An image recognition-based spoiler automatic polishing method

Abstract

The application discloses an image recognition-based spoiler automatic polishing method, which acquires a gray-scale image of a region near a spoiler point to be polished by using an image sensor, performs image analysis according to the gray-scale distribution, determines polishing parameters of the point to be polished, and executes a polishing operation according to the polishing parameters. The polishing parameters include polishing duration, polishing force, polishing rotating speed, and for a double-polishing-head or multi-polishing-head system, the polishing parameters further include polishing head selection. The application can realize full-automatic polishing of the spoiler and automatic acquisition of program polishing experience parameters through the joint design and application of the image sensor, the pressure sensor, the illumination sensor and the mechanical arm.

Description

Technical Field

[0001] This invention belongs to the field of automotive spoiler technology, specifically relating to an automatic spoiler polishing method based on image recognition. Background Technology

[0002] As the outermost trim and protective component of a car, the spoiler requires a high degree of streamlined appearance and smoothness. Generally, automotive spoilers are injection molded products, and the injection gate needs to be cut and polished before they can be packaged and shipped.

[0003] Traditional polishing operations are performed manually. While some facilities use robotic arms for automated polishing, these systems typically rely on experience or experiments to determine parameters such as rotation speed, time, and movement path. They then implement fixed polishing procedures, including switching between polishing heads in dual-head systems, all following a fixed program. This lack of real-time feedback and correction regarding polishing results means that some automatically polished spoilers still require manual finishing after inspection. Furthermore, while fixed-parameter automated polishing programs offer high overall efficiency once the desired parameters are obtained, achieving the ideal parameters is complex and time-consuming. In summary, both the experimental process for obtaining polishing parameters and the production process of polishing finished products suffer from a lack of effective real-time feedback, resulting in a time-consuming, labor-intensive, inefficient, and low-automation-level process. Summary of the Invention

[0004] To address the above problems, this invention designs an automatic spoiler grinding method based on image recognition. Through the combined design and application of image sensors, pressure sensors, illuminance sensors, and robotic arms, fully automatic spoiler grinding and automatic acquisition of programmed grinding experience parameters can be achieved.

[0005] This invention discloses an automatic spoiler polishing method based on image recognition. The method utilizes a real-time image sensor to acquire grayscale images of the area near the spoiler to be polished. Image analysis is performed based on the grayscale distribution to determine polishing parameters for the area to be polished, and then the polishing operation is executed according to these parameters. The method includes the following polishing steps:

[0006] S1. Acquire and analyze images, determine grinding parameters: The robotic arm drives the camera to be aligned with the point to be ground, acquires and analyzes the image of the point to be ground, and determines the grinding parameters of the point to be ground.

[0007] S2. Switch the grinding head according to the grinding parameters. The robotic arm drives the grinding head to approach the point to be ground. The initial coordinates of the point to be ground are preset according to the type of spoiler. If it is a single grinding head, there is no need to switch. The grinding head is directly driven to approach the point to be ground.

[0008] S3. Drive the grinding head to press against the point to be ground based on the difference between the pressure sensor sampling value and the preset grinding force;

[0009] S4. Issue a speed control command to drive the speed control device to rotate the grinding head.

[0010] S5. Perform polishing according to the set or calculated polishing time.

[0011] The image recognition-based automatic spoiler polishing method described in this invention is called an adaptive fully automatic polishing method, which corresponds to fixed-parameter automatic program polishing. The point to be polished is generally the central target point that drives the polishing head to move when the robotic arm performs the polishing operation, and it is also the center point of the image area. A certain deviation is allowed, such as ±1mm, and the error depends on the installation error of the image sensor and the robotic arm. In fact, the polishing operation is also to polish a local area with the target point as the center. It is impossible to polish only one point. Therefore, it is sometimes described as the area to be polished.

[0012] Regarding the points to be polished: Regardless of whether it is a new or mature model of spoiler, the same model of spoiler is a product of the automated production line with the same mold. The points and features to be polished are consistent. The points and sequence to be polished are determined in advance based on the actual situation after the spoiler is formed. Even if the points initially determined for a new type of spoiler are not ideal, the automatic polishing process of this invention can be used to analyze the images after each point has been polished, add or delete polishing points, thereby fixing the polishing points and sequence. This is beneficial for both fixed parameter automatic program polishing and adaptive fully automatic polishing.

[0013] For single-head grinding systems, the selection of grinding heads is not required; they can be applied directly. For ease of explanation, this invention is uniformly described as the grinding head selection and / or grinding head switching process.

[0014] Furthermore, the execution order of steps S3 and S4 is interchanged.

[0015] Furthermore, the automatic polishing method also includes polishing confirmation;

[0016] The polishing confirmation process includes, after the previous polishing is completed, the robotic arm drives the camera to re-align with the polished point, re-acquire the feature image, and analyze and confirm it. If the requirements are met, the process proceeds to the next point to be polished, and the next polishing process begins. Otherwise, the polishing parameters are recalculated, and the polishing is repeated, in a loop, until the requirements are met or the program exits due to limitations.

[0017] Furthermore, the polishing parameters include polishing time, polishing force, polishing speed, and polishing head selection. For a single polishing head system, the polishing head selected is a fixed polishing head. The process of acquiring and analyzing images and determining polishing parameters includes:

[0018] S11. Statistically analyze the grayscale distribution characteristics of the image within the target area. The target area includes a central region with a fixed length and width, centered on the image center point. The target area can be determined using a gate of fixed size. The point to be polished is generally the image center point or near the image center point. Calculate the grayscale variance of all pixels within the target area, or calculate the grayscale difference for adjacent pixels in rows and columns, and then calculate the variance for the grayscale difference in the two-dimensional row and column distribution. When calculating the variance, the mean is usually calculated simultaneously. The mean of different objects represents different physical meanings. For the straight part of the spoiler, the grayscale variance is used as the benchmark, and for the arc part, the variance of the grayscale difference is used as the comparison benchmark.

[0019] S12. Determine grinding parameters: This includes a method combining parameter grading and calculation. Any two of the three parameters—grinding time, grinding force, and grinding speed—are set to several grades. Based on the characteristics of the point to be ground and established principles, the graded grinding parameters for the point to be ground are first determined. The remaining parameters are calculated based on the correlation between grinding energy consumption and grinding effect, combined with image grayscale analysis results and the determined parameters. Since there are countless combinations of grinding head selection, grinding time, grinding force, and grinding speed to achieve the final grinding effect of a single point to be ground, dynamic optimization is a common research direction in the field of control. However, for the practical engineering application of this invention, the research on complex optimization schemes is not an urgent need. What is more important now is to have a reasonable and simple practical scheme. Therefore, it is simple and practical to pre-determine some parameters based on experience and then calculate individual parameters.

[0020] Furthermore, the method combining parameter grading and calculation includes:

[0021] S121. Determine the gear values: For a dual-grinding head system, first select the grinding head, define several fixed values ​​for the grinding force between the minimum and maximum values, define several fixed values ​​for the grinding speed between the minimum and maximum values, and determine the corresponding gear selection according to the characteristics of the point to be ground, the gray scale variance, or the gray scale difference variance.

[0022] S122. Calculate polishing time:

[0023] Let H1 be the first grayscale variance threshold and H2 be the second grayscale variance threshold, D1 be the first grayscale difference variance threshold and D2 be the second grayscale difference variance threshold, and let H be the grayscale variance and D be the grayscale difference variance of the target area pixels in real time. If the graded polishing force determined in step S121 is F and the polishing speed is V, then when using a fine polishing head, the polishing time is...

[0024] t=η1H / FV (Formula 1)

[0025] or

[0026] t = η²D / FV (Formula 2)

[0027] Wherein, η1 is the grayscale efficiency parameter of the fine grinding head, that is, the product of grinding force, grinding speed and grinding time required to achieve the desired grinding effect under unit grayscale variance; η2 is the grayscale difference efficiency parameter of the fine grinding head, that is, the product of grinding force, grinding speed and grinding time required to achieve the desired grinding effect under unit grayscale difference variance. Both parameters can be obtained through simple experiments. That is, for a certain grayscale variance or grayscale difference variance, fix the grinding force and grinding speed, and measure the actual grinding time to obtain the corresponding efficiency parameter. Generally, at least 10 experiments are conducted, and the final statistical mean is used as the empirical efficiency parameter.

[0028] When using a coarse grinding head, the grinding time is...

[0029] t=η3(H-H0) / FV (Formula 3)

[0030] or

[0031] t=η4(D-D0) / FV (Formula 4)

[0032] Wherein, η3 is the coarse grinding head grayscale efficiency parameter, which is the product of grinding force, grinding speed, and grinding time required to achieve the desired grinding effect under unit grayscale variance; η4 is the coarse grinding head grayscale difference efficiency parameter, which is the product of grinding force, grinding speed, and grinding time required to achieve the desired grinding effect under unit grayscale difference variance; H0 is the grayscale variance threshold to prevent over-grinding, and D0 is the grayscale difference variance threshold to prevent over-grinding. η3 and η4 can also be obtained empirically through simple experiments. That is, for a certain grayscale variance or grayscale difference variance, fix the grinding force and grinding speed, and measure the actual grinding time to obtain the corresponding efficiency parameters. Generally, at least 10 experiments are conducted, and the final statistical mean is used as the empirical efficiency parameter; H0 ranges from 0 to 10, and D0 ranges from 0 to 5.

[0033] Furthermore, the switching of the grinding head includes determining the selected grinding head according to the grinding parameters in step S1, controlling the robotic arm to rotate so that the selected grinding head faces the point to be ground.

[0034] Furthermore, the step of driving the grinding head to approach the grinding point based on the difference between the pressure sensor sampling value and the preset grinding force includes: when the pressure sensor sampling value is less than the lower limit of the preset value, the robotic arm drives the grinding head to move towards the grinding point, making the grinding head closer to the grinding point; when the pressure sensor sampling value is greater than the upper limit of the preset value, the robotic arm drives the grinding head to move away from the grinding point. Note that this type of movement is generally a small distance movement. The lower limit of the preset value is the preset grinding force minus the allowable error, and the upper limit of the preset value is the preset grinding force plus the allowable error. The pressure sensor is also referred to as a force-feed sensor or force-feed system in engineering.

[0035] Furthermore, the steps also include first confirming and adjusting the ambient light intensity to meet the illumination requirements for spoiler shooting before image capture. When the illumination intensity measured by the illumination sensor in real time is lower than the minimum threshold value, the system controls the light source to be brightened until the illumination requirements are met.

[0036] Since artificial lighting generally does not exceed the maximum illuminance threshold, when the measured illuminance is higher than the maximum threshold, it indicates that the sunlight outside is too strong during the day, triggering an alarm in the system and requiring manual intervention, such as closing some doors and windows.

[0037] The illuminance confirmation process is generally an independent and autonomous process used to monitor changes in the brightness of the light source. When the brightness change exceeds the threshold, the system control module issues a light adjustment command to adjust the light brightness and ensure that the brightness of the light source remains constant. This process is completed automatically after the system is turned on and does not require any software intervention.

[0038] On the other hand, a method for determining fixed-program grinding parameters for a spoiler, also known as empirical grinding parameter determination, is characterized by: after the previous grinding is completed, grinding confirmation is performed, and all combined grinding parameters during the grinding process are recorded; if a single grinding operation meets the requirements, then the single grinding parameter is the fixed-program grinding parameter for that location; if two or more grinding operations are performed, parameter optimization is required, and the parameter optimization method includes:

[0039] The grinding head switching strategy remains unchanged. Different grinding heads are used, and for the combined grinding parameters recorded at the aforementioned points, the maximum grinding time (Tmax) and the maximum grinding speed (Vmax) are taken. Then the grinding force...

[0040] F=(F1*t1*V1+F2*t2*V2+…+Fn* tn*Vn) / (Tmax*Vmax);

[0041] When the calculated grinding force exceeds the threshold Ft, the threshold value Ft shall prevail. In this case, the grinding time needs to be extended appropriately. The new grinding time is as follows:

[0042] t=( F1*t1*V1+F2*t2*V2+…+Fn* tn*Vn) / (Ft*Vmax);

[0043] The basic idea behind this optimization scheme is to ensure that the effective energy consumption of multiple polishing processes is consistent with the effective energy consumption of a single polishing process.

[0044] For routine product polishing, the robotic arm's movement sequence is pre-set based on the spoiler type or structural parameters. The camera is aimed at the point to be polished, and the polishing parameters for that point are calculated before the polishing operation is performed. If it is necessary to confirm the polishing effect of that point, the robotic arm is usually driven back to re-photograph the polished point and perform image processing and analysis. Based on the analysis results, if the results are not satisfactory, new polishing parameters are determined and the polishing is repeated. After confirming that the requirements are met, the parameters of that point are recorded for subsequent empirical parameter determination.

[0045] Furthermore, grinding parameters were determined for multiple spoilers of the same model, and multiple sets of grinding experience parameters were obtained. The average of these experience parameters was then statistically averaged, and the average value was used as the final grinding experience parameter for this type of spoiler.

[0046] The advantages and beneficial effects of this invention are as follows: The automatic spoiler grinding method based on image recognition designed in this invention achieves full automation of the spoiler grinding process by adding an image sensor for real-time detection and confirmation of the grinding effect, eliminating the need for manual intervention. Since programmed grinding is more suitable for mass production and streamlined operations, this method can be used to obtain programmed grinding parameters for new spoiler models, greatly reducing the pressure of obtaining reliable parameters through manual experience and actual operation experiments, while significantly improving the efficiency of parameter acquisition. Undoubtedly, this method can also be directly used in the normal grinding process of spoilers. Compared to programmed grinding, it adds a synchronous detection and analysis process; if deficiencies are found, supplementary grinding is performed, which can significantly improve the efficiency of manual detection and manual supplementary grinding. For situations where the yield of a single-pass yield from programmed grinding is low, the significance of this invention is obvious. Compared to programmed grinding, the grinding time for a single spoiler in this invention is uncertain and may vary slightly, but the overall efficiency and yield are significantly higher than programmed grinding with fixed parameters. Attached Figure Description

[0047] Figure 1 This is a flowchart of an automatic spoiler polishing method based on image recognition;

[0048] Figure 2 This is a schematic diagram of an automatic spoiler polishing system based on image recognition.

[0049] Figure 3 This is a partial schematic diagram of the robotic arm and the grinding head;

[0050] Figure 4 This is a schematic diagram of the gas path control principle.

[0051] The markings in the image are as follows:

[0052] 1. Robotic arm; 2. Two-way grinding head; 3. Speed ​​control device; 4. System control module; 5. Image sensor; 6. Pressure sensor;

[0053] First grinding head A1, second grinding head B2;

[0054] First pressure regulating valve Q1, second pressure regulating valve Q2, third pressure regulating valve Q3, first solenoid directional valve V1, second solenoid directional valve V2, third solenoid directional valve V3, separator V4, fifth solenoid directional valve V5, air inlet Z1;

[0055] The first channel A of the fifth solenoid directional valve and the second channel B of the fifth solenoid directional valve. Detailed Implementation

[0056] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and examples. The following examples are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0057] Example 1, such as Figure 1 As shown, an automatic spoiler polishing method based on image recognition utilizes real-time grayscale images of the area near the spoiler to be polished obtained by an image sensor. Image analysis is performed based on the grayscale distribution to determine polishing parameters for the area to be polished, and then the polishing operation is executed according to these parameters. The method includes the following polishing steps:

[0058] S1. Acquire and analyze images, determine grinding parameters: The robotic arm drives the camera to be aligned with the point to be ground, acquires and analyzes the image of the point to be ground, and determines the grinding parameters of the point to be ground.

[0059] S2. Switch the grinding head according to the grinding parameters, and drive the robotic arm to move the grinding head close to the point to be ground. The initial coordinates of the point to be ground are preset according to the type of spoiler.

[0060] S3. Drive the grinding head to press against the point to be ground based on the difference between the pressure sensor sampling value and the preset grinding force;

[0061] S4. Issue a speed control command to drive the speed control device to rotate the grinding head.

[0062] S5. Perform polishing according to the set or calculated polishing time.

[0063] The image recognition-based automatic spoiler polishing method described in this invention is called an adaptive fully automatic polishing method, which corresponds to fixed-parameter automatic program polishing. The point to be polished is generally the central target point that drives the polishing head to move when the robotic arm performs the polishing operation, and it is also the center point of the image area. A certain deviation is allowed, such as ±1mm, and the error depends on the installation error of the image sensor and the robotic arm. In fact, the polishing operation is also to polish a local area with the target point as the center. It is impossible to polish only one point. Therefore, it is sometimes described as the area to be polished.

[0064] Regarding the points to be polished: Regardless of whether it is a new or mature model of spoiler, the same model of spoiler is a product of the automated production line with the same mold. The points and features to be polished are consistent. The points and sequence to be polished are determined in advance based on the actual situation after the spoiler is formed. Even if the points initially determined for a new type of spoiler are not ideal, the automatic polishing process of this invention can be used to analyze the images after each point has been polished, add or delete polishing points, thereby fixing the polishing points and sequence. This is beneficial for both fixed parameter automatic program polishing and adaptive fully automatic polishing.

[0065] For single-head grinding systems, the selection of grinding heads is not required; they can be applied directly. For ease of explanation, this invention is uniformly described as the grinding head selection and / or grinding head switching process.

[0066] Preferably, the polishing parameters include polishing time, polishing force, polishing speed, and polishing head selection. For a single polishing head system, the polishing head selected is a fixed polishing head. The process of acquiring and analyzing images and determining polishing parameters includes:

[0067] S11. Statistically analyze the grayscale distribution characteristics of the image within the target area. The target area includes a central region with a fixed length and width centered on the image center point. The target area can be determined using a gate of fixed size. The point to be polished is generally the image center point or near the image center point. In this embodiment, for a 1024*1024 pixel image, a 512*512 pixel region centered on the image center point is taken. Alternatively, a 256*256 pixel region can be taken. The size of this region can be determined based on the total resolution and the actual range of the captured image to reflect the point to be polished. The local grayscale features (also known as target points) should be considered. If they are too large, they will reduce the feature representation ability of the statistical parameters. If they are too small, they will not be able to fully reflect the grayscale change features of the target points. The grayscale variance of all pixels in the target area should be calculated, or the grayscale difference of adjacent pixels in rows and columns should be calculated first, and then the variance of the grayscale difference of the two-dimensional row and column distribution should be calculated. When calculating the variance, the mean is usually calculated at the same time. The mean of different objects represents different physical meanings. For the straight part of the spoiler, the grayscale variance is used as the benchmark, and for the arc part, the variance of the grayscale difference is used as the comparison benchmark.

[0068] S12. Determine grinding parameters: This includes a method combining parameter grading and calculation. Any two of the three parameters—grinding time, grinding force, and grinding speed—are set to several grades. Based on the characteristics of the point to be ground and established principles, the graded grinding parameters for the point to be ground are first determined. The remaining parameters are calculated based on the correlation between grinding energy consumption and grinding effect, combined with image grayscale analysis results and the determined parameters. Since there are countless combinations of grinding head selection, grinding time, grinding force, and grinding speed to achieve the final grinding effect of a single point to be ground, dynamic optimization is a common research direction in the field of control. However, for the practical engineering application of this invention, the research on complex optimization schemes is not an urgent need. What is more important now is to have a reasonable and simple practical scheme. Therefore, it is simple and practical to pre-determine some parameters based on experience and then calculate individual parameters.

[0069] Preferably, the method combining parameter grading and calculation includes:

[0070] S121. Determine the gear values: For a dual-grinding head system, first select the grinding head, define several fixed values ​​for the grinding force between the minimum and maximum values, define several fixed values ​​for the grinding speed between the minimum and maximum values, and determine the corresponding gear selection according to the characteristics of the point to be ground, the gray scale variance, or the gray scale difference variance.

[0071] This embodiment describes the method for selecting grinding heads: Firstly, the grinding head should be selected based on the characteristics of the location. Specifically, for prominent protrusions such as the gate area, a coarse grinding head should be used for rapid rough grinding, followed by a fine grinding head at medium or slow speed for finishing. Generally, straight sections of the spoiler should be quickly finished with a fine grinding head, while other sections should be finished with a fine grinding head at medium or slow speed. Secondly, the statistical characteristics of the image should be considered. When the measured variance is less than a first threshold, the grinding at that location meets the requirements. When the measured variance is greater than or equal to the first threshold but less than a second threshold, the location should be finished with a fine grinding head. When the measured variance is greater than or equal to the second threshold, the location should first be quickly rough ground with a coarse grinding head, followed by a fine grinding head at medium or slow speed for finishing. For single-grinding-head cases, this grinding head selection process is ignored.

[0072] This embodiment describes the method for selecting grinding force and grinding speed: Three fixed grinding forces are defined as 20N, 10N, and 5N, and three grinding speeds as 200r / m, 100r / m, and 50r / m, respectively, to calculate the required grinding time. The method for selecting the three grinding forces and speeds is as follows: 20N and 200r / m are selected for the coarse grinding head; 10N and 100r / m are selected for the straight sections of the spoiler; 10N and 50r / m are selected for the curved arc sections with a small curvature (generally less than 1); and 5N and 50r / m are selected for the curved arc sections with a large curvature (generally greater than or equal to 1).

[0073] Based on the selection of the grinding head, grinding force F, and grinding speed V, calculate the grinding time t.

[0074] S122. Calculate polishing time:

[0075] Let H1 be the first grayscale variance threshold and H2 be the second grayscale variance threshold, D1 be the first grayscale difference variance threshold and D2 be the second grayscale difference variance threshold, and let H be the grayscale variance and D be the grayscale difference variance of the target area pixels in real time. If the graded polishing force determined in step S121 is F and the polishing speed is V, then when using a fine polishing head, the polishing time is...

[0076] t=η1H / FV (Formula 1)

[0077] or

[0078] t = η²D / FV (Formula 2)

[0079] Wherein, η1 is the grayscale efficiency parameter of the fine grinding head, that is, the product of grinding force, grinding speed and grinding time required to achieve the desired grinding effect under unit grayscale variance; η2 is the grayscale difference efficiency parameter of the fine grinding head, that is, the product of grinding force, grinding speed and grinding time required to achieve the desired grinding effect under unit grayscale difference variance. Both parameters can be obtained through simple experiments. That is, for a certain grayscale variance or grayscale difference variance, fix the grinding force and grinding speed, and measure the actual grinding time to obtain the corresponding efficiency parameter. Generally, at least 10 experiments are conducted, and the final statistical mean is used as the empirical efficiency parameter.

[0080] When using a coarse grinding head, the grinding time is...

[0081] t=η3(H-H0) / FV (Formula 3)

[0082] or

[0083] t=η4(D-D0) / FV (Formula 4)

[0084] Wherein, η3 is the coarse grinding head grayscale efficiency parameter, which is the product of grinding force, grinding speed, and grinding time required to achieve the desired grinding effect under unit grayscale variance; η4 is the coarse grinding head grayscale difference efficiency parameter, which is the product of grinding force, grinding speed, and grinding time required to achieve the desired grinding effect under unit grayscale difference variance; H0 is the grayscale variance threshold to prevent over-grinding, and D0 is the grayscale difference variance threshold to prevent over-grinding. η3 and η4 can also be obtained through simple experiments to obtain empirical values ​​of these parameters. That is, for a certain grayscale variance or grayscale difference variance, fix the grinding force and grinding speed, and measure the actual grinding time to obtain the corresponding efficiency parameters. Generally, at least 10 experiments are conducted, and the final statistical mean is used as the empirical efficiency parameter; H0 ranges from 0 to 10, and is 5 in this embodiment; D0 ranges from 0 to 5, and is 2 in this embodiment.

[0085] In this embodiment, for the dual-grinding head system, after the coarse grinding head is used to complete the grinding, image detection and analysis can be performed again to obtain new measured grayscale variance or grayscale difference variance. Then, the grinding time parameter of the fine grinding head can be calculated according to Formula 1 or Formula 2. Alternatively, when using program grinding, the grinding time parameter of the fine grinding head can be calculated directly according to Formula 1 or Formula 2, and high grinding force and fast speed can be selected first.

[0086] Preferably, the switching of the grinding head includes determining the selected grinding head according to the grinding parameters in step S1, controlling the robotic arm to rotate so that the selected grinding head faces the point to be ground.

[0087] Preferably, the step of driving the grinding head to approach the grinding point based on the difference between the pressure sensor sampling value and the preset grinding force includes: when the pressure sensor sampling value is less than the lower limit of the preset value, the robotic arm drives the grinding head to move towards the grinding point, making the grinding head closer to the grinding point; when the pressure sensor sampling value is greater than the upper limit of the preset value, the robotic arm drives the grinding head to move away from the grinding point. Note that this type of movement is generally a small distance movement. The lower limit of the preset value is the preset grinding force minus the allowable error, and the upper limit of the preset value is the preset grinding force plus the allowable error. In this embodiment, the allowable error is 1N. When the preset grinding force is 10N, the lower limit of the preset value is 9N, and the upper limit of the preset value is 11N. The pressure sensor is also referred to as a force-feed sensor or force-feed system in engineering.

[0088] Preferably, the steps further include: firstly confirming and adjusting the ambient light intensity to meet the illuminance requirements for spoiler photography before image capture; the confirmation and adjustment of the ambient light intensity to meet the illuminance requirements for spoiler photography includes collecting the illuminance measured in real time by the illuminance sensor, and if it is lower than the minimum threshold value, the system controls the light source to brighten until the illuminance requirements are met.

[0089] Since artificial lighting generally does not exceed the maximum illuminance threshold, when the measured illuminance is higher than the maximum threshold, it indicates that the sunlight outside is too strong during the day, triggering a system alarm and requiring manual intervention, such as closing some doors and windows. In this embodiment, the minimum threshold is set to 200 lx (lux), and the maximum threshold is set to 500 lx. That is, under normal circumstances, the indoor ambient illuminance is controlled below 500 lx. On sunny days, the requirement can be met by opening and closing curtains, or by designing the size and orientation of the factory windows to ensure that the maximum threshold is never exceeded. On cloudy days or at night, active lighting can be used to achieve this.

[0090] The illuminance confirmation process is generally an independent and autonomous process used to monitor changes in the brightness of the light source. When the brightness change exceeds the threshold, the system control module issues a light adjustment command to adjust the light brightness and ensure that the brightness of the light source remains constant. This process is completed automatically after the system is turned on and does not require any software intervention.

[0091] This invention uses local grayscale variance or grayscale difference variance as the basis for determining grinding parameters. Its fundamental premise is that the spoiler emerging from the mold is generally smooth, although some local points or line segments may exhibit steps or roughness, meaning there are significant differences in grayscale distribution within a local area. If roughness is prevalent throughout the entire square area of ​​the image recognition, the method of this invention is not very practical, as actual spoilers generally do not exhibit such phenomena.

[0092] Example 2 differs from Example 1 in that the order of steps S3 and S4 is reversed.

[0093] Example 3 differs from Example 1 in that the method for obtaining grinding parameters further includes grinding confirmation;

[0094] The polishing confirmation process includes, after the previous polishing is completed, the robotic arm drives the camera to re-align with the polished point, re-acquire the feature image, and analyze and confirm it. If the requirements are met, the process proceeds to the next point to be polished, and the next polishing process begins. Otherwise, the polishing parameters are recalculated, and the polishing is repeated, in a loop, until the requirements are met or the program exits due to limitations.

[0095] This method is generally used when obtaining empirical grinding parameters for a fixed grinding procedure on a spoiler of the same model. It can also be used for new spoiler models, when grinding parameters differ significantly between spoilers of the same model, or when higher surface finish is required.

[0096] Example 4: A method for determining fixed-program grinding parameters for a spoiler, also known as empirical grinding parameter determination, characterized in that: after the previous grinding is completed, grinding confirmation is performed, and all combined grinding parameters during the grinding process are recorded; if a single grinding meets the requirements, then the single grinding parameter is the fixed-program grinding parameter for that point; if two or more grindings are performed, parameter optimization is required, and the parameter optimization method includes:

[0097] The grinding head switching strategy remains unchanged. Different grinding heads are used, and for the combined grinding parameters recorded at the aforementioned points, the maximum grinding time (Tmax) and the maximum grinding speed (Vmax) are taken. Then the grinding force...

[0098] F=(F1*t1*V1+F2*t2*V2+…+Fn* tn*Vn) / ( Tmax*Vmax);

[0099] When the calculated grinding force exceeds the threshold Ft, the threshold value Ft shall prevail. In this case, the grinding time needs to be extended appropriately. The new grinding time is as follows:

[0100] t=( F1*t1*V1+F2*t2*V2+…+Fn* tn*Vn) / (Ft*Vmax);

[0101] The basic idea behind this optimization scheme is to ensure that the effective energy consumption of multiple polishing processes is consistent with the effective energy consumption of a single polishing process.

[0102] For routine product polishing, the robotic arm's movement sequence is pre-set based on the spoiler type or structural parameters. The camera is aimed at the point to be polished, and the polishing parameters for that point are calculated before the polishing operation is performed. If it is necessary to confirm the polishing effect of that point, the robotic arm is usually driven back to re-photograph the polished point and perform image processing and analysis. Based on the analysis results, if the results are not satisfactory, new polishing parameters are determined and the polishing is repeated. After confirming that the requirements are met, the parameters of that point are recorded for subsequent empirical parameter determination.

[0103] Grinding parameters were determined for multiple spoilers of the same model, and multiple sets of grinding experience parameters were obtained. The average of these experience parameters was then statistically averaged, and the average value was used as the final grinding experience parameter for this type of spoiler.

[0104] The above embodiments are generally applied to a novel spoiler grinding system, namely an automatic spoiler grinding system based on image recognition, such as... Figures 2-4As shown, the system includes a robotic arm 1, a grinding head 2, a speed control device 3, a system control module 4, an image sensor 5, and a pressure sensor 6. The grinding head 2 is mounted on the robotic arm 1 and can be either a unidirectional or bidirectional grinding head; the illustration only shows a bidirectional grinding head. The speed control device 3 includes either a pneumatic control device or a motor control device, both of which are existing technologies. The system control module 4 is electrically connected to the speed control device 3, the image sensor 5, and the pressure sensor 6, respectively performing functions such as grinding head speed control, image sensor signal acquisition and analysis, and pressure sensor information acquisition and control of the contact degree between the grinding head and the spoiler. The system reacts to contact pressure. The image sensor 5 is mounted on the robotic arm and electrically connected to the system control module 4. It is used to capture images of local areas of the spoiler to be polished or already polished. The pressure sensor 6 is mounted on the robotic arm or the polishing head to sense the contact pressure between the polishing head and the spoiler. The system is designed with an image sensor to dynamically determine polishing parameters and work with the pressure sensor to complete fully automatic polishing. It can also achieve fully automatic acquisition of fixed program polishing parameters. The pressure sensor is mounted on the non-rotating part of the axial connection on the back of the sandpaper on the polishing head and mainly senses axial pressure. The image sensor includes a camera with a pixel count of not less than 1024*1024.

[0105] The image sensor is generally a telephoto camera. Its field of view is adjusted to clearly capture the features of the entire surface of the spoiler to be polished. Generally, the central area of ​​the image is a range of 0.1 to 1 square centimeters, with the point to be polished as the center point. The central area occupies 1 / 16 to 15 / 16 of the length and width of the entire image. This is generally achieved by adjusting the camera's installation position, setting the focal length, and adjusting the magnification. In this embodiment, the central area occupies about 1 / 2 of the length and width of the entire image. During image processing, a fixed gate is used to process image features of a fixed size and fixed area. Image analysis is performed after each single image is acquired.

[0106] The automatic polishing system also includes an illuminance sensor and a light source (not shown in the figure). The illuminance sensor is installed in a fixed position in the polishing workshop, with a fixed relative position to the light source. It is electrically connected to the system control module and is used to monitor changes in the brightness of the light source. When the brightness change exceeds a threshold, the system control module issues a light adjustment command to adjust the light brightness and ensure that the brightness of the light source remains constant. The illuminance sensor is generally installed near the spoiler and does not affect the movement of the robotic arm during the polishing process. The illuminance sensor is set up to ensure that the ambient light level of the spoiler remains basically constant. This way, the grayscale value change of the image of the polished area captured by the image sensor is within a reasonable range, and the image recognition and analysis of the polishing effect will not be affected by ambient light. When the ambient natural light level meets the requirements during the day, the light source can be turned off.

[0107] The illuminance sensor and the light source are synchronously installed on the robotic arm with a fixed relative position. The light source is used to illuminate the local area of ​​the spoiler that is to be polished or has been polished.

[0108] The system speed control device adopts an air circuit control device, including a one-way large grinding head circuit control device or a two-way grinding head circuit control device. The two-way grinding head includes a first grinding head A1 and a second grinding head B2. The dual-grinding head automatic grinding system includes two image sensors, which are installed and photographed in the grinding direction of the two grinding heads respectively.

[0109] The system control module sends air path control commands to the air path control device, which executes the commands to control the grinding head speed. For bidirectional grinding heads, it controls the grinding speed of the first and second grinding heads respectively; for unidirectional grinding heads, it only controls that unidirectional grinding head. The air path control device includes a separator, a pressure regulating channel valve, a pressure regulating channel solenoid directional valve, and an output channel solenoid directional valve to achieve grinding speed regulation. The pressure regulating channel valve and the pressure regulating channel solenoid directional valve correspond one-to-one to form a pressure regulating combination. The system control module realizes the motion control of the robotic arm and the control of the air path control device.

[0110] The pressure regulating channel includes a first pressure regulating valve Q1, a second pressure regulating valve Q2, and a third pressure regulating valve Q3. The pressure regulating channel electromagnetic reversing valve includes a first electromagnetic reversing valve V1, a second electromagnetic reversing valve V2, and a third electromagnetic reversing valve V3. The output channel electromagnetic reversing valve includes a fifth electromagnetic reversing valve V5. The intake air passes through the separator V4 and is connected to the first pressure regulating valve Q1, the second pressure regulating valve Q2, and the third pressure regulating valve Q3 through the air path. Then, each of these valves is connected to the first electromagnetic reversing valve V1, the second electromagnetic reversing valve V2, and the third electromagnetic reversing valve V3, and then simultaneously connected to the fifth electromagnetic reversing valve V5. The first channel A and the second channel B of the fifth electromagnetic reversing valve V5 are respectively connected to the first grinding head A1 and the second grinding head B2.

[0111] The first electromagnetic reversing valve V1, the second electromagnetic reversing valve V2, and the third electromagnetic reversing valve V3 are all 2-position 3-way electromagnetic reversing valves, and the fifth electromagnetic reversing valve V5 is a 3-position 5-way electromagnetic reversing valve.

[0112] The first grinding head A1 is fitted with coarse sandpaper, and the second grinding head B2 is fitted with fine sandpaper.

[0113] Another image recognition-based automatic spoiler grinding system uses a motor control device for speed control, including at least one adjustable speed motor. The adjustable speed motor is connected to the grinding head through a transmission device, and the speed of the adjustable speed motor is controlled by the system control module.

[0114] The motor control device includes one adjustable speed motor, which is connected to the first grinding head and the second grinding head respectively through a transmission switching device; or the motor control device includes two adjustable speed motors, which are connected to the first grinding head and the second grinding head respectively through a transmission device; when the motor control module is equipped with two adjustable speed motors, the two adjustable speed motors are each connected to one of the grinding heads in the bidirectional grinding head through separate transmission devices, and the two motors are controlled separately by the system control module, and the adjustable speed motors are equipped with several fixed adjustable speeds.

[0115] The basic principle of this invention is as follows: By analyzing the grayscale image of the area near the spoiler to be polished, which is acquired in real time by an image sensor, the grayscale distribution characteristics are obtained. Based on these characteristics, combined with the features of each point after the spoiler is demolded, and empirical polishing strategies, the combination of polishing parameters such as the selection of polishing head, polishing time, polishing force, and polishing speed for each point to be polished is automatically determined. The basic principle for determining the polishing parameters is that if each polishing point is not smooth enough, the grayscale variance or grayscale difference variance of that area will exceed a preset threshold. The greater the exceedance, the more obvious the step is, and coarse polishing may be required. Furthermore, at least one of the polishing time, polishing force, and polishing speed should be relatively large. Since increasing any one or more of the grinding time, grinding force, and grinding speed can complete the grinding of spoilers with obvious or large steps, there are infinitely many combinations. Considering that each parameter has its effective range, and that engineering experience generally adopts a graded design for grinding force and grinding speed, this invention follows this design and designs an automatic graded principle. This simplifies the multi-parameter optimization process into a single-parameter determination process, thereby realizing a more practical automatic determination of grinding parameters and achieving fully automatic grinding of spoilers.

[0116] The above descriptions are merely a few relatively systematic and comprehensive embodiments of the automatic spoiler grinding method based on image recognition of the present invention. The main method is to classify grinding force and grinding speed, and then calculate the grinding time. In fact, any two of the grinding time, grinding force, and grinding speed can be classified, and the remaining one can be calculated. These combinations should also be considered within the scope of protection of the present invention, and will not be listed one by one here.

Claims

1. An automatic spoiler polishing method based on image recognition, characterized in that, The polishing process includes the following steps: S1. Acquire and analyze images and determine polishing parameters: The robotic arm drives the camera to be aligned with the point to be polished. The image sensor acquires grayscale images of the area near the point to be polished in real time. Based on the grayscale distribution, the image is analyzed to determine the polishing parameters of the point to be polished. The polishing parameters include polishing time, polishing force, polishing speed, and polishing head selection. S11. Statistically analyze the grayscale distribution characteristics of the image within the target area, and calculate the grayscale variance of all pixels within the target area, or calculate the grayscale difference for adjacent pixels in rows and columns, and then calculate the variance of the grayscale difference. The target area includes a region with fixed length and width with the target point to be polished or the center point of the image as the origin. For the straight part of the spoiler, the grayscale variance is used as the benchmark, and for the arc part, the variance of the grayscale difference value is used as the comparison benchmark. S12. Determine the grinding parameters: This includes a method that combines parameter grading and calculation. Any two of the three parameters, namely grinding time, grinding force, and grinding speed, are set to several grades. Based on the characteristics of the point to be ground and the established principles, the graded grinding parameters of the point to be ground are determined first. The remaining parameters are calculated and determined based on the correlation between grinding energy consumption and grinding effect, combined with the image grayscale analysis results and the already determined parameters. The method combining parameter grading and calculation includes: S121. Determine the gear values: For a dual-grinding head system, first select the grinding head, define several fixed values ​​for the grinding force between the minimum and maximum values, define several fixed values ​​for the grinding speed between the minimum and maximum values, and determine the corresponding gear selection based on the characteristics of the point to be ground, the gray scale variance, or the gray scale difference variance. S122. Calculate polishing time: Let H1 be the first grayscale variance threshold and H2 be the second grayscale variance threshold, D1 be the first grayscale difference variance threshold and D2 be the second grayscale difference variance threshold, and let H be the grayscale variance and D be the grayscale difference variance of the target area pixels in real time. If the graded polishing force determined in step S121 is F and the polishing speed is V, then when using a fine polishing head, the polishing time is... t=η1H / FV (Formula 1) or t = η²D / FV (Formula 2) Wherein, η1 is the grayscale efficiency parameter of the fine grinding head, that is, the product of grinding force, grinding speed and grinding time required to achieve the desired grinding effect under unit grayscale variance; η2 is the grayscale difference efficiency parameter of the fine grinding head, that is, the product of grinding force, grinding speed and grinding time required to achieve the desired grinding effect under unit grayscale difference variance. When using a coarse grinding head, the grinding time is... t=η3(H-H0) / FV (Formula 3) or t=η4(D-D0) / FV (Formula 4) Wherein, η3 is the grayscale efficiency parameter of the coarse grinding head, which is the product of grinding force, grinding speed and grinding time required to achieve the desired grinding effect under unit grayscale variance; η4 is the grayscale difference efficiency parameter of the coarse grinding head, which is the product of grinding force, grinding speed and grinding time required to achieve the desired grinding effect under unit grayscale difference variance; H0 is the grayscale variance threshold to prevent over-grinding, and D0 is the grayscale difference variance threshold to prevent over-grinding. S2. Switch the grinding head according to the grinding parameters, and drive the robotic arm to move the grinding head closer to the point to be ground; S3. Drive the grinding head to press against the point to be ground based on the difference between the pressure sensor sampling value and the preset grinding force; S4. Issue a speed control command to drive the speed control device to rotate the grinding head. S5. Perform polishing according to the set or calculated polishing time.

2. The automatic spoiler polishing method based on image recognition according to claim 1, characterized in that, The execution order of steps S3 and S4 is interchanged.

3. The automatic spoiler polishing method based on image recognition according to claim 1, characterized in that, This also includes polishing confirmation; The polishing confirmation process includes, after the previous polishing is completed, the robotic arm drives the camera to re-align with the polished point, re-acquire the feature image, and analyze and confirm it. If the requirements are met, the process proceeds to the next point to be polished for shooting, analysis, and polishing. Otherwise, the polishing parameters are recalculated, and polishing is performed again, repeating the process until the requirements are met or the program exits.

4. The automatic polishing method for spoilers based on image recognition according to claim 1, characterized in that, The switching of the grinding head includes determining the selected grinding head according to the grinding parameters in step S1, controlling the robotic arm to rotate so that the selected grinding head faces the point to be ground.

5. The automatic polishing method for spoilers based on image recognition according to claim 1, characterized in that, The step of driving the grinding head to move closer to the grinding point based on the difference between the pressure sensor sampling value and the preset grinding force includes the following: when the pressure sensor sampling value is less than the lower limit of the preset value, the robotic arm drives the grinding head to move towards the grinding point so that the grinding head is closer to the grinding point; when the pressure sensor sampling value is greater than the upper limit of the preset value, the robotic arm drives the grinding head to move away from the grinding point.

6. The automatic polishing method for spoilers based on image recognition according to claim 1, characterized in that, The steps also include first confirming and adjusting the ambient light intensity to meet the illuminance requirements for spoiler shooting before image capture. When the illuminance measured by the illuminance sensor in real time is lower than the minimum threshold value, the system controls the light source to be brightened until the illuminance requirements are met.

7. The automatic spoiler polishing method based on image recognition according to claim 1, characterized in that, After the previous polishing is completed, a polishing confirmation is performed, and all combined polishing parameters during the polishing process are recorded. If a single polishing operation meets the requirements, the parameters for that single polishing operation are the fixed program polishing parameters for that location. If two or more polishing operations are performed, parameter optimization is required. The parameter optimization methods include: The grinding head switching strategy remains unchanged. Different grinding heads are used, and for the combined grinding parameters recorded at the aforementioned points, the maximum grinding time (Tmax) and the maximum grinding speed (Vmax) are taken. Then the grinding force... F=(F1*t1*V1+F2*t2*V2+…+Fn* tn*Vn) / ( Tmax*Vmax); When the calculated grinding force exceeds the threshold Ft, the threshold value Ft shall prevail. In this case, the grinding time needs to be extended appropriately. The new grinding time is as follows: t=( F1*t1*V1+F2*t2*V2+…+Fn* tn*Vn) / (Ft*Vmax).

8. The automatic polishing method for spoilers based on image recognition according to claim 7, characterized in that, Grinding parameters were determined for multiple spoilers of the same model, and multiple sets of grinding experience parameters were obtained. The average of these experience parameters was then statistically averaged, and the average value was used as the final grinding experience parameter for that model of spoiler.

Images