Vehicle body paint surface imaging device and defect detection system
By using a body paint imaging device in automobile paint inspection, combined with array-shaped light sources and simulated environmental light sources, images under various lighting modes are collected, and deep learning models are used for detection, which solves the problem of high missed detection rate in the existing technology. problem, achieving more accurate paint surface defect detection.
Patent Information
- Application Number
- CN202323458829.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Utility models(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-19
- Publication Date
- 2025-01-21
- Estimated Expiration
- 2033-12-19
AI Technical Summary
The existing automotive paint detection methods mainly rely on visual methods and visual methods, and have problems such as strong subjectivity, great influence on light angles, and the inability to identify defects without curvature changes, resulting in high missed detection rates.
A car body paint imaging device is used, including array-shaped light sources arranged horizontally and vertically and simulated environmental light sources. Images under multiple lighting modes are collected by switching modules, and defects are detected using a deep learning model.
It can effectively reveal defects with curvature changes, such as particles, shrinkage cavities, paint scars, etc., and identify defects such as poor paint, bottom exposure, and solvent contamination without curvature changes, reducing misjudgments and improving detection accuracy.
Smart Images
Figure CN222379564U_ABST
Abstract
Description
Technical Field
[0001] The utility model relates to the technical field of machine vision detection, and in particular to a vehicle body paint surface imaging device and a defect detection system. Background Art
[0002] Cars have become one of the indispensable tools for people's work and life, and people have high demands for their appearance quality. The paint surface is the car's clothes and the carrier of the car's appearance quality. However, in the process of painting and assembly, defects such as paint scars, particles, and scratches often occur. The existence of appearance defects will inevitably lead to sales and production disputes in car sales. In order to avoid the above disputes, it is very necessary to test the paint surface of the whole car before the car leaves the factory. In addition, with the gradual increase in human labor, the production line has an increasingly strong demand for automated procedures, and the automatic detection and repair of paint defects has become a new development trend.
[0003] At present, the main means of inspecting automobile paint surfaces are visual inspection and visual detection. Among them, the visual inspection method is greatly affected by the proficiency of the person being inspected and is highly subjective. In addition, since the paint surface is a highly reflective surface, it is greatly affected by the angle of illumination, and it is inevitable that there will be many missed inspections by human eyes. Moreover, long-term inspection will cause eye fatigue, which will also cause missed inspections of appearance defects. The current light source scheme of the visual inspection method uses the phase shift of structured stripe light for defect recognition, which can only target defects with curvature changes on the surface of the target being tested, that is, only targets particles, shrinkage cavities, paint scars, paint residues, etc. on the paint surface, but cannot identify defects such as poor paint, exposed bottom, solvent pollution, etc. without curvature changes. At the same time, the stripe light only has a light and shadow gradient in a single vertical direction, resulting in missed inspections of defects parallel to the stripe light. Utility Model Content
[0004] The purpose of the utility model is to provide a vehicle body paint surface imaging device and a defect detection system, which can make paint surface defects easier to reveal and can be used in conjunction with a variety of light sources to avoid missing defects.
[0005] In order to achieve the above purpose, the technical solution adopted by the utility model is as follows:
[0006] In the first aspect, the utility model provides a vehicle body paint imaging device, including a light source module, a switching module and a collection module, the light source module including a first light source and a second light source, the first light source is an array-shaped light source arranged horizontally and vertically, and the second light source is simulated ambient light; the switching module is connected to the light source module, and is used to switch the first light source or the second light source; the collection module is used to collect the vehicle body paint image under the illumination of the light source module.
[0007] Furthermore, the first light source includes a first chessboard array light source in which black and white are alternately distributed, and a second chessboard array light source in which the black and white are complementary to the first chessboard array light source.
[0008] Furthermore, the second light source includes a first environment simulation light and a second environment simulation light; the color temperature of the first environment simulation light is 6500±200K, which is used to simulate sunlight; the color temperature of the second environment simulation light is 5000±200K, which is used to simulate light on a cloudy day.
[0009] In the second aspect, the utility model provides a vehicle body paint defect detection system, comprising a first moving component, a second moving component, a prediction component and the above-mentioned vehicle body paint imaging device; the first moving component is used to transport the vehicle to be inspected to the inspection station or to transport the vehicle that has completed the inspection away from the inspection station; the acquisition module of the vehicle body paint imaging device is fixed on the second moving component, and the acquisition module is driven by the second moving component to move along a preset trajectory to acquire vehicle body paint images of different areas under the illumination of the light source module; the prediction component is used to receive the vehicle body paint images acquired by the acquisition module, and obtain the detection results based on a preset prediction model.
[0010] Furthermore, it also includes an offset acquisition component for acquiring the offset of the vehicle to be inspected at the inspection station, and the second moving component corrects the preset trajectory based on the offset.
[0011] Furthermore, it also includes a plurality of distance acquisition components, which are arranged around the acquisition module and are used to obtain the coordinates of each preset measurement position in the field of view of the acquisition module and the distance between the acquisition module and the vehicle to be detected.
[0012] Beneficial effects of the utility model:
[0013] 1. The first light source of the utility model is an array light source arranged horizontally and vertically. When the array light generated by the array light source is projected onto the paint surface of the vehicle body to be tested, the image is smooth and natural for qualified paint surfaces. When there are concave-convex defects on the paint surface, no matter where the concave-convex defects are, they will be distorted locally under the illumination of the array light arranged horizontally and vertically, effectively revealing the defects with curvature changes on the surface of the target to be tested, that is, particles, shrinkage holes, paint scars, paint residues, etc. on the paint surface. The second light source is simulated ambient light. Through environmental simulated lighting, the color difference defects of the vehicle body paint surface are captured, the misjudgment caused by the lighting environment or hardware is reduced, and the defects such as poor paint without curvature changes, exposed bottom, and solvent pollution are effectively identified.
[0014] 2. The utility model is connected to the light source module through a switching module, which is used to switch the first light source or the second light source, and then collect images under multiple lighting modes for the same car body paint surface, and output defect detection results based on the image synthesis, which effectively avoids defect omissions and improves detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] In order to more clearly illustrate the specific implementation of the utility model or the technical solution in the prior art, the drawings required for use in the implementation or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the utility model.
[0016] Figure 1 is a schematic diagram of the composition of the vehicle body paint surface imaging device described in an embodiment of the present application;
[0017] Figure 2 is a front schematic diagram of the vehicle body paint surface imaging device according to an embodiment of the present application;
[0018] Figure 3 Schematic diagram of the composition of the vehicle body paint defect detection system described in the embodiment of the present application;
[0019] Figure 4 is a schematic diagram of the layout of the robotic arm described in an embodiment of the present application;
[0020] Figure 5 is a schematic diagram of the first chessboard array light in an embodiment of the present application;
[0021] Figure 6 is a schematic diagram of the second chessboard array light in an embodiment of the present application;
[0022] Figure 7 This is a schematic diagram of the conventional streak light imaging capability;
[0023] Figure 8 It is a schematic diagram of the optical imaging capability of the array-shaped light arranged horizontally and vertically as described in the embodiments of the present application.
[0024] In the figure, 1 is a vehicle body paint imaging device, 11 is a light source module, 12 is a collection module, 13 is a control module, 14 is a distance collection component, 2 is a second moving component, 21 is a slide rail, 22 is a mechanical arm, 3 is an offset acquisition component, 4 is a detection result display screen, 5 is a vehicle to be detected, and 6 is a distortion area. DETAILED DESCRIPTION
[0025] The following will describe the implementation of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific implementations, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.
[0026] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention. The drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.
[0027] Example 1, see Figure 1 and Figure 2 As shown, the present application provides a vehicle body paint imaging device 1, which includes a light source module 11, a switching module and a collection module 12, the light source module 11 includes a first light source and a second light source, the first light source is an array-shaped light source arranged horizontally and vertically, and the second light source is simulated ambient light; the switching module is connected to the light source module 11, and is used to switch the first light source or the second light source; the collection module 12 is used to collect the vehicle body paint image under the illumination of the light source module 11.
[0028] Since the first light source is an array-shaped light source arranged horizontally and vertically, when the array light generated by the array-shaped light source is projected onto the paint surface of the vehicle body to be tested, the imaging is smooth and natural for a qualified paint surface. When there are concave-convex defects on the paint surface, no matter the direction of the concave-convex defects, they will be distorted locally under the irradiation of the array light arranged horizontally and vertically, effectively revealing the defects with curvature changes on the surface of the target to be tested, namely, particles, shrinkage holes, paint scars, paint residues, etc. on the paint surface.
[0029] Since the second light source simulates ambient light, the paint surface of the vehicle body is illuminated by the ambient simulated light to capture the color difference defects of the paint surface of the vehicle body, reduce misjudgment caused by lighting environment or hardware, and effectively identify defects such as poor paint without curvature change, exposed bottom, solvent contamination, etc.
[0030] The light source module 11 described in the present application is a program-controlled adjustable array light source, that is, it is pre-set by a program to form a first light source and a second light source, and an automatic switching of the first light source and the second light source is achieved through a program-integrated switching module to improve the efficiency of vehicle body paint image acquisition.
[0031] In some embodiments, the first light source includes a first checkerboard array light source with black and white alternately distributed and a second checkerboard array light source with black and white complementary to the first checkerboard array light source. Figure 5 As shown, the first chessboard array light source irradiates the paint surface of the vehicle body to form a first chessboard array light, and the first chessboard array light is arranged in a horizontal and vertical array, effectively revealing the concave and convex defects of the paint surface of the vehicle body. Figure 6 As shown, the second chessboard array light source is irradiated on the paint surface of the vehicle body to form a second chessboard array light, and the second chessboard array light is also arranged in a horizontal and vertical array, and the second chessboard array light and the first chessboard array light are complementary in black and white, thereby being able to more fully detect the overall defects of the vehicle body paint surface and avoid missed detection.
[0032] Detection effect description: see Figure 7 As shown in Figure 1, when a traditional stripe light source is used for image acquisition, if the defect is parallel to the stripe light, no distortion will be produced, resulting in missed defect detection. Figure 8 As shown, when an array-shaped light source arranged horizontally and vertically is used for image acquisition, it will not be limited by the direction of the defect, and a corresponding distortion area 6 will be generated, thereby fully detecting the overall defects of the vehicle body paint surface and avoiding missed detection.
[0033] In some embodiments, the second light source includes a first environment simulation light and a second environment simulation light; the first environment simulation light has a color temperature of 6500±200K, which is used to simulate sunlight; the second environment simulation light has a color temperature of 5000±200K, which is used to simulate illumination on a cloudy day. Preferably, the present application combines the vehicle body paint visual observation scene, and specifically defines two standard light sources with color temperatures of 6500K and 5000K, respectively, and presents a scattered light effect, simulating the light environment of daytime sunlight and daytime cloudy days, respectively, and capturing the color difference defects of the vehicle paint surface through sensors from the customer's perspective, reducing misjudgment caused by lighting environment or hardware and other reasons.
[0034] In some embodiments, the vehicle body paint imaging device 1 further includes a control module 13, which is connected to the light source module 11 and the acquisition module 12, and controls the light source module 11 and the acquisition module 12 to perform a response action in response to a detection instruction.
[0035] Example 2, see Figure 3As shown, the present application provides a vehicle body paint defect detection system, including a first moving component, a second moving component 2, a prediction component and the vehicle body paint imaging device described in Example 1. The first moving component is used to transport the vehicle 5 to be inspected to the inspection station or to transport the vehicle that has completed the inspection away from the inspection station. The acquisition module 12 of the vehicle body paint imaging device 1 is fixed on the second moving component 2, and the acquisition module 12 is driven by the second moving component 2 to move along a preset trajectory to acquire vehicle body paint images of different areas illuminated by the light source module 11. The prediction component is used to receive the vehicle body paint image acquired by the acquisition module 12, and obtain the detection result based on the preset prediction model.
[0036] In some embodiments, see Figure 4 As shown, the second moving assembly 2 includes two slide rails 21 arranged in the direction of travel of the vehicle 5 to be detected and a mechanical arm 22 moving along the slide rails 21. The two slide rails 21 are arranged on the left and right sides of the vehicle 5 to be detected. It should be noted that the number of mechanical arms 22 is adaptively adjusted according to the number of slide rails 21. When the number of slide rails 21 is 0, the number of mechanical arms 22 is 4; when the number of slide rails 21 is 1, the number of mechanical arms 22 is 3; when the number of slide rails 21 is 2, the number of mechanical arms 22 is 2. In particular, when the number of slide rails 21 is 1 or 2, the mechanical arm 22 can consider the shape-adaptive detection solution.
[0037] The preset trajectory of the robot arm 22 is the robot arm motion trajectory obtained by sampling and debugging a standard vehicle before the system goes online.
[0038] The prediction component includes a detection algorithm module, and the detection algorithm mainly includes an algorithm detection for concave-convex defects and an algorithm detection for color difference defects.
[0039] The working principle of the algorithm for detecting concave-convex defects is as follows: the vehicle body paint imaging device 1 collects images and inputs them into the positive sample deep learning model for abnormality detection, and the detected output abnormal images are sent to the negative sample deep learning model for defect classification detection, and finally the defect location and its name are output.
[0040] Among them, the positive sample deep learning model adopts the positive sample learning method, takes qualified images as samples for learning, and obtains a deep learning model. Unlike the traditional model that outputs higher than the model confidence, this algorithm outputs image results that are lower than the model confidence.
[0041] Among them, the negative sample deep learning model is obtained by labeling and learning a limited number of defects to obtain a classification detection model.
[0042] By combining the positive sample deep learning model with the negative sample deep learning model, the problem of low detection rate of small sample models in visual detection in this scenario can be effectively overcome.
[0043] The working principle of the color difference defect algorithm detection is the same as that of the concave-convex defect algorithm detection: the vehicle body paint imaging device 1 collects images and inputs them into the positive sample deep learning model for abnormality detection, and the detected output abnormal image is sent to the negative sample deep learning model for defect classification detection, and finally the defect location and its name are output.
[0044] Among them, the image detected by the algorithm for concave-convex defects comes from the imaging under the first light source mode, and the image detected by the algorithm for chromatic aberration defects comes from the imaging under the second light source mode.
[0045] Among them, if the defect results detected under multiple different light source modes belong to the same position, they will be merged and output; if they do not belong to the same position, they must all be reflected in the final output result.
[0046] In some embodiments, the vehicle body paint defect detection system also includes an offset acquisition component 3, which is used to obtain the offset of the vehicle to be detected at the detection station, and the second mobile component 2 corrects the preset trajectory based on the offset. Exemplarily, the offset acquisition component 3 is composed of 4 to 6 groups of two-dimensional cameras or three-dimensional cameras, and the X-direction offset of the vehicle body is calculated by coordinate calibration and combining the vehicle body features obtained by shooting with the standard vehicle features. Among them, the X-direction offset refers to the direction of movement of the vehicle body on the slide rail, that is, the direction of travel of the vehicle. In particular, if the detection station is a precision positioning station, the offset acquisition component 3 can be shielded or cancelled, and the mechanical arm 22 and the slide rail 21 of the second mobile component 2 will move according to the standard preset trajectory.
[0047] In some embodiments, the vehicle body paint defect detection system also includes a plurality of distance acquisition components 14, which are arranged around the acquisition module 12 and are used to obtain the coordinates of each preset measurement position within the field of view of the acquisition module 12 and the distance between the acquisition module 12 and the vehicle to be detected 5, so as to prevent a collision accident between the vehicle body paint imaging device 1 and the vehicle to be detected 5 due to the close distance.
[0048] The vehicle body paint defect detection system described in this application comprises the following steps when working specifically:
[0049] Step 1: The red vehicle of a certain model to be detected moves to the detection area, triggering the in-place light barrier.
[0050] Step 2: After receiving the arrival information of the vehicle 5 to be detected, the offset acquisition component 3 performs photogrammetry on the vehicle body and calculates the X-direction offset of the vehicle body to be 6mm. The offset acquisition component 3 inputs the X-direction offset of 0.6mm into the control element of the second moving component 2.
[0051] Step 3: The control element of the second moving component 2 corrects and compensates the standard preset trajectory according to the offset, and directs the robot arm 22 and the slide rail 21 to move along the compensated trajectory. During the movement of the robot arm 22 and the slide rail 21, the vehicle body paint imaging device 1 fixed at the front end of the robot arm 22 switches four different light source modes as described in this application at each target point for taking photos. The target points are the positions where the robot arm drives the imaging system to stay on the surface of the entire vehicle during the system debugging process to ensure that the standard vehicle paint is fully covered by the imaging system's field of view. During the image taking process, the vehicle body paint imaging device 1 synchronously transmits the collected images to the detection algorithm module of the prediction component for defect detection and defect location.
[0052] Preferably, to ensure full coverage of the vehicle body inspection area, two adjacent image fields need to overlap, and an overlap of 1 / 5 is optimal. At the same time, to ensure the maximization and effectiveness of the imaging field of view, the axis of the field of view of the acquisition module 12 of the vehicle body paint imaging device 1 coincides with the normal direction of the target point.
[0053] Step 4: After the detection algorithm module completes the detection, it outputs the defect information to the central control module, including category, size, coordinates, etc. The central control module binds the defect information with the basic information such as the vehicle model VIN code, production date, batch, etc. and stores it in the server.
[0054] Step 5: When all the robotic arms have completed their reset according to the compensated trajectory, the robotic arm control system transmits the vehicle release signal to the production line. The production line scheduling system starts the vehicle and leaves the inspection station and enters the buffer area.
[0055] Step 6: When the vehicle runs to the repair station in the repair area and triggers the production line code reading system, the central control module receives the VIN code information of the vehicle at the production line repair station, and sends the corresponding vehicle detection information to the corresponding detection result display screen 4 of the repair station for visual display, so as to facilitate the repair personnel to refer to and search and repair.
[0056] The above embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Any equivalent substitution or modification made by a person skilled in the art based on the present invention is within the protection scope of the present invention.
Claims
1. A vehicle body paint imaging device, characterized in that: It includes a light source module, a switching module and a collection module, wherein the light source module includes a first light source and a second light source, wherein the first light source is an array light source arranged horizontally and vertically, and the second light source is a simulated ambient light; The switching module is connected to the light source module and is used to switch the first light source or the second light source; The acquisition module is used to acquire the vehicle body paint surface image illuminated by the light source module.
2. The vehicle body paint surface imaging device according to claim 1, characterized in that: The first light source includes a first checkerboard array light source in which black and white are alternately distributed, and a second checkerboard array light source in which the black and white are complementary to the first checkerboard array light source.
3. The vehicle body paint surface imaging device according to claim 1 or 2, characterized in that: The second light source includes a first environment simulation light and a second environment simulation light; The color temperature of the first environment simulated light is 6500±200K, which is used to simulate sunlight; The color temperature of the second environment simulated light is 5000±200K, which is used to simulate the lighting on a cloudy day.
4. A vehicle body paint defect detection system, characterized in that: It comprises a first moving component, a second moving component, a prediction component and the vehicle body paint surface imaging device according to any one of claims 1 to 3; The first moving component is used to transport the vehicle to be inspected to the inspection station or to transport the vehicle that has completed the inspection away from the inspection station; The acquisition module of the vehicle body paint imaging device is fixed on the second moving component, and the acquisition module is driven by the second moving component to move along a preset trajectory to acquire vehicle body paint images of different areas illuminated by the light source module; The prediction component is used to receive the vehicle body paint surface image acquired by the acquisition module, and obtain the detection result based on the preset prediction model.
5. The vehicle body paint defect detection system according to claim 4, characterized in that: It also includes an offset acquisition component for acquiring the offset of the vehicle to be inspected at the inspection station, and the second moving component corrects the preset trajectory based on the offset.
6. The vehicle body paint defect detection system according to claim 4, characterized in that: It also includes a plurality of distance acquisition components, which are arranged around the acquisition module and are used to obtain the coordinates of each preset measurement position in the field of view of the acquisition module and the distance between the acquisition module and the vehicle to be detected.