A method and device for detecting defects of rotating body based on linear array tracking focus

Abstract

The present invention relates to a method and device for detecting defects of a rotating body based on linear array focus tracking, the method comprising: mounting a rotating body to be tested on a controllable turntable, driving the rotating body to rotate to different measuring positions by the controllable turntable, adjusting the spatial posture of the linear array imaging module according to the relative posture information between the linear array imaging module and the current measured area of ​​the rotating body at different measuring positions, so that the current measured area of ​​the rotating body is completely within the focal depth range of the linear array imaging module, and then obtaining the texture information and depth information of the current measured area of ​​the rotating body surface; data fusion is performed on the relative posture information, depth information and texture information obtained at each measuring position, so as to extract the defect information and three-dimensional morphology information of the rotating body surface. Compared with the prior art, the present invention can perform real-time focus tracking during the rotation of the rotating body, ensuring that the imaging module can clearly image under a lens group with high magnification and small depth of field.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial defect detection, and in particular to a method and device for detecting defects of a rotating body based on linear array focus tracking. Background Art

[0002] Cylindrical and conical components are very common in industrial scenarios, such as bearings, commutators, pipes, ingots, etc. Scratches, cracks, pitting, etc. on the surface of these workpieces will affect their working performance and even directly cause the failure of the components. The defect detection system can quickly complete quality assessment, improve the yield rate, and improve the process in a targeted manner to improve production efficiency. Usually the size of the defect is hundreds of microns, and the width of a small scratch may be only about 20um. Therefore, the detection system needs to have a high enough spatial resolution and a sufficient sampling speed to meet the rapid detection of industrial sites.

[0003] The most commonly used traditional shape-based defect detection technology is the line laser measuring instrument. The measuring instrument projects a linear laser onto the undulating object being measured. The modulation of the object's shape causes the line laser to bend accordingly. After the detection camera obtains the curve, the height value of each point on the object corresponding to this linear laser can be obtained by solving the geometric relationship. Furthermore, the object is scanned in the vertical direction of the linear laser, and the collected data is processed to obtain the three-dimensional information of the object. However, this method has poor recognition ability for fine and shallow defects, and has poor measurement effect on high-reflectivity surfaces such as metals and wafers. In addition, there are problems such as high cost and high installation accuracy requirements. Therefore, it is difficult to be widely used in the field of rotating body defect detection.

[0004] With the rapid development of computer vision technology, using cameras to obtain image information and identify and locate scratches and damage in images has gradually become the mainstream. Use a linear array camera to scan the rotating body for image stitching to obtain the appearance of a complete circle, and use machine vision-related algorithms to achieve defect detection, such as a rotating body part defect detection device disclosed in the invention with publication number CN115901624A. However, the current system is only applicable to standard cylinders. Due to the small depth of field of high-magnification objective lenses and the inclination of the generatrix of special-shaped rotating bodies, the object distances at both ends of the imaging line are different, resulting in serious defocus of the object; on the other hand, the roundness of the rotating body itself and the concentricity of the installation process will also cause changes in the object distance, resulting in defocus, affecting the imaging quality, and then false detection or missed detection of defects. In order to ensure that the system's resolution reaches 10um, a high-magnification lens group is required to detect relatively small defects on the metal surface. Therefore, it is necessary to adjust the spatial position and posture of the visual imaging system in real time to complete the continuous focus tracking of the linear array camera to ensure the availability of the overall system. High resolution will cause the problem of limited depth of field of the objective lens. For example, when the resolution is 7 microns, the actual depth of field of the objective lens is about 50 microns. In addition, the camera can only obtain two-dimensional information such as defect length, width or diameter, and cannot obtain depth data, which is still a long way from restoring the real defect information. Summary of the invention

[0005] The purpose of the present invention is to overcome the defect of the above-mentioned prior art that the imaging module cannot clearly and automatically focus on the entire curved surface at a large magnification, and to provide a rotating body defect detection method and device based on linear array tracking focus.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A method for detecting defects of a rotating body based on linear array focus tracking comprises the following steps:

[0008] The rotating body to be measured is mounted on a controllable turntable, an objective lens is set to collect light beams, a linear array imaging module is set to collect texture information on the surface of the rotating body, a linear array tracking module is set to obtain relative posture information between the linear array imaging module and the rotating body and depth information of the current measured area on the surface of the rotating body, and a posture adjustment module is set to adjust the spatial posture of the linear array tracking module and the linear array imaging module;

[0009] The controllable turntable is used to drive the rotating body to rotate to different measurement positions. At different measurement positions, the relative posture information between the linear array imaging module and the rotating body is obtained according to the linear array tracking module, and the spatial postures of the linear array imaging module and the linear array tracking module are adjusted so that the current measured area of ​​the rotating body is completely within the focal depth range of the linear array imaging module, and then the texture information and depth information of the current measured area on the surface of the rotating body are obtained;

[0010] The relative posture information, depth information and texture information obtained at each measurement position are fused to extract the defect information and three-dimensional morphology information of the rotating body surface.

[0011] Furthermore, the process of the linear array focus tracking module acquiring the relative posture information between the linear array imaging module and the rotating body specifically includes:

[0012] Generate a bias line laser that can identify the defocus size and direction of the object and irradiate it onto the surface of the rotating body;

[0013] The focus tracking camera in the linear array focus tracking module obtains the reflected light spot of the offset line laser on the surface of the rotating body. The distance from each point on the offset line laser to the surface of the rotating body will cause each point corresponding to the reflected light spot to expand horizontally. The expansion size is calculated in pixels and converted into the actual size {d1, d2, …, d n};

[0014] The calculation expression of the actual size of the n points on the current measured area of ​​the rotating body is:

[0015] d i =P Z C xi ,i=1,2,…,n

[0016] Where i is the number of the point in the current measured area of ​​the rotating body, P Z is the pixel size of the tracking camera in the linear array tracking module, C xi is the coordinate of the edge of the reflected light spot corresponding to the i-th point in the measured area in the imaging picture of the tracking camera;

[0017] The calculation formula for the defocus amount of each point in the current measured area of ​​the rotating body is:

[0018]

[0019] Where i is the number of the point in the current measured area of ​​the rotating body, δ i is the defocus value corresponding to the i-th point in the current measured area of ​​the rotating body, f1 is the focal length of the objective lens, f2 is the focal length of the focusing lens of the linear array tracking module, and d i is the extended size of the reflected light spot corresponding to the i-th point in the current measured area of ​​the rotating body, and a is the initial width of the bias line laser provided by the line bias module;

[0020] Finally, according to the actual size of the expansion, the defocus value of each point in the current measured area of ​​the rotating body is calculated, and the coordinate transformation matrix is ​​used to calculate the relative posture information between the linear array imaging module and the current measured area of ​​the rotating body.

[0021] Furthermore, the process of adjusting the spatial posture of the linear array imaging module and the linear array focus tracking module is specifically as follows:

[0022] According to the relative posture information, a six-axis motion system is used to perform three-dimensional translation and pitch, yaw and roll angle adjustment on the linear array imaging module, so that the focal plane of the objective lens is parallel to the best fitting plane of the rotating body surface, and the linear array imaging module is located at the clearest imaging position.

[0023] Furthermore, the process of acquiring the depth information is specifically as follows:

[0024] The reflected light spot of the biased line laser on the surface of the rotating body is obtained through the linear array tracking module. According to the linear relationship between the depth and the spot size, the undulation data of the surface of the rotating body is obtained with the focal plane of the objective lens as the original plane. In combination with the relative posture information between the turntable rotation axis of the rotating body and the linear array tracking module, the three-dimensional morphology data of the surface of the rotating body is obtained as the depth information.

[0025] Furthermore, the position of the objective lens is moved axially, and the spatial postures of the linear array imaging module and the linear array focus tracking module are adjusted by changing the position of the objective lens, so that the linear array imaging module is located at the clearest imaging position.

[0026] Furthermore, the process of fusing the relative pose information, depth information and texture information includes:

[0027] For a regular rotating body, the posture information of each measurement position is converted into spatial coordinates, the depth information is superimposed on the spatial coordinates of the corresponding measurement position, and data fitting and error correction processing are performed to obtain a high-precision point cloud; the pixel points of the texture information of each measurement position are matched to the high-precision point cloud to form a multimodal point cloud with texture information;

[0028] For irregular rotating bodies, the pose information, texture information and depth information of each measured position are sent to the pre-trained neural network for coordinate matching, depth superposition and texture assignment to obtain a multimodal point cloud with texture information.

[0029] Furthermore, the method further includes processing the multimodal data of the surface of the rotating body obtained after data fusion, so as to extract surface defect information;

[0030] The multimodal data processing process includes:

[0031] Input the point cloud data containing spatial coordinates and color texture in the multimodal data into the pre-trained neural network to classify the defect type;

[0032] The point cloud is clustered according to the pose information and defect type information, and the points belonging to the same defect type are divided into a point cloud set;

[0033] According to the point cloud clustering results, the three-dimensional model of the rotating body is reconstructed, and defect information is marked on the three-dimensional model. The defect information includes defect type, defect area, defect center coordinates and defect circumscribed rectangle.

[0034] The present invention also provides a rotating body defect detection device for realizing the rotating body defect detection method based on linear array focus tracking as described above, which is used for defect detection of a rotating body, and comprises a controllable turntable, a linear array focus tracking module, a beam splitter, an objective lens, a dichroic mirror, a calculation control unit, a posture adjustment module, an illumination light source and a linear array imaging module, wherein the linear array imaging module comprises a tube lens and a linear array camera, and the linear array focus tracking module comprises a fixed-focus light source, a linear bias module and a focusing lens, and a focus tracking camera; the linear array focus tracking module, the beam splitter, the objective lens, the dichroic mirror, the illumination light source and the linear array imaging module are all mounted on the posture adjustment module; the rotating body is mounted on the controllable turntable;

[0035] The laser emitted by the fixed-focus light source passes through the line bias module to form a bias line laser, which is reflected by the beam splitter into the objective lens and converged on the surface of the rotating body to form a fixed-focus line laser; the objective lens collects the light reflected by the fixed-focus line laser from the rotating body, propagates to the dichroic mirror, and then reflects through the focusing lens to enter the tracking camera; the calculation control unit calculates the defocus amount and spatial posture according to the captured fixed-focus laser, and drives the posture adjustment module to perform real-time focusing;

[0036] The illumination light source emits visible light so that the surface of the rotating body is evenly illuminated. The reflected light generated by the illumination of the rotating body is transmitted through the dichroic mirror into the cylindrical mirror and irradiated into the linear array imaging module, and the linear array imaging module collects the texture information of the surface of the rotating body.

[0037] Furthermore, the line bias module includes a collimating lens, a biased rectangular aperture and a cylindrical lens. The laser emitted by the fixed-focus light source becomes parallel light after passing through the collimating lens. The parallel light blocks half of the light through the biased rectangular aperture to form a biased laser. The biased laser is focused in one direction through the cylindrical lens, and remains parallel light in the other direction, thereby forming a biased line laser.

[0038] Furthermore, the calculation control unit acquires the image of the focus tracking camera in real time and calculates the spatial posture, drives the posture adjustment module to adjust the posture, and forms a closed-loop feedback control system to achieve real-time focus tracking.

[0039] Compared with the prior art, the present invention has the following advantages:

[0040] (1) The real-time linear array focus tracking method proposed in the present invention can detect whether a rotating body is within the focal depth of an imaging module on a line, and is suitable for clear imaging of a rotating body at a high magnification; the posture adjustment module calculates the posture adjustment data in real time according to the result of the focus tracking optical path and completes the adjustment, thereby realizing real-time clear imaging of the linear array camera under different depths of field and complex surface objects; therefore, the optical focus tracking proposed in the present invention is a universal focusing solution for curved objects, which solves the problem that the imaging module cannot automatically focus on the entire curved surface clearly at a large magnification, and can be widely used in curved surface automatic focus tracking imaging.

[0041] (2) The real-time linear array focus tracking module proposed in the present invention cooperates with the posture adjustment module to output high-precision point cloud information of the surface of the object under test through posture data and depth data on the basis of closed-loop control and real-time focus tracking, thereby reflecting the morphology of the object surface.

[0042] (3) The linear array focus tracking module and the linear array imaging module of the present invention work in different bands through dichroic mirrors without interfering with each other, thus ensuring real-time adjustment and real-time measurement, improving the versatility of the method, and making it applicable to measurement needs in more different scenarios and conditions.

[0043] (4) The multimodal point cloud proposed by the present invention integrates 3D shape data and texture information, realizes multi-dimensional measurement of rotating bodies, and provides more references and information sources for defect detection. This advanced data fusion method will further improve the accuracy and reliability of defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a structural schematic diagram of a rotating body defect detection device provided in an embodiment of the present invention;

[0045] Figure 2 A schematic diagram of the composition of a linear array focus tracking module and a linear array imaging module provided in an embodiment of the present invention;

[0046] Figure 3 A schematic flow chart of a method for detecting defects of a rotating body based on linear array focus tracking provided in an embodiment of the present invention;

[0047] Figure 4 This is a principle diagram of the optical path of a linear array focus tracking module provided in an embodiment of the present invention;

[0048] Figure 5 A schematic diagram of the relationship between the defocus condition of a measured area and the shape of a reflected light spot provided in an embodiment of the present invention;

[0049] Figure 6 It is a schematic structural diagram of a rotating body defect detection device for measuring a rotating body with a regular surface and high reflectivity provided in an embodiment of the present invention;

[0050] Figure 7 It is a schematic structural diagram of a rotating body defect detection device for measuring rotating bodies with irregular surfaces and medium and low reflectivity provided in an embodiment of the present invention;

[0051] Figure 8 A first effect comparison diagram of the rotating body defect detection method based on linear array focus tracking of the present invention and the existing method based on linear array focus tracking of the present invention is not used;

[0052] Fig. 9 A second effect comparison diagram of the rotating body defect detection method based on linear array focus tracking of the present invention and the existing method based on linear array focus tracking of the present invention is not used;

[0053] In the figure, 1. fixed-focus light source, 2. line bias module, 3. beam splitter, 4. objective lens, 5. rotating body, 6. dichroic mirror, 7. focus tracking camera, 8. calculation control unit, 9. posture adjustment module, 10. illumination light source, 11. tube lens, 12. linear array camera, 13. focusing lens. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0055] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0057] Example 1

[0058] like Figure 3 As shown, this embodiment provides a method for detecting defects of a rotating body based on linear array focus tracking, comprising the following steps:

[0059] The rotating body to be measured is mounted on a controllable turntable, an objective lens is set to collect light beams, a linear array imaging module is set to collect texture information on the surface of the rotating body, and a linear array tracking module is set to obtain relative posture information between the linear array imaging module and the current measured area of ​​the rotating body;

[0060] The controllable turntable is used to drive the rotating body to rotate to different measurement positions. At different measurement positions, the spatial position of the linear array imaging module is adjusted according to the relative position information between the linear array imaging module and the current measured area of ​​the rotating body, so that the current measured area of ​​the rotating body is completely within the focal depth range of the linear array imaging module, and then the texture information and depth information of the current measured area on the surface of the rotating body are obtained;

[0061] The relative posture information, depth information and texture information obtained at each measurement position are fused to extract the defect information on the surface of the rotating body.

[0062] Equivalently, the above method comprises the following steps:

[0063] S0: Place the rotating body mounted on the controllable turntable at the initial measurement angle; set the objective lens to collect the light beam, set the linear array imaging module to collect the texture information of the surface of the rotating body, set the linear array tracking module to obtain the relative posture information between the linear array imaging module and the current measured area of ​​the rotating body; set the posture adjustment module to adjust the spatial posture of the linear array tracking module and the linear array imaging module;

[0064] S1: Use the linear array tracking module to obtain the relative posture information between the linear array imaging module and the current measured area of ​​the rotating body;

[0065] S2: Based on the acquired relative posture information, the spatial posture of the linear array imaging module is adjusted using the posture adjustment module to ensure that the current measured area on the surface of the rotating body is completely within the focal depth range of the linear array imaging module;

[0066] S3: using a linear array imaging module to obtain texture information of the current measured area on the surface of the rotating body, and using a linear array tracking module to obtain depth information of the current measured area;

[0067] S4: Control the turntable to rotate the rotating body to the next measurement position, and repeat steps S1 to S3 until the measurement of all the regions of interest of the rotating body is completed;

[0068] S5: fusing the relative posture information, depth information and texture information obtained at each measuring position to obtain multimodal data of the surface of the measured rotating body;

[0069] S6: Process the multimodal data of the surface of the rotating body to be measured and accurately extract the surface defect information.

[0070] Specifically, in step S1, the specific steps of the linear array focus tracking module acquiring the relative posture information between the linear array imaging module and the current measured area of ​​the rotating body are:

[0071] S101: generating a bias line laser capable of identifying the defocus size and direction of the object, and irradiating the laser onto the surface of the rotating body;

[0072] S102: The reflected light on the surface of the rotating body is modulated by the defocus direction and size. Due to the linear relationship between the width of the reflected light spot and the defocus amount, the linear array tracking module will collect light spots of different widths;

[0073] like Figure 4 and Figure 5 As shown in the figure, the calculation formula for the defocus amount of each point in the current measured area of ​​the rotating body is:

[0074]

[0075] Where i is the number of the point in the current measured area of ​​the rotating body, δ i is the defocus value corresponding to the i-th point in the current measured area of ​​the rotating body, f1 is the focal length of the objective lens, f2 is the focal length of the focusing lens of the linear array tracking module, and d i is the extended size of the reflected light spot corresponding to the i-th point in the current measured area of ​​the rotating body, and a is the initial width of the bias line laser provided by the line bias module.

[0076] S103: The linear array tracking module calculates the defocus value of each point in the current measured area of ​​the rotating body according to the spot size of the reflected biased line laser on the surface of the rotating body, thereby calculating the relative posture information between the linear array imaging module and the current measured area of ​​the rotating body, including the x, y, and z axis distances and the pitch, yaw, and roll relative angles.

[0077] In step S2, the specific steps of adjusting the spatial posture of the linear array imaging module by using the posture adjustment module are as follows:

[0078] According to the measured relative posture information, a six-axis motion system is used to perform three-dimensional translation and pitch, yaw and roll angle adjustment on the linear array imaging module, so that the focal plane of the objective lens is parallel to the normal plane of the rotating body surface, and the distance is the focal length of the objective lens, thereby achieving the focusing of the imaging module.

[0079] In step S3, the process of acquiring the depth information is specifically as follows:

[0080] The reflected light spot of the biased line laser on the surface of the rotating body is obtained through the linear array tracking module. According to the linear relationship between the depth and the spot size, the undulation data of the surface of the rotating body is obtained with the focal plane of the objective lens as the original plane. In combination with the relative posture information between the turntable rotation axis of the rotating body and the linear array tracking module, the three-dimensional morphology data of the surface of the rotating body is obtained as the depth information.

[0081] In step S3, when the spatial posture of the linear array imaging module is adjusted to the optimal position, the linear array imaging module and the linear array tracking module synchronously collect data so that the collected depth data and texture data correspond to the same measured area of ​​the rotating body.

[0082] That is, the acquisition of texture information and depth information is synchronized. Specifically, when the linear array imaging module is adjusted to the optimal position, the linear array focus tracking module is also in the optimal spatial position. The two modules collect images at the same time to ensure that the depth data and texture data correspond to the same measured area. The linear array focus tracking module and the linear array imaging module together constitute a real-time, closed-loop control measurement system to achieve the measurement of complex rotating surfaces such as cones and transformation curved rotating surfaces. Usually, the visible light band actually used by the illumination light source is separated from the infrared band of the line light source used for focus tracking. A dichroic mirror with specific parameters is used to ensure that the light of the two bands does not interfere with each other during the focus tracking and measurement process.

[0083] In step S5, the process of fusing the relative posture information, depth information and texture information includes:

[0084] For a regular rotating body, the posture information of each measurement position is converted into spatial coordinates, the depth information is superimposed on the spatial coordinates of the corresponding measurement position, and data fitting and error correction processing are performed to obtain a high-precision point cloud; the pixel points of the texture information of each measurement position are matched to the high-precision point cloud to form a multimodal point cloud with texture information;

[0085] Data fitting and error correction processing specifically include:

[0086] The least squares method is used to fit the plane circle, eliminate the errors such as eccentricity and tilt introduced by the workpiece clamping, and obtain a high-precision point cloud. The formula is as follows:

[0087] The equation of a circle is:

[0088] x 2 +y 2 +ax+by+c=0

[0089] For parameters a, b, c, there are the following calculation formulas:

[0090]

[0091] Where Xi, Yi is the two-dimensional coordinate of the i-th point on the rotating body, N is the number of images collected in one revolution of the rotating body, i = 1, 2, …, N.

[0092] For irregular rotating bodies, the pose information, texture information and depth information of each measured position are sent to the pre-trained neural network MVSNet. The generalization ability of the neural network is used to automatically complete the coordinate matching, depth superposition and texture assignment of the irregular rotating body to form a multimodal point cloud with texture information.

[0093] In step S6, the processing of multimodal data includes:

[0094] Input the point cloud data containing spatial coordinates and color texture in the multimodal data into the pre-trained neural network to classify the defect type;

[0095] The point cloud is clustered according to the spatial position and defect type information, and points belonging to the same defect type are divided into a point cloud set;

[0096] According to the point cloud clustering results, the three-dimensional model of the rotating body is reconstructed, and defect information is marked on the three-dimensional model. The defect information includes defect type, defect area, defect center coordinates and defect circumscribed rectangle.

[0097] Specifically include:

[0098] S601: inputting the point cloud data containing spatial coordinates and color texture in the multimodal data into the neural network for classification, and the categories are flawless, pitted, scratched, colored, cracked, and spotted, a total of 6 categories;

[0099] S602: clustering the point cloud according to the spatial position and category information, and dividing the points belonging to the same defect into a point cloud set;

[0100] S603: Reconstruct the three-dimensional model of the rotating body, and mark the defects on this basis to form the final result of the rotating body defect measurement. The output defect information includes defect type, defect area, center coordinates, and circumscribed rectangle.

[0101] Example 2

[0102] The present embodiment provides a rotating body defect detection device for realizing a rotating body defect detection method based on linear array focus tracking as in embodiment 1, which is used for defect detection of a rotating body 5, and comprises a controllable turntable, a linear array focus tracking module, a beam splitter 3, an objective lens 4, a dichroic mirror 6, a calculation control unit 8, a posture adjustment module 9, an illumination light source 10 and a linear array imaging module, wherein the linear array imaging module comprises a tube lens 11 and a linear array camera 12, and the linear array focus tracking module comprises a fixed focus light source 1, a linear bias module 2 and a focusing lens 13, and a focus tracking camera 7; the linear array focus tracking module, the beam splitter 3, the objective lens 4, the dichroic mirror 6, the illumination light source 10 and the linear array imaging module are all mounted on the posture adjustment module 9; the rotating body 5 is mounted on the controllable turntable;

[0103] The laser emitted by the fixed-focus light source 1 passes through the line bias module 2 to form a bias line laser, which is reflected by the beam splitter 3 and enters the objective lens 4 to converge into a fixed-focus line laser on the surface of the rotating body 5; the objective lens 4 collects the light reflected by the fixed-focus line laser from the rotating body 5, propagates to the dichroic mirror 6, and then reflects through the focusing lens 13 to enter the tracking camera 7; the calculation control unit 8 calculates the defocus amount and spatial posture according to the captured fixed-focus laser, and drives the posture adjustment module 9 to perform real-time focusing;

[0104] The illumination light source 10 emits visible light so that the surface of the rotating body 5 is evenly illuminated. The reflected light generated by the illumination of the rotating body 5 is transmitted through the dichroic mirror 6 into the tube mirror 11 and illuminates the camera, which collects the texture information of the surface of the rotating body 5.

[0105] Optionally, the line bias module 2 includes a collimating lens, a biased rectangular aperture and a cylindrical lens. The laser emitted by the fixed-focus light source 1 becomes parallel light after passing through the collimating lens. The parallel light passes through the biased rectangular aperture to block half of the light, forming a biased laser; the biased laser passes through the cylindrical lens to be focused in one direction, and remains parallel light in the other direction, thereby forming a biased line laser.

[0106] The reflected light of the biased line laser on the surface of the rotating body is modulated by the defocus direction and size of the rotating body surface and presents different shapes. The relative spatial position of the objective lens focal plane and the measured plane is calculated according to the corresponding relationship.

[0107] The linear array focus tracking module composed of the focus tracking camera 1, the fixed focus light source 2, the linear bias module 3 and the posture adjustment module 9 together constitute a real-time linear array focus tracking mechanism. Specifically:

[0108] The optical paths of the fixed-focus light source 2, the line bias module 3 and the focus tracking camera 1 are coupled through the dichroic mirror 6, the beam splitter 3 and the objective lens 4. The three units are jointly mounted on the posture adjustment module 9. The calculation control unit 8 obtains the image of the focus tracking camera in real time and calculates the defocus amount, drives the posture adjustment module 9 to adjust the posture, and forms a closed-loop feedback control system to realize real-time focus tracking.

[0109] like Figure 6As shown, the rotating body defect detection device of this embodiment measures a rotating body with a regular surface and high reflectivity, specifically a metal cylinder. The measurement process is as follows: the rotating body to be measured is mounted on a controllable turntable and placed in the initial position. The infrared fixed-focus light source 1 is expanded into parallel light by a collimating lens, and a biased line laser is generated through a biased rectangular aperture and a cylindrical lens, which is finally irradiated on the rotating body to be measured through a beam splitter and an objective lens; the line laser is reflected through the objective lens, the beam splitter, and the dichroic mirror into the tracking camera, and the relative posture is calculated according to the shape of the light spot, and the posture adjustment module is executed to achieve focusing; dark field illumination is used to ensure uniform illumination and collect texture information, and depth information and posture information are collected synchronously; the turntable rotates a certain angle to start the next measurement until the measurement is completed. The traditional method is used to fuse the posture information, depth information, and texture information, and the least squares fitting circle is used to eliminate error points to form a point cloud. The pixel points of the texture information are further matched to the high-precision point cloud to form a multimodal point cloud with texture information.

[0110] like Figure 8 and Fig. 9 As shown, the effect comparison between the rotating body defect detection method based on linear array tracking focus of the present invention and the existing method based on linear array tracking focus of the present invention is shown. Through the comparison, it can be found that the detection effect clarity of the method of the present invention is higher.

[0111] Example 3

[0112] like Figure 7 As shown, this embodiment is substantially the same as Embodiment 2, except that this embodiment is used to measure irregular surfaces and rotating bodies with medium and low reflectivity. This embodiment uses bright field light source illumination and a linear array camera for imaging, and the brightness of the defective area is lower than that of the smooth object surface. The fusion of depth information, posture information, and texture information on the surface of the rotating body is relatively complex. The neural network MVSNet is used for data fusion, error points are eliminated, coordinate matching is completed, and a multimodal point cloud with texture information is formed.

[0113] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for detecting defects of a rotating body based on linear array focus tracking, characterized in that: The following steps are involved: The rotating body to be measured is mounted on a controllable turntable, an objective lens is set to collect light beams, a linear array imaging module is set to collect texture information on the surface of the rotating body, a linear array tracking module is set to obtain relative posture information between the linear array imaging module and the rotating body and depth information of the current measured area on the surface of the rotating body, and a posture adjustment module is set to adjust the spatial posture of the linear array tracking module and the linear array imaging module; The controllable turntable is used to drive the rotating body to rotate to different measurement positions. At different measurement positions, the relative posture information between the linear array imaging module and the rotating body is obtained according to the linear array tracking module, and the spatial postures of the linear array imaging module and the linear array tracking module are adjusted so that the current measured area of ​​the rotating body is completely within the focal depth range of the linear array imaging module, and then the texture information and depth information of the current measured area on the surface of the rotating body are obtained; The relative posture information, depth information and texture information obtained at each measurement position are fused to extract the defect information and three-dimensional shape information of the rotating body surface; The process of the linear array focus tracking module acquiring the relative posture information between the linear array imaging module and the rotating body specifically includes: Generate a bias line laser that can identify the defocus size and direction of the object and irradiate it onto the surface of the rotating body; The focus tracking camera in the linear array focus tracking module obtains the reflected light spot of the offset line laser on the surface of the rotating body. The distance from each point on the offset line laser to the surface of the rotating body will cause each point corresponding to the reflected light spot to expand horizontally. The expansion size is calculated in terms of the number of pixels. For n points on the current measured area of ​​the rotating body, the actual size {d1, d2, …, d n }; The calculation expression for converting the pixel size to the actual size of n points on the current measured area of ​​the rotating body is: d i =P Z C xi ,i=1,2,…,n Where i is the number of the point in the current measured area of ​​the rotating body, P Z is the pixel size of the tracking camera in the linear array tracking module, C xi is the coordinate of the edge of the reflected light spot corresponding to the i-th point in the measured area in the imaging picture of the tracking camera; The calculation formula for the defocus amount of each point in the current measured area of ​​the rotating body is: Where i is the number of the point in the current measured area of ​​the rotating body, δ i is the defocus value corresponding to the i-th point in the current measured area of ​​the rotating body, f1 is the focal length of the objective lens, f2 is the focal length of the focusing lens of the linear array tracking module, and d i is the extended size of the reflected light spot corresponding to the i-th point in the current measured area of ​​the rotating body, and a is the initial width of the bias line laser provided by the line bias module; Finally, according to the actual size of the expansion, the defocus value of each point in the current measured area of ​​the rotating body is calculated, and the coordinate transformation matrix is ​​used to calculate the relative posture information between the linear array imaging module and the current measured area of ​​the rotating body.

2. The method for detecting defects of a rotating body based on linear array focus tracking according to claim 1, characterized in that: The process of adjusting the spatial posture of the linear array imaging module and the linear array focus tracking module is specifically as follows: According to the relative posture information, a six-axis motion system is used to perform three-dimensional translation and pitch, yaw and roll angle adjustment on the linear array imaging module, so that the focal plane of the objective lens is parallel to the best fitting plane of the rotating body surface, and the linear array imaging module is located at the clearest imaging position.

3. The method for detecting defects of a rotating body based on linear array focus tracking according to claim 1, characterized in that: The process of obtaining the depth information is specifically as follows: The reflected light spot of the biased line laser on the surface of the rotating body is obtained through the linear array tracking module. According to the linear relationship between the depth and the spot size, the undulation data of the surface of the rotating body is obtained with the focal plane of the objective lens as the original plane. In combination with the relative posture information between the turntable rotation axis of the rotating body and the linear array tracking module, the three-dimensional morphology data of the surface of the rotating body is obtained as the depth information.

4. The method for detecting defects of a rotating body based on linear array focus tracking according to claim 1, characterized in that: The objective lens position is moved axially, and the focusing conditions of the linear array imaging module and the linear array focus tracking module with the rotating body are adjusted by changing the position of the objective lens, so that the linear array imaging module is located at the clearest imaging position.

5. The method for detecting defects of a rotating body based on linear array focus tracking according to claim 1, characterized in that: The process of fusing relative pose information, depth information and texture information includes: For a regular rotating body, the posture information of each measurement position is converted into spatial coordinates, the depth information is superimposed on the spatial coordinates of the corresponding measurement position, and data fitting and error correction processing are performed to obtain a high-precision point cloud; the pixel points of the texture information of each measurement position are matched to the high-precision point cloud to form a multimodal point cloud with texture information; For irregular rotating bodies, the pose information, texture information and depth information of each measured position are sent to the pre-trained neural network for coordinate matching, depth superposition and texture assignment to obtain a multimodal point cloud with texture information.

6. The method for detecting defects of a rotating body based on linear array focus tracking according to claim 1, characterized in that: The method further includes processing the multimodal data of the surface of the rotating body obtained after data fusion, thereby extracting surface defect information; The multimodal data processing process includes: Input the point cloud data containing spatial coordinates and color texture in the multimodal data into the pre-trained neural network to classify the defect type; The point cloud is clustered according to the pose information and defect type information, and the points belonging to the same defect type are divided into a point cloud set; According to the point cloud clustering results, the three-dimensional model of the rotating body is reconstructed, and defect information is marked on the three-dimensional model. The defect information includes defect type, defect area, defect center coordinates and defect circumscribed rectangle.

7. A rotating body defect detection device implementing a rotating body defect detection method based on linear array focus tracking as claimed in any one of claims 1 to 6, used for defect detection of a rotating body (5), characterized in that: The invention comprises a controllable turntable, a linear array focus tracking module, a beam splitter (3), an objective lens (4), a dichroic mirror (6), a calculation control unit (8), a posture adjustment module (9), an illumination light source (10) and a linear array imaging module, wherein the linear array imaging module comprises a tube lens (11) and a linear array camera (12), and the linear array focus tracking module comprises a fixed focus light source (1), a linear bias module (2), a focusing lens (13) and a focus tracking camera (7); the linear array focus tracking module, the beam splitter (3), the objective lens (4), the dichroic mirror (6), the illumination light source (10) and the linear array imaging module are all mounted on the posture adjustment module (9); and the rotating body (5) is mounted on the controllable turntable; The laser light emitted by the fixed-focus light source (1) passes through the line bias module (2) to form a bias line laser light, which is reflected by the beam splitter (3) into the objective lens (4) and converges on the surface of the rotating body (5) to form a fixed-focus line laser light; the objective lens (4) collects the light reflected by the rotating body (5) from the fixed-focus line laser light, propagates to the dichroic mirror (6), and then reflects through the focusing lens (13) to enter the tracking camera (7); the calculation control unit (8) calculates the defocus amount and the spatial posture according to the captured fixed-focus laser light, and drives the posture adjustment module (9) to perform real-time focusing; The illumination light source (10) emits visible light so that the surface of the rotating body (5) is evenly illuminated. The reflected light generated by the illumination of the rotating body (5) is transmitted through the dichroic mirror (6) into the cylindrical mirror (11) and illuminates the linear array imaging module, which collects texture information on the surface of the rotating body (5).

8. The rotating body defect detection device according to claim 7, characterized in that: The line bias module (2) comprises a collimating lens, a biased rectangular aperture and a cylindrical lens. The laser emitted by the fixed-focus light source (1) becomes parallel light after passing through the collimating lens. Half of the parallel light is blocked by the biased rectangular aperture to form a biased laser. The biased laser is focused in one direction by the cylindrical lens, while the other direction remains parallel light, thus forming a biased line laser.

9. The rotating body defect detection device according to claim 7, characterized in that: The calculation control unit (8) acquires the image of the focus tracking camera (7) in real time and calculates the spatial posture, and drives the posture adjustment module (9) to adjust the posture, thereby forming a closed-loop feedback control system to achieve real-time focus tracking.

Images