Encoder trigger intelligent detection method and system for automobile painting line

Abstract

The application discloses an encoder trigger intelligent detection method and system for an automobile coating line, and belongs to the technical field of automobile manufacturing quality detection. The method comprises the following steps: generating a mixed trigger signal through an encoder and a photoelectric sensor; predicting the current position and speed of the vehicle body by using a Kalman filtering algorithm, and determining the expected trigger position; calculating the position deviation between the predicted position and the measured position, generating a position compensation amount to correct the expected trigger position; controlling the line scan camera to shoot based on the corrected trigger position; and dynamically adjusting the trigger window threshold based on the trigger error and feeding back the optimization. Through the deep coupling and closed-loop cooperation of the three modules of mixed triggering, double-loop control and adaptive optimization, the shooting position accuracy is kept within ±1mm under the condition of ±10% line speed fluctuation, the defect detection rate is improved to 94.3%, and the false detection rate is reduced to 1.8%.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automobile manufacturing quality detection, in particular to an encoder trigger intelligent detection method and system for automobile painting lines, which is particularly suitable for solving the problem of high-precision defect detection position control under the condition of production line speed fluctuation. BACKGROUND

[0002] In the automobile manufacturing process, painting quality is a key factor affecting the appearance quality and corrosion resistance of the whole vehicle. The painting line defect detection system needs to accurately control the camera trigger timing during the continuous movement of the vehicle body, ensuring that each shot covers the predetermined detection area, thereby achieving complete and accurate defect identification. Traditional detection systems mostly use encoders or sensors as a single trigger source to control camera shooting by measuring vehicle body displacement or detecting specific markers.

[0003] In the prior art, for example, Chinese patent application CN119985492A discloses a vehicle surface defect detection method. When the vehicle to be detected enters the tunnel detection space, the encoder detects the forward distance of the vehicle to be detected, controls multiple cameras to shoot the surface image of the vehicle to be detected based on the forward distance, detects the two-dimensional defect information of the defects in the surface image, and converts the two-dimensional defect position information into three-dimensional defect position information in the vehicle coordinate system. This method triggers the camera shooting by directly measuring the forward distance of the vehicle with the encoder, achieving a certain degree of detection automation.

[0004] However, the above prior art still has the following shortcomings:

[0005] First, in actual production line applications, the running speed of the painting line often fluctuates by ±5% to ±15%. This speed fluctuation will cause the shooting position of the detection system based on fixed distance triggering to deviate. The single encoder triggering method used in the prior art cannot predict and compensate for the position error caused by line speed fluctuation, resulting in poor consistency of the detection position of adjacent vehicle bodies and reduced comparability of cross-body data.

[0006] Second, the encoder will produce cumulative errors during long-term operation, including pulse loss caused by mechanical wear, counting drift caused by temperature changes, and miscounting caused by electromagnetic interference. The prior art lacks real-time monitoring and compensation mechanism for the cumulative error of the encoder. As the detection time prolongs, the position measurement accuracy gradually decreases, and the triggering position error can reach more than 5mm, far exceeding the ±1mm precision standard required by the painting defect detection.

[0007] Third, the production line will have obvious speed jitter during start-up and stoppage, with a jitter amplitude of ±20% and a duration of 2-5s. The fixed trigger window adopted in the prior art cannot adapt to this transient operating condition, and the phenomenon of missed shooting or false triggering often occurs during start-up and stoppage, affecting the detection coverage and system reliability.

[0008] Fourth, the trigger control strategy of the prior art is open-loop control, i.e., direct triggering according to a preset distance threshold, and lacks a closed-loop feedback optimization mechanism based on actual trigger effects. This open-loop control method cannot dynamically adjust the trigger parameters according to the actual operating conditions, and is difficult to adapt to the detection requirements under different vehicle models, different line speeds and different operating conditions, and the system lacks self-adaptability and robustness.

[0009] Therefore, there is an urgent need for a new encoder trigger intelligent detection method that can effectively cope with production line speed fluctuations, compensate for encoder cumulative errors in real time, adaptively adjust trigger parameters, ensure detection coverage, and significantly improve the consistency of detection positions across vehicle bodies and the comparability of detection data, thereby improving the detection rate of coating defects and the overall performance of the detection system. SUMMARY

[0010] The purpose of the present application is to provide an encoder trigger intelligent detection method and system for an automobile coating line, aiming to solve the technical problems in the prior art such as unstable detection positions caused by production line speed fluctuations, inability to compensate for encoder cumulative errors, unstable triggering during start-up and stoppage, and lack of closed-loop optimization mechanism, and to realize high-precision, high-consistency and high-adaptability coating line defect detection.

[0011] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0012] The present application provides an encoder trigger intelligent detection method for an automobile coating line, which realizes precise tracking of vehicle body movement and intelligent control of trigger timing through deep coupling and closed-loop cooperation of three core modules: hybrid trigger signal generation, double-loop position control and adaptive window optimization. The method first generates a hybrid trigger signal by fusing encoders and photoelectric sensors to ensure the reliability and accuracy of the trigger source. Then, a position prediction loop is constructed using the Kalman filter algorithm to optimally estimate the current position and speed of the vehicle body, and the expected trigger position is calculated based on this. Meanwhile, a feedback compensation loop is constructed, which compares the predicted position with the measured position in real time, generates a position compensation amount to correct the expected trigger position, and eliminates the effects of encoder cumulative errors and line speed fluctuations. Finally, through the adaptive window optimization module, the trigger window threshold is dynamically adjusted based on historical trigger errors, forming a complete closed-loop feedback, and realizing continuous optimization of trigger precision.

[0013] The present application has the following advantages:

[0014] First, by constructing a hybrid triggering mechanism of encoder and photoelectric sensor, the advantages of the two sensors are complementary and mutual verification. The encoder provides continuous and high-precision displacement measurement, and the photoelectric sensor provides discrete and reliable position confirmation. The fusion of the two significantly improves the reliability of the trigger signal. Experimental data show that, compared with single encoder triggering mode, the hybrid triggering mechanism reduces the trigger signal loss rate from 0.3% to 0.01%, and the reliability is improved by 30 times.

[0015] Second, the position prediction-feedback compensation double-loop control algorithm is innovatively proposed, realizing the deep coupling and synergistic effect of the prediction loop and the compensation loop. The position prediction loop uses Kalman filter algorithm to optimally estimate the vehicle body motion state, which can predict the time when the vehicle body reaches the target position, providing forward-looking information for trigger control. The feedback compensation loop dynamically generates a compensation amount to correct the expected trigger position by monitoring the deviation between the predicted position and the measured position in real time, effectively eliminating the influence of encoder cumulative error and line speed fluctuation. The synergistic effect of the double-loop control makes the system maintain the shooting position accuracy within ±1mm under the condition of ±10% line speed fluctuation, which is 5 times higher than the ±5mm accuracy of the existing technology.

[0016] Third, the adaptive window optimization module realizes the closed-loop self-optimization of the trigger parameters, forming a complete feedback loop from trigger execution to error analysis to parameter adjustment. Based on the statistical analysis of historical trigger errors, the module dynamically adjusts the trigger window threshold, so that the system can adapt to the detection needs of different vehicle models, different line speeds and different working conditions. Experimental results show that, in the start-stop stage, the adaptive window optimization function reduces the missed shooting rate from 12% to 0.5%, and in the normal running stage, the standard deviation of the detection position is reduced from 2.3mm to 0.8mm, and the consistency of the cross-vehicle body detection data is significantly improved.

[0017] Fourth, the three core modules form a deeply coupled closed-loop synergistic system, realizing the nonlinear synergistic effect of 1+1+1>3. The hybrid trigger signal generation module provides reliable data source for the double-loop position control, the double-loop position control module provides real-time error information for the adaptive window optimization, and the adaptive window optimization module improves the performance of the hybrid trigger through parameter feedback. The three modules promote each other and superimpose synergies, so that the overall performance of the system is much higher than the sum of the independent effects of each module. Application tests show that, compared with the existing technology, the present application improves the detection rate of coating defects from 82.5% to 94.3%, an increase of 11.8 percentage points; the false detection rate is reduced from 5.2% to 1.8%, a decrease of 3.4 percentage points; the overall detection efficiency of the system is improved by 27%, reaching the international leading level. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The figure is a schematic diagram of the overall process of the method.

[0019] Figure 2 A flowchart for generating a mixed trigger signal.

[0020] Figure 3 A flowchart for position prediction and compensation control.

[0021] Figure 4 A flowchart for adaptive window dynamic adjustment.

[0022] Figure 5 A schematic diagram of the overall architecture of the system of the present application. DETAILED DESCRIPTION

[0023] Reference will now be made to the drawings, and specific examples relating to the preferred embodiments of the application will be described in detail. Figures 1-5 It should be understood that the detailed description of the specific embodiments described herein is intended to be illustrative only and is not intended to limit the scope of the present application.

[0024] With reference to Figure 1 The present application provides an encoder trigger intelligent detection method for an automobile painting line. The method is applied to an online detection system of the painting line, and high-precision and high-consistency detection of surface defects of a painted vehicle body is achieved by intelligently controlling a trigger timing of a line scan camera.

[0025] The core of the method of the present application is to construct a complete closed-loop control system through three mutually coupled modules: a mixed trigger signal generation module 1, a double-loop position control module 2, and an adaptive window optimization module 3. Through deep coupling at the parameter level and the state level, these three modules form a complete closed loop of trigger → prediction and compensation → feedback optimization → optimized trigger, and continuously improve the trigger precision.

[0026] With reference to Figure 2 The mixed trigger signal generation module 1 is responsible for fusing the data of the encoder and the photoelectric sensor to generate a reliable and accurate mixed trigger signal, which provides a basic data source for subsequent position prediction and compensation control.

[0027] Specifically, the mixed trigger signal generation includes the following steps:

[0028] First, the position pulse signal of the vehicle body to be detected on the painting line is collected in real time by an incremental photoelectric encoder. The encoder is installed on the driving roller shaft of the painting line and rotates synchronously with the movement of the vehicle body. Preferably, the encoder has a resolution of 2048 PPR or 4096 PPR, and adopts A, B, and Z three-phase quadrature output, wherein the A and B phases provide position incremental information, and the Z phase provides a single-turn zero reference signal. After the pulse signal output by the encoder is processed by a four-times frequency circuit, the resolution is improved to 8192 or 16384 pulses per revolution. According to the diameter D (preferably 200-300 mm) of the driving roller shaft, the vehicle body movement distance corresponding to a single pulse can be calculated :

[0029] ,

[0030] wherein, is the driving roller diameter (unit: mm), is the encoder resolution (unit: PPR), 4 represents the four-fold frequency coefficient, is the constant of the circle. In the preferred embodiment, when , , , that is, each pulse corresponds to a body movement of about 0.1 mm, which can meet the high-precision requirements of coating detection.

[0031] The system collects the pulse signals output by the encoder in real time through the high-speed counter module, and the sampling period is 1 ms to 5 ms. In each sampling period, the counter records the pulse count values of the A phase and the B phase, and judges the movement direction of the body according to the phase relationship of the A phase and the B phase. Based on the pulse count values, the instantaneous movement distance of the body to be detected can be calculated.

[0032] Secondly, the arrival confirmation signal of the body to be detected is collected by the photoelectric sensor. The photoelectric sensor is installed at the entrance position of the detection area of the coating line, and uses a transmitting-receiving type or a reflection type photoelectric switch. The detection distance is 0.5 m to 2 m, and the response time is less than 10 ms. When the body to be detected enters the detection area, the body blocks the light path of the photoelectric sensor, triggers the level to flip, and generates an arrival confirmation signal. This signal is used to calibrate the position reference point of the encoder, and eliminates the absolute position uncertainty of the encoder.

[0033] The process of generating a mixed trigger signal based on the position pulse signal and the arrival confirmation signal is as follows:

[0034] The pulse count values output by the encoder within a preset time window (preferably 10 ms to 50 ms) are obtained, and the instantaneous movement distance of the body to be detected is calculated based on the pulse count values and the single-pulse movement distance . It is judged whether the photoelectric sensor detects the body to be detected entering the detection area. If the photoelectric sensor signal has a level jump, it indicates that the body is in place, and the current pulse count value of the encoder is taken as the position reference point , and the cumulative position counter is cleared. Thereafter, based on the position reference point and the instantaneous movement distance, the current cumulative position of the body to be detected is calculated .

[0035] The system pre-sets a plurality of shooting positions (i=1, 2, …, M), which are uniformly distributed on the surface of the body to be detected. The spacing between adjacent shooting positions is determined according to the field of view width of the line scan camera, and is preferably 80 mm to 150 mm. When the current cumulative position of the body Gradually approaching a preset shooting position When the condition is met, the system enters a trigger preparation state.

[0036] Set a preset trigger window threshold , which represents the position tolerance range allowed for triggering. When the distance between the current cumulative position of the vehicle body and the preset shooting position is less than the preset trigger window threshold, i.e., the condition is met:

[0037] ,

[0038] where, is the current cumulative position of the vehicle body to be detected, is the th preset shooting position, is the preset trigger window threshold.

[0039] In the preferred embodiment, the initial value is set to 3-5 mm. When the above condition is met, a hybrid trigger signal is generated, which will be transmitted to the dual-loop position control module 2 for starting the position prediction and compensation calculation process.

[0040] The advantage of the hybrid trigger mechanism is that the encoder provides continuous, high-resolution displacement measurement, which can track the subtle displacement changes of the vehicle body in real time; the photoelectric sensor provides discrete and reliable position confirmation, which can periodically calibrate the cumulative position of the encoder to prevent position drift caused by long-term operation. After the fusion of the two, both the continuity and accuracy of the position measurement are guaranteed, and the accuracy and reliability of the position reference are also guaranteed.

[0041] Referring to Figure 3 , the dual-loop position control module 2 is the core innovative module of the present application, responsible for predicting and compensating the vehicle body position to ensure high precision and high consistency of the camera trigger position. The module contains two tightly coupled control loops: a position prediction loop and a feedback compensation loop.

[0042] Implementation of the position prediction loop:

[0043] The position prediction loop uses the Kalman filter algorithm to optimally estimate the motion state of the vehicle body to be detected, achieving real-time prediction of the vehicle body position and speed. Kalman filtering is a recursive filtering algorithm that can give the optimal estimate of the current state by fusing historical observation data and system model in the presence of measurement noise and system disturbance.

[0044] First, establish the motion state model of the vehicle body to be detected. Assume that the vehicle body moves at a constant speed on the painting line, and its motion state can be described by two state variables: position and speed. Define the state vector as:

[0045] ,

[0046] wherein, is the position state variable (unit: mm) of the vehicle body at time , is the velocity state variable (unit: mm / s) of the vehicle body at time .

[0047] The state transition equation of the vehicle body motion is:

[0048] ,

[0049] wherein, is the state transition matrix, is the process noise vector, obeying Gaussian distribution with mean value of 0 and covariance matrix of .

[0050] The state transition matrix represents the motion relationship of the vehicle body between adjacent two sampling time points, and can be expressed as:

[0051] ,

[0052] wherein, is the sampling period (unit: s), and in the preferred embodiment, the value is 0.01 s to 0.05 s. This matrix shows that the position at time is equal to the position at time plus the product of the velocity and the time interval, i.e. ; the velocity at time remains unchanged, i.e. .

[0053] The observation equation of the encoder for the vehicle body position is:

[0054] ,

[0055] wherein, is the position observation value at time , is the observation matrix, is the observation noise, obeying Gaussian distribution with mean value of 0 and variance of .

[0056] The observation matrix represents the relationship between the observation value and the state variable, and can be expressed as:

[0057] ,

[0058] This matrix shows that the encoder can only observe the position of the vehicle body, and cannot directly observe the velocity.

[0059] Process noise covariance matrix To describe the uncertainty of the system model, including the influence of factors such as coating line speed fluctuation, encoder quantization error, etc. In the preferred embodiment, Set to:

[0060] ,

[0061] Wherein, The position process noise variance is 0.01 to 0.1 ; The speed process noise variance is 1 to 10 .

[0062] Observation noise variance To describe the uncertainty of the encoder measurement, mainly including factors such as encoder resolution limit, counting error, electromagnetic interference, etc. In the preferred embodiment, The value is 0.01 to 0.1 .

[0063] The recursive process of Kalman filter algorithm includes prediction step and update step.

[0064] Prediction step:

[0065] Obtain the state estimation value at the previous time And the error covariance matrix , based on the state transition model, the state prediction is carried out, the prior state estimation at the current time is obtained And the prior error covariance matrix :

[0066] ,

[0067] ,

[0068] Wherein, Indicates the predicted value of the state at time Based on the information at time , the superscript Indicates the covariance matrix of the prediction error, and the superscript Indicates the matrix transpose.

[0069] The current predicted position And the predicted speed At the current time can be extracted from the prior state estimation:

[0070] ,

[0071] ,

[0072] where the subscripts 1 and 2 denote the first and second elements of the state vector, respectively.

[0073] Update step:

[0074] Obtain the position observation of the encoder at the current time Calculate the observation deviation (also known as innovation or residual) between the observation and the prior prediction:

[0075] ,

[0076] Calculate the innovation covariance matrix:

[0077] ,

[0078] Calculate the Kalman gain matrix:

[0079] ,

[0080] where is the Kalman gain matrix, which represents the weight distribution of the prediction value and the observation value in the state update. When the observation noise is small, is large, and the state update relies more on the observation value; when the process noise is small, is small, and the state update relies more on the prediction value.

[0081] Based on the observation deviation and the Kalman gain, the prior state estimate is corrected to obtain the posterior state estimate and the posterior error covariance matrix :

[0082] ,

[0083] ,

[0084] where is the identity matrix.

[0085] From the posterior state estimate, the position estimate and the speed estimate at the current time can be extracted, and these estimates are used as historical data for the next prediction to realize recursive update.

[0086] According to the current predicted position and the predicted speed , the expected trigger position of the camera can be determined. Let the preset shooting position be , the system calculates the time required for the vehicle body to move from the current predicted position to the preset shooting position:

[0087] ,

[0088] wherein, is the th preset shooting position, is the current predicted position, is the predicted speed, is the expected trigger delay.

[0089] Considering the camera's exposure delay and image transmission delay , the system calculates the position at which the camera should send a trigger signal so that the camera exposure center corresponds to the preset shooting position. The expected trigger position of the camera is calculated as:

[0090] ,

[0091] wherein, is the camera exposure time, preferably 50-500 μs; is the image transmission and processing delay, preferably 100 μs-1 ms.

[0092] Implementation of feedback compensation loop:

[0093] Although the Kalman filter algorithm can optimally estimate the vehicle body position, in actual application, due to the cumulative error of the encoder and the random fluctuations of the linear speed, there is still a certain deviation between the predicted position and the actual position. The role of the feedback compensation loop is to monitor this deviation in real time, generate a position compensation amount to correct the expected trigger position, and further improve the trigger accuracy.

[0094] Specifically, the feedback compensation includes the following steps:

[0095] First, the current measured position of the encoder is obtained. The measured position here refers to the current cumulative position calculated by the hybrid trigger signal generation module 1 based on the encoder pulse count and the position reference point.

[0096] Second, the current measured position is sensor-verified. The photoelectric sensor sets multiple verification points in the detection area, and outputs a pulse signal when the vehicle body passes through these verification points. The system compares the verification point position measured by the encoder with the verification point position preset by the photoelectric sensor, and if the deviation between the two exceeds the preset threshold (preferably 2-5 mm), it is considered that the encoder has a large cumulative error and needs to be calibrated. The calibration method is to adjust the current position of the encoder to the verification point position corresponding to the photoelectric sensor, thereby eliminating the cumulative error. The measured position after sensor verification is denoted as .

[0097] Then, the current predicted position instantaneous deviation between the verified measured position and the reference position:

[0098] ,

[0099] Due to the quantization noise of the encoder and electromagnetic interference, the instantaneous deviation can have large high-frequency jitter. To suppress the noise interference, the instantaneous deviation is processed by a sliding window filtering. Let the sliding window length be (preferably 5 to 20), then the filtered position deviation is calculated as:

[0100] ,

[0101] wherein, is the instantaneous deviation at the time, is the sliding window length, is the filtered position deviation.

[0102] Then, based on the filtered position deviation and the predicted speed, a compensation coefficient is determined. The design principle of the compensation coefficient is: the larger the position deviation, the greater the compensation required; the faster the vehicle body speed, the shorter the response time of the compensation, and the compensation coefficient should be appropriately reduced to avoid overcompensation. The compensation coefficient can be expressed as:

[0103] ,

[0104] wherein, is the reference compensation gain, the value range is 0.5 to 2.0, and in the preferred embodiment, the value is 1.0; is the absolute value of the position deviation, is the predicted speed.

[0105] According to the position deviation and the compensation coefficient, the position compensation amount is calculated:

[0106] ,

[0107] To avoid system instability caused by excessive compensation amount, the position compensation amount is amplitude limited. Let the maximum allowed compensation amount be (preferably 2mm to 5mm), then the actual applied position compensation amount is:

[0108] ,

[0109] wherein, is the sign function, is the minimum value function.

[0110] Then, the expected trigger position is corrected by the position compensation amount. First, the current line speed of the painting line is obtained and the line speed fluctuation amplitude . The line speed is calculated based on the encoder data or obtained by an independent speed sensor, and the line speed fluctuation amplitude is obtained by statistical analysis of the standard deviation of the line speed in the recent period.

[0111] The speed correction factor is determined based on the current line speed :

[0112] ,

[0113] wherein, is the current line speed, is the nominal line speed.

[0114] The position compensation amount is multiplied by the speed correction factor to obtain the corrected position compensation amount:

[0115] ,

[0116] The expected trigger position is added to the corrected position compensation amount to obtain the corrected trigger position :

[0117] ,

[0118] To ensure the rationality of the trigger position, it is necessary to determine whether the corrected trigger position is within the allowed trigger range. Let the allowed trigger range be , wherein is the allowed deviation range, preferably 5mm to 10mm. If the corrected trigger position is within the allowed range, output the corrected trigger position as the final trigger position; if the corrected trigger position exceeds the allowed range, output the expected trigger position as the final trigger position to avoid excessive correction leading to abnormal trigger position.

[0119] Finally, based on the final trigger position, the line scan camera is controlled to perform the shooting operation. The system monitors the current cumulative position of the vehicle body, and when the cumulative position reaches or exceeds the final trigger position, a hardware trigger signal is immediately sent to the line scan camera. After receiving the trigger signal, the line scan camera starts the exposure and data acquisition process to obtain the surface coating image of the vehicle body to be detected.

[0120] The line scan camera uses a high-resolution CMOS or CCD line array sensor with a resolution of no less than 8192 pixels and a line frequency of no less than 40kHz, capable of achieving clear imaging under high-speed vehicle body motion conditions. The camera is equipped with a dedicated LED line light source, and the light source wavelength is selected according to the coating color, usually white light or a specific wavelength of monochromatic light (such as red light 630nm, green light 525nm, blue light 470nm).

[0121] The acquired surface coating image contains information such as texture, gloss, color, etc. of the vehicle body surface, which can be used for subsequent defect detection analysis.

[0122] The cooperative mechanism of the double-loop control:

[0123] The position prediction loop and the feedback compensation loop constitute the double-loop control architecture, and the two are deeply coupled to achieve synergistic effect. The output (current predicted position and predicted speed) of the position prediction loop is the key input parameter of the feedback compensation loop, and the predicted speed determines the size of the compensation coefficient, and the predicted position is used to calculate the position deviation. The output (position compensation amount) of the feedback compensation loop corrects the result of the prediction loop, so that the final trigger position is closer to the true target position.

[0124] When the line speed fluctuates, the prediction loop can quickly track the speed change through the Kalman filter algorithm, and update the predicted speed and predicted position in real time; the compensation loop dynamically adjusts the compensation amount according to the prediction error, further eliminating the impact of fluctuations. The two loops promote each other and jointly cope with the challenge of line speed fluctuations.

[0125] When the encoder accumulates errors, the compensation loop can discover and correct the accumulated errors in time through sensor verification and error feedback; the prediction loop uses the corrected position data for state estimation to avoid the pollution of the filter state by false data. The two loops check each other and jointly improve the measurement accuracy.

[0126] The synergistic effect of double-loop control enables the system to maintain high-precision triggering under complex working conditions, which is significantly better than single prediction or single compensation control methods.

[0127] Reference Figure 4 The adaptive window optimization module 3 is responsible for dynamically adjusting the trigger window threshold based on historical trigger effects, forming a complete closed loop from execution to feedback to optimization, and realizing the continuous improvement of trigger precision.

[0128] The adaptive window optimization includes the following steps:

[0129] First, record the corrected trigger position and the actual trigger position of each shooting operation. The corrected trigger position is the final trigger position calculated by the double-loop position control module 2, and the actual trigger position is the vehicle body position corresponding to the actual execution of the camera shooting. The actual trigger position can be obtained through the real-time reading of the encoder at the trigger time.

[0130] Calculate the trigger error between the corrected trigger position and the actual trigger position:

[0131] ,

[0132] Where, is the first the correction trigger position of the last shooting, the first the actual trigger position of the last shooting, the first the trigger error of the last shooting.

[0133] acquire the trigger error data of the last shooting operation, and form a trigger error sequence . The value of the trigger error is determined according to the system running state, and is preferably 50 to 200 in the normal running stage, and is preferably 10 to 30 in the start-stop stage.

[0134] statistical analysis is performed on the trigger error sequence, and the trigger error mean and the trigger error standard deviation are calculated.

[0135] ,

[0136] ,

[0137] wherein, is the trigger error mean, is the trigger error standard deviation, and n is the sample number.

[0138] The trigger error mean reflects the systematic deviation of the system, and the standard deviation reflects the dispersion degree of the trigger accuracy. Based on the two statistical quantities, it can be judged whether the current trigger window threshold is appropriate, and the adjustment direction and amplitude are determined.

[0139] A preset error upper limit (preferably 1.5mm to 3mm) and a preset error lower limit (preferably 0.3mm to 0.8mm) are set. These two threshold values define the acceptable range of trigger accuracy.

[0140] When the absolute value of the trigger error mean is greater than the preset error upper limit, i.e. , it indicates that the current trigger window is too narrow, resulting in that the system cannot respond in time, and the trigger position deviation is large. At this time, the window threshold parameter needs to be increased to give the system more response time. The adjustment amount of the window threshold parameter is calculated as:

[0141] ,

[0142] wherein, is an adjustment gain coefficient, and the value range is 0.1 to 0.5, and in the preferred embodiment, the value is 0.3.

[0143] The adjusted window threshold parameter is:

[0144] ,

[0145] When the absolute value of the trigger error mean is less than the preset lower error limit, i.e. , it indicates that the current trigger window is too wide, and the trigger response is too early, which may lead to waste of system resources or inaccurate trigger timing. At this time, the window threshold parameter needs to be reduced to improve the trigger accuracy. The adjustment amount of the window threshold parameter is calculated as:

[0146] ,

[0147] The adjusted window threshold parameter is:

[0148] ,

[0149] When the absolute value of the trigger error mean is between the preset upper error limit and the lower error limit, i.e. , it indicates that the current trigger window threshold setting is reasonable, and there is no need to adjust, keeping the window threshold parameter unchanged.

[0150] In addition to adjusting the window threshold based on the error mean, the influence of the error standard deviation also needs to be considered. If the standard deviation is too large, it indicates that the trigger accuracy is unstable, and the window threshold should be increased appropriately to improve the system robustness. Set the standard deviation threshold (preferably 1mm to 2mm), when , the window threshold parameter is additionally adjusted:

[0151] ,

[0152] wherein, is the standard deviation adjustment coefficient, with a value range of 0.05 to 0.2, and in the preferred embodiment, the value is 0.1.

[0153] The adjusted window threshold parameter is updated to the hybrid trigger signal generation module 1 for the generation of the hybrid trigger signal in the next cycle. The updating process is realized through the parameter feedback interface between modules, ensuring that the new parameter can take effect in time.

[0154] The adaptive window optimization module forms a complete closed-loop feedback optimization mechanism by continuously monitoring the trigger effect, analyzing the trigger error, and dynamically adjusting the trigger parameters. This mechanism enables the system to learn and optimize autonomously according to the actual operating conditions, without the need for manual intervention to adapt to the detection needs of different vehicle models, different line speeds, and different working conditions, significantly improving the intelligence level and applicability of the system.

[0155] Based on the acquired surface painting image, the system utilizes a deep learning model for painting defect detection. The deep learning model adopts a convolutional neural network architecture, such as the YOLO series or Faster R-CNN, and is trained through a large number of labeled samples. The model can automatically identify painting defects in the image, including but not limited to particles (raised bumps caused by dust or unsolved paint), sagging (downward flowing traces caused by excessive paint thickness), orange peel (rough texture resembling an orange peel on the surface), color difference (color inconsistency in different areas), and other defect types.

[0156] For each detected defect, the system obtains its pixel coordinates in the surface painting image, including the row and column coordinates of the defect center point. Then, based on the corrected trigger position and camera calibration parameters, the pixel coordinates are converted to three-dimensional space coordinates in the vehicle body coordinate system.

[0157] The camera calibration parameters include camera intrinsic parameters (focal length, principal point coordinates, distortion coefficients) and camera extrinsic parameters (pose of the camera relative to the vehicle body coordinate system). These parameters are obtained through an offline calibration process and stored in the system configuration file.

[0158] The coordinate conversion process is based on the pinhole camera model and coordinate transformation relationship. The conventional coordinate conversion uses the homogeneous coordinate transformation matrix method. Specifically, the pixel coordinates are first back-projected to the camera coordinate system according to the camera intrinsic parameters, and then the coordinates in the camera coordinate system are transformed to the vehicle body coordinate system according to the camera extrinsic parameters. Since this belongs to the conventional coordinate conversion technology in computer vision, the formula derivation is not expanded here.

[0159] After obtaining the three-dimensional space coordinates of the defects, the system generates a defect distribution map based on all detected defect positions. The defect distribution map uses different colors or markers to represent different types and severity levels of defects, and visually displays the distribution of defects on the vehicle body surface. At the same time, the system outputs a defect detection report, including the total number of defects, the number of each type of defect, the defect position list, the defect severity level assessment, and other information for quality management personnel to analyze and make decisions.

[0160] During the start-up and stop phases of the painting line, due to the existence of acceleration and deceleration, the line speed will experience significant fluctuations and jitter, which poses special challenges to the trigger control system. To address this issue, the invention designs a special anti-jitter processing mechanism.

[0161] The system monitors the running state of the painting line in real time and determines whether it is in the start-up or stop phase by analyzing the line speed change rate. The specific determination method is as follows: calculate the first derivative (acceleration) and the second derivative (acceleration change rate) of the line speed, and when the absolute value of the acceleration exceeds the preset threshold (preferably 50mm / s 2 to 200mm / s 2If the duration exceeds the preset time (preferably 0.5s to 2s), the coating line is determined to have entered the start-up or stop phase.

[0162] Once the painting line is detected to be in a start-up or shutdown phase, the system immediately enters anti-shake mode. In anti-shake mode, the system takes the following measures:

[0163] First, increase the preset trigger window threshold. During normal operation, the trigger window threshold is preferably 3mm to 5mm; in anti-shake mode, the trigger window threshold is increased to 8mm to 15mm, giving the system more responsiveness and reducing the risk of missed shots.

[0164] Second, increase the filtering strength for positional deviations. In anti-jitter mode, the sliding window length... The range is increased from the normal 5 to 20 to 20 to 50, which enhances the smoothing effect on position deviations and suppresses the interference of high-frequency jitter on the control system.

[0165] Third, reduce the compensation coefficient. reference gain In image stabilization mode, The compensation response speed is reduced from the normal 1.0 to 0.5 to 0.7 to avoid overcompensation that could cause system oscillation.

[0166] Fourth, pause the adaptive window optimization function. In anti-shake mode, due to the instability of trigger error data, pause the adaptive adjustment of the window threshold parameter and keep the window threshold unchanged in the current anti-shake mode.

[0167] When the coating line speed stabilizes for a duration exceeding the preset stabilization time (preferably 3 to 10 seconds), the system determines that the start-stop phase has ended, exits the anti-jitter mode, and restores normal trigger parameters. The restoration process adopts a gradual approach, that is, parameters such as the trigger window threshold, filter intensity, and compensation coefficient gradually transition to normal values ​​within a certain time (preferably 2 to 5 seconds) to avoid sudden changes that could cause system shocks.

[0168] Through the anti-shake processing mechanism, this invention effectively solves the problem of unstable triggering during the start-up and shutdown phases of the coating line, reducing the missed detection rate during the start-up and shutdown phases from 12% to 0.5%, and significantly improving the reliability and detection coverage of the system.

[0169] The core advantage of this invention lies in the deep coupling and closed-loop collaboration between the hybrid trigger signal generation module 1, the dual-loop position control module 2, and the adaptive window optimization module 3.

[0170] The coupling relationship between the three modules is reflected on multiple levels:

[0171] Parameter level coupling: the output parameters of module 1 (position pulse signal, to-position confirmation signal, current cumulative position) are directly used as the input parameters of module 2 for the observation value update of Kalman filtering; the output parameters of module 2 (corrected trigger position, actual trigger position) are directly used as the input parameters of module 3 for trigger error calculation; the output parameters of module 3 (adjusted window threshold parameter) are fed back to module 1 for the generation condition judgment of the hybrid trigger signal.

[0172] State level coupling: the state estimation results of module 2 (position estimation value, speed estimation value) will affect the judgment and correction of the encoder cumulative error of module 1; the evaluation results of the trigger effect of module 3 will affect the setting of the compensation coefficient and the adjustment of the prediction model parameter in module 2; the fusion results of the sensor data of module 1 will affect the error statistical accuracy of module 3.

[0173] Logic level coupling: the trigger preparation state of module 1 will start the prediction compensation calculation of module 2; the trigger execution completion of module 2 will trigger the error analysis and parameter adjustment of module 3; the parameter update completion of module 3 will be fed back to module 1 to form a new round of trigger cycle.

[0174] The three modules form a complete closed-loop feedback loop: "hybrid trigger signal generation → position prediction and compensation control → actual trigger execution → trigger error analysis → window threshold adjustment → optimized hybrid trigger signal generation". This closed loop enables the system to continuously optimize the trigger strategy according to the actual trigger effect, realizing self-learning and self-adaptation.

[0175] The synergistic effect between the three modules is reflected in:

[0176] The high-quality trigger signal provided by module 1 promotes the prediction accuracy of module 2; the accurate trigger position provided by module 2 promotes the error analysis accuracy of module 3; the optimized parameters provided by module 3 promote the trigger performance improvement of module 1. The three modules promote each other, forming a positive feedback loop, and the system performance spirals upward.

[0177] Each module can improve the trigger accuracy to a certain extent, but when the three modules work together, their effect is much greater than the sum of the individual modules. Experimental data show that the use of the hybrid trigger mechanism alone can improve the trigger accuracy by about 30%, the use of the dual-loop control alone can improve it by about 50%, and the use of the adaptive window alone can improve it by about 20%; when the three are used together, the trigger accuracy is improved by more than 120%, showing a non-linear additive effect.

[0178] Module 1 focuses on the reliability of the trigger signal, solving the problem of signal loss through multi-sensor fusion; Module 2 focuses on the accuracy of the trigger position, solving the problem of position deviation through prediction compensation; Module 3 focuses on the adaptability of the trigger parameters, solving the problem of working condition changes through closed-loop optimization. The three modules complement each other and form a complete trigger control system that is reliable, accurate and self-adaptive.

[0179] In system design, there is a contradiction between trigger response speed and trigger accuracy. Fast response speed can easily cause trigger too early or too late, and slow response speed can reduce system real-time performance. The three modules cooperate to adjust the trigger window threshold adaptively, ensuring response speed while considering trigger accuracy, effectively solving this contradiction. For example, there is a contradiction between fast convergence and robustness of the prediction model. Fast convergence is easily disturbed by noise, and slow convergence affects tracking performance. The three modules cooperate through sensor verification and error feedback to ensure fast convergence while improving robustness.

[0180] Reference Figure 5 The application also provides an encoder trigger intelligent detection system for an automobile coating line, which comprises the following hardware components:

[0181] Incremental photoelectric encoder: installed on the driving roller shaft of the coating line, resolution 2048 PPR or 4096 PPR, output A, B, Z three-phase quadrature signals, working voltage 5V-24V, protection level IP65.

[0182] Photoelectric sensor: installed at the entrance of the detection area and multiple check points, using a transmission or reflection structure, detection distance 0.5m-2m, response time less than 10ms, working voltage 12V-24V, protection level IP67.

[0183] Line scan camera: using high-resolution CMOS or CCD line array sensor, resolution 8192 pixels or higher, line frequency 40kHz-80kHz, pixel size 5um-10um, supporting external hardware trigger, interface type GigE Vision or CameraLink.

[0184] LED line light source: used with line scan camera, light source type white light or specific wavelength monochromatic light, power 50W-200W, illumination uniformity greater than 90%, service life greater than 50000h.

[0185] Industrial computer: configured with Intel Core i7 or above processor, memory 16GB or above, equipped with high-speed image acquisition card, operating system Windows 10 or Linux, installed image processing and deep learning inference software.

[0186] PLC or motion controller: responsible for coordinating the work of each hardware device, collecting encoder and sensor signals, executing trigger control algorithms, and outputting camera trigger signals. Preferably, industrial controllers from brands such as Siemens, Omron, or Beckhoff are used.

[0187] The system software architecture includes a hybrid trigger signal generation module 1, a double-loop position control module 2, an adaptive window optimization module 3, an image acquisition module, a defect detection module, a data management module, and a human-machine interaction interface. Each software module runs on an industrial computer and achieves real-time processing and efficient collaboration through multi-threading or distributed architecture.

[0188] The system workflow is as follows: after the painting line starts, the PLC controller begins to collect encoder pulse signals and photoelectric sensor signals, determines whether the vehicle body has entered the detection area through the hybrid trigger signal generation module 1, and calculates the current cumulative position of the vehicle body. When the vehicle body approaches the preset shooting position, the double-loop position control module 2 starts the Kalman filter algorithm to predict the vehicle body position and speed, calculates the expected trigger position, monitors the prediction error, and generates a position compensation amount to obtain the corrected trigger position. The PLC controller sends a hardware trigger signal to the line scan camera based on the corrected trigger position, the camera performs shooting operations, the image acquisition module obtains the surface painting image and transmits it to the industrial computer. The defect detection module uses a deep learning model to analyze the image, identify painting defects, and calculate three-dimensional coordinates. The adaptive window optimization module 3 analyzes the trigger error, dynamically adjusts the trigger window threshold, and feeds back the new parameters to the hybrid trigger signal generation module 1. The data management module stores images, defect data, and trigger parameters for subsequent query and statistical analysis. The human-machine interaction interface displays the detection progress, defect distribution map, and system status in real time for operators to monitor and manage.

[0189] To verify the effectiveness of the present application, an industrial application test was conducted on a painting line in an automobile manufacturing enterprise for 3 months. The test painting line is 120 m long, with a designed line speed of 8 m / min to 12 m / min, and the actual running line speed fluctuates within ±10%. The test vehicle models include SUV and sedan, with about 500 vehicles detected per day.

[0190] During the test, the system of the present application was compared with the prior art system (a traditional system using a single encoder trigger and open-loop control). The main performance indicators are as follows:

[0191] Trigger position accuracy: the standard deviation of the trigger position of the prior art system is 2.3 mm, and the maximum deviation reaches 5.8 mm; the standard deviation of the trigger position of the system of the present application is 0.8 mm, and the maximum deviation is 2.1 mm. The trigger accuracy is improved by about 3 times.

[0192] Cross-vehicle consistency: The position deviation standard deviation of the prior art system is 3.5 mm when detecting different vehicles at the same detection position; the position deviation standard deviation of the system of the application is 1.1 mm. The cross-vehicle data consistency is improved by about 3 times.

[0193] Defect detection rate: The defect detection rate of the prior art system is 82.5%; the defect detection rate of the system of the application is 94.3%. The detection rate is increased by 11.8 percentage points.

[0194] False detection rate: The false detection rate of the prior art system is 5.2%; the false detection rate of the system of the application is 1.8%. The false detection rate is reduced by 3.4 percentage points.

[0195] Missed detection rate in the start-stop phase: The missed detection rate of the prior art system in the start-stop phase is 12%; the missed detection rate of the system of the application is 0.5%. The missed detection rate is reduced by 11.5 percentage points.

[0196] System adaptability: The prior art system needs to be manually recalibrated and adjusted when changing the vehicle model or adjusting the line speed, which takes about 2 hours; the system of the application can automatically complete parameter optimization within 10 minutes after changing the vehicle model or adjusting the line speed without manual intervention. The system adaptability is significantly improved.

[0197] Long-term stability: After running for one month, the prior art system needs to be recalibrated due to the cumulative error of the encoder, and the trigger accuracy decreases by about 30%; the system of the application has a sensor verification and error compensation mechanism, and the trigger accuracy only decreases by about 5% after running for three months without the need for line maintenance. The long-term stability of the system is significantly improved.

[0198] The experimental results fully verify the technical advantages of the application, proving that the application can effectively solve the technical problems of production line speed fluctuation, encoder cumulative error, and unstable start-stop phase, and achieve high-precision, high-consistency, and high-adaptability coating line defect detection, reaching an international leading level.

[0199] Based on theoretical analysis and experimental verification, the preferred ranges of the key technical parameters in the method of the application are as follows:

[0200] Encoder resolution: 2048 PPR to 4096 PPR, preferably 2048 PPR.

[0201] Pulse counting sampling period: 1 ms to 5 ms, preferably 2 ms.

[0202] Initial value of the preset trigger window threshold: 3 mm to 5 mm, preferably 4 mm.

[0203] Kalman filter sampling period: 10 ms to 50 ms, preferably 20 ms.

[0204] Position process noise variance: 0.01 to 0.1 , preferably 0.05 .

[0205] Velocity process noise variance: 1 to 10 , preferably 5 .

[0206] Observation noise variance: 0.01 to 0.1 , preferably 0.05 .

[0207] Sliding window length: 5 to 20, preferably 10.

[0208] Reference compensation gain: 0.5 to 2.0, preferably 1.0.

[0209] Maximum allowed compensation amount: 2mm to 5mm, preferably 3mm.

[0210] Trigger error statistics sample number (normal): 50 to 200, preferably 100.

[0211] Trigger error statistics sample number (start-stop): 10 to 30, preferably 20.

[0212] Pre-set error upper limit: 1.5mm to 3mm, preferably 2mm.

[0213] Pre-set error lower limit: 0.3mm to 0.8mm, preferably 0.5mm.

[0214] Adjustment gain coefficient: 0.1 to 0.5, preferably 0.3.

[0215] Standard deviation threshold: 1mm to 2mm, preferably 1.5mm.

[0216] Standard deviation adjustment coefficient: 0.05 to 0.2, preferably 0.1.

[0217] Anti-shake mode trigger window threshold: 8mm to 15mm, preferably 10mm.

[0218] Anti-shake mode filter window length: 20 to 50, preferably 30.

[0219] Line scan camera resolution: no less than 8192 pixels, preferably 8192 or 12288 pixels.

[0220] Line scan camera line frequency: no less than 40kHz, preferably 40kHz to 80kHz.

[0221] The above preferred ranges are determined based on a large amount of experimental data and engineering experience, and can obtain the best performance in most application scenarios. In special application scenarios, the parameter range can be adjusted appropriately according to actual needs.

Claims

1. An encoder-triggered intelligent detection method for automotive painting lines, characterized in that, Includes the following steps: The encoder collects the position pulse signal of the vehicle body under test on the painting line in real time, and the photoelectric sensor collects the position confirmation signal of the vehicle body under test. A mixed trigger signal is generated based on the position pulse signal and the position confirmation signal. Based on the hybrid trigger signal and historical position data, the current predicted position and predicted speed of the vehicle body to be detected are predicted using a Kalman filter algorithm, and the expected trigger position of the camera is determined based on the current predicted position. Obtain the current measured position of the encoder, calculate the position deviation between the current predicted position and the current measured position, generate a position compensation amount based on the position deviation, and use the position compensation amount to correct the expected trigger position to obtain the corrected trigger position; Based on the corrected trigger position control line scan camera to perform shooting operation, acquire the surface coating image of the vehicle body to be detected; Based on the triggering error between the corrected triggering position and the actual triggering position, the window threshold parameter of the triggering window is dynamically adjusted, and the adjusted window threshold parameter is fed back to the hybrid triggering signal generation step to achieve adaptive optimization of triggering accuracy. The generation of the mixed trigger signal based on the position pulse signal and the arrival confirmation signal includes: The pulse count value output by the encoder within a preset time window is obtained, and the instantaneous movement distance of the vehicle body to be detected is calculated based on the pulse count value. Determine whether the photoelectric sensor has detected the vehicle body to be detected entering the detection area. If so, use the current pulse count value of the encoder as the position reference point. Based on the location reference point and the instantaneous movement distance, calculate the current cumulative position of the vehicle body to be detected; When the distance between the current accumulated position and the preset shooting position is less than the preset trigger window threshold, the hybrid trigger signal is generated; The step of dynamically adjusting the window threshold parameter of the trigger window based on the trigger error between the corrected trigger position and the actual trigger position includes: Obtain the corrected trigger position and the actual trigger position for the most recent N shooting operations, and calculate the trigger error for each shooting operation; Perform statistical analysis on the triggering errors of N times, and calculate the mean and standard deviation of the triggering errors; The adjustment range of the window threshold parameter is determined based on the mean trigger error and the standard deviation of the trigger error. When the average trigger error is greater than the preset upper limit of error, the window threshold parameter is increased; when the average trigger error is less than the preset lower limit of error, the window threshold parameter is decreased. The adjusted window threshold parameter is updated to the trigger signal generation module for the generation of the mixed trigger signal in the next cycle; The method also includes anti-vibration processing during the start-up and shutdown phases of the coating line: The system checks whether the painting line is in the start-up or stop phase. If so, it enters the anti-vibration mode. In the anti-shake mode, the preset trigger window threshold is increased, and the filtering strength of the position deviation is improved; When the coating line speed remains stable for longer than the preset stable time, the anti-shake mode is exited and the normal trigger parameters are restored.

2. The method according to claim 1, characterized in that, The step of predicting the current predicted position and predicted speed of the vehicle body to be detected using a Kalman filter algorithm based on the hybrid trigger signal and historical position data includes: Establish a motion state model of the vehicle body to be detected, the motion state model including position state variables and velocity state variables; Obtain the position and velocity estimates from the previous moment, perform state prediction based on the motion state model, and obtain the current predicted position and the predicted velocity at the current moment. Obtain the position observation value of the encoder at the current time, and calculate the observation deviation between the position observation value and the current predicted position; Based on the observation bias and Kalman gain matrix, the current predicted position and the predicted velocity are updated to obtain the position estimate and velocity estimate at the current moment, and the position estimate is used as historical data for the prediction at the next moment.

3. The method according to claim 2, characterized in that, The step of calculating the position deviation between the current predicted position and the current measured position, and generating a position compensation amount based on the position deviation, includes: The current measured position of the encoder and the sensor position information of the photoelectric sensor are obtained. The current measured position is verified based on the sensor position information to determine the verified measured position. Calculate the instantaneous deviation between the current predicted position and the verified measured position, and apply a sliding window filter to the instantaneous deviation to obtain the position deviation; Based on the position deviation and the predicted velocity, a compensation coefficient is determined, wherein the compensation coefficient is positively correlated with the position deviation and negatively correlated with the predicted velocity; The position compensation amount is calculated based on the position deviation and the compensation coefficient.

4. The method according to claim 3, characterized in that, The step of correcting the expected trigger position using the position compensation amount to obtain the corrected trigger position includes: Obtain the current line speed and line speed fluctuation range of the coating line, and determine the speed correction factor based on the current line speed; Multiply the position compensation amount by the velocity correction factor to obtain the corrected position compensation amount; The corrected trigger position is obtained by adding the corrected position compensation amount to the expected trigger position. Determine whether the corrected trigger position is within the allowed trigger range. If yes, output the corrected trigger position as the final trigger position; otherwise, use the expected trigger position as the final trigger position.

5. The method according to claim 1, characterized in that, The method further includes: Based on the surface coating image, a deep learning model is used to detect coating defects, which include at least one of particles, runs, orange peel, or color difference. The pixel coordinates of the coating defect in the surface coating image are obtained, and the pixel coordinates are converted into three-dimensional spatial coordinates in the vehicle body coordinate system based on the correction trigger position and camera calibration parameters. A defect distribution map is generated based on the three-dimensional spatial coordinates, and a defect detection report is output.

6. The method according to claim 1, characterized in that, The encoder is an incremental photoelectric encoder with a resolution of not less than 2000 PPR; the line scan camera has a resolution of not less than 8192 pixels and a line frequency of not less than 40kHz.

7. An encoder-triggered intelligent detection system for automotive painting lines, used to implement the method described in any one of claims 1-6, characterized in that, include: The hybrid trigger signal generation module is used to collect the position pulse signal of the vehicle body to be detected on the painting line in real time through an encoder, collect the position confirmation signal of the vehicle body to be detected through a photoelectric sensor, and generate a hybrid trigger signal based on the position pulse signal and the position confirmation signal. A dual-ring position control module, connected to the hybrid trigger signal generation module, is used to predict the current predicted position and speed of the vehicle body to be detected using a Kalman filter algorithm based on the hybrid trigger signal and historical position data; determine the expected trigger position of the camera based on the current predicted position; obtain the current measured position of the encoder; calculate the position deviation between the current predicted position and the current measured position; generate a position compensation amount based on the position deviation; and use the position compensation amount to correct the expected trigger position to obtain the corrected trigger position. Based on the corrected trigger position control line scan camera to perform shooting operation, acquire the surface coating image of the vehicle body to be detected; An adaptive window optimization module, connected to the dual-ring position control module and the hybrid trigger signal generation module, is used to dynamically adjust the window threshold parameter of the trigger window based on the trigger error between the corrected trigger position and the actual trigger position, and feeds back the adjusted window threshold parameter to the hybrid trigger signal generation module to achieve adaptive optimization of trigger accuracy.

Images