultaneously providing a timing state diagnosis function by the DPU, and regularly reporting the utilization rate of a CPU, the memory and the hard disk of the DPU to the detection server.

The main functions of the detection service module are divided into two parts, namely a signal monitoring part and a defect processing part. The defect processing part is mainly responsible for synchronizing the defect feedback results of each DPU, informing the DPUs of giving up the graph or uploading the defect information of the graph to the detection server, storing the uploaded defect information into a database on the detection server, and informing the terminal of the defect information; at the same time, the detection server monitors the status of each DPU.

The main function of the data storage module is to transmit the cached images in the DPU to the archive server according to the policy of the detection server or manually by the user. Meanwhile, the local machine runs a file acquisition service, allows the external part to acquire historical defect images under a certain path of the local machine in a safe connection mode, and simultaneously converts historical images before a certain period of time into lossy compression for storage if the format of the historical images is lossless compression, thereby saving the space of a hard disk.

The HMI of the terminal mainly has the functions of displaying production information and latest defect detection data, including various display graphs of incoming material information, defect grades and 3D information, and can view depth information, defect distribution, types and the like of concerned defects based on the HMI. The client can also recall historical data or export data based on the HMI. The terminal can also display 3D defect information and a corresponding 2D defect map.

The image processing algorithm in this application has taken into account the degree of depth detection demand of 3D defect and 2D image detection's degree of depth study classification mode simultaneously, the system adopts the mode of priority processing 3D defect detection to carry out defect identification, 3D defect synchronous acquisition 2D grey level image data after to the discernment, carry out specific processing, the defect secondary classification and the filtration strategy of having set for a plurality of logics and dimension for effectual filtration pseudo-defect, 3D + 2D's defect show mode simultaneously, be convenient for the customer to observe and verify actual defect. A double-layer processing framework combining 3D detection and 2D detection can adapt to detection requirements under high-speed production, point cloud data output in the shortest time is guaranteed, meanwhile, each frame of image completes image acquisition, edge detection and target detection, the image containing suspected defects is stored in an image cache, otherwise, the image is deleted, and then the suspected target images in the cache queue are further processed, wherein the processing comprises defect segmentation, image storage, feature extraction, classification and image post-processing.

According to the method and the system, the alarm rule can be customized through a standard configuration interface, when the system detects a set alarm defect, an automatic alarm (sound-light alarm or picture alarm) can be given, and the I/O output switch quantity can be selected as a production brake signal.

In the embodiment of the present application, the camera model: LMI-Gotat series cameras. Data form collected by the camera: 1. and according to the set sampling frequency and the resolutions in the X, Y and Z directions, carrying out data transmission on the coordinates of each point in the three-dimensional space of the scanned area in a mode of firstly carrying out row by row and then carrying out column by column to obtain the three-dimensional coordinates of any sampling point. (point cloud mode) 2. after the laser line is swept for a certain distance, the contour data of the measured object along the laser line plane is derived (contour mode), and the brightness value of the acquired data point can be obtained in both modes. The collection process comprises the following steps: aiming at the possible defects of the sample, the sample is rotated by using the operating platform, the laser line of the camera is scanned along the direction parallel to the axis of the sample, the section condition of the scanned sample along the axis is obtained, and a three-dimensional plane is obtained after the scanning process is finished.

The electric system of this application integration corresponding light source, imaging device, three electric equipment controlgear. In order to ensure seamless connection of various devices and complete set functions, special interface devices are selected for electrical integration to ensure stability and reliability of long-distance, high-speed and mass image data transmission. Special cables, optical cables and network connecting equipment are selected to ensure data connection between image acquisition equipment such as cameras and the like and the surface detection assembly, and ensure data communication connection between the surface detection assembly and the universal server.

The application's mechanical mechanism mainly plays the thermal-insulated protection of dust removal, bears all kinds of core and detects the original paper, possesses the accurate adjustable in mechanical device simultaneously. The mechanical mechanism must have sufficient strength, heat insulation, dust prevention, shock resistance and internal temperature monitoring functions.

Designing a camera adjusting mechanism: in order to ensure that the imaging position of the 3D sensor is consistent with the optical design, the imaging angle can be adjusted to be a rotation angle or a pitching angle under a set angle, the adjustable range is +/-10 degrees, the rotation angle is +/-180 degrees, the adjustment angle is linear change, and any angle can be adjusted according to the actual requirements of a site. The camera is rotatable along the optical axis. The multi-degree-of-freedom precise adjustment of the camera is realized; according to the adjustment requirements of multiple degrees of freedom, each degree of freedom is provided with an independent worm and gear adjusting mechanism and a locking device, so that the device is suitable for the field vibration environment.

Purging and cooling modes: the stable work of camera, light source, electrical system part in the assurance sensor detection box, the system has designed the refrigeration machine and cools off to the components and parts of box inside, ensures the stable work of system. Meanwhile, the multi-level filtering axial flow blower and the specific opening angle of the system are designed, so that the blowing of the detection window is realized, the dust in the production process is prevented from being accumulated on the detection window, and the convenience of maintenance is improved.

This application is based on 3D line laser Profile image sensor, with the non-contact mode, the dynamic point cloud data that acquires roll surface Profile (Profile) in real time, adopt customized algorithm to handle image data with the roll surface defect that produces specially in the work to the coating roller, acquire the surface feature and the depth information of target image, carry out AI deep learning intelligent analysis, discernment, defect depth measurement, make final defect judgement or hierarchical judgement, accomplish real-time online roll surface state detection, in time report to the police or linkage field control system, and then stop the production of defect from the source. The method solves the problem of fluctuation of the state of the coating roller surface on site, eliminates the surface defects of the strip steel caused by abnormal roller surface, can stop the defects from the source end, and improves the quality control level.

The electromechanical design and cooling system, the box body airtightness and protection are suitable for the severe environment of the steel production field, and the stable work and the quick maintenance operation of the equipment are ensured. The system reserved communication interface is used for uploading data such as image data, defect information, grade information and the like to a user quality management platform for subsequent data mining application.

In the application, the detection mode is real-time online, continuous, non-contact, highly automatic and intelligent detection and identification, the abnormal state of the coating roller surface can be clearly imaged, meanwhile, the required 3D height information is acquired based on profile data, meanwhile, the data transmission is strong in real-time performance and low in time delay, and the on-site high-precision quality control requirement is met.

All functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated into one unit; the integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional unit.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A coating roller surface defect detection system adopting a 3D line laser contour technology is characterized by comprising a 3D imaging unit and a processing unit;

the 3D imaging unit comprises a 3D line laser profile image sensor and is used for periodically scanning a coating roller in line motion at a preset period T, acquiring roller surface image data and sending the roller surface image data to the processing unit;

the processing unit is configured to: preprocessing the roller surface image; and intelligently analyzing the roll surface image data by using an image processing algorithm so as to identify the roll surface defects and judge the defect grade.

2. The system of claim 1, wherein the 3D imaging unit is configured to:

scanning a coating roller in online motion at regular intervals with a preset period T by using a 3D line laser profile image sensor to obtain 3D profile data and a 2D gray level image;

the processing unit is configured to:

and preferentially processing and analyzing the 3D contour data to identify the defects, and when the defects are identified, synchronously acquiring and analyzing the 2D gray level image, and performing image segmentation, feature extraction and target detection to identify the defects.

3. The system of claim 2, wherein the processing unit comprises a defect surface segmentation module configured to: outputting three-dimensional point cloud data in a multi-row outline form finally, detecting the inflection point of each outline, and solving the average value of the inflection point row index values of the corresponding positions detected by the plurality of outlines to obtain a tangent plane for dividing the three-dimensional plane;

the inflection point is determined from the profile data:

for each contour, the horizontal and vertical coordinates of each point on the contour can be obtained, for the point I, the average value of the 5 points before the index I and the 5 points after the index I is obtained as a statistical result, the longitudinal gradient of the point I is calculated, and the first 4 points with the maximum gradient change of the vertical coordinates of each contour are obtained as the inflection points of the contour.

4. The system of any one of claims 1-3, wherein the processing unit comprises a post-segmentation curved surface defect detection module configured to:

fitting each curved surface by a least square method;

the fitted plane is compared to the actual segmentation plane data points to detect defective points.

5. The system of claim 4, wherein the post-segmentation curved surface defect detection module is configured to:

initializing a coefficient matrix and a result matrix;

judging whether the coefficient matrix is a singular matrix;

solving an inverse matrix, and solving A, B, C, D coefficients in the plane equation Ax + By + Cz + D = 0 in sequence after fitting.

6. The system of claim 4, wherein the post-segmentation curved surface defect detection module is configured to:

each block is subjected to block fitting, the divided blocks are subjected to fitting according to 100 × 100 blocks, the area blocks of the sampling points 100 × 100 are subjected to surface fitting, and the sampling point areas less than 100 are subjected to fitting according to the result of the remainder of 100, so that the three-dimensional plane formed by each block is subjected to fitting;

calculating the distance between the fitted plane and the actual sampling point, discarding the first 20% of data points with the largest distance, continuing plane fitting on the basis of the rest data points, finally, enabling the rest data points to be located on the same plane as much as possible through multiple iterations, having an iterative optimization process, finally obtaining the plane fitting equation Ax + By + Cz + D = 0 of the pixel block sampling point, and finally screening out possible defect points By using the fitting equation.

7. The coating roller surface defect detection system adopting 3D line laser profile technology according to any one of claims 1-3, characterized in that the 3D line laser profile sensor adopts line laser imaging technology of triangulation principle, optimized optical design and blue or red laser.

8. The coating roller surface defect detection system adopting the 3D line laser profile technology according to any one of claims 1-3, characterized by further comprising an alarm unit for: when the defect of the preset level is detected, alarm information is sent out, and the alarm mode comprises audible and visual alarm or picture alarm.

9. The system for detecting defects of a coating roller surface by using the 3D line laser contour technology is characterized in that the processing unit is also used for preprocessing image reconstruction, image transformation, image enhancement, restoration and correction of the roller surface image.

10. The coating roller surface defect detection system adopting the 3D line laser contour technology is characterized by further comprising a terminal, wherein the terminal is used for displaying production information and latest defect detection data, and various display graphs comprising material incoming information, defect grade and 3D information are included.

Images