Vision-based tire chip mounting position deviation real-time detection method and system

Abstract

The invention discloses a tire chip mounting position deviation real-time detection method and system based on vision, particularly relates to the technical field of tire manufacturing quality control, and is used for solving the problem that chip deviation cannot be detected in real time due to a visual blind area in a rolling process in the prior art. The method comprises the following steps: continuously acquiring chip area images when a rolling mechanism acts; extracting chip contour feature points and rubber blank mark points in the image to generate a dynamic displacement vector; tracking a tread pattern unit displacement track to generate a relative strain gradient; decomposing the displacement vector, extracting a texture flow direction angle of the rubber blank, and calculating and correcting a real-time deviation angle; when the correction deviation angle exceeds the limit or the ratio of the displacement component to the strain gradient exceeds the threshold value, irreversible deviation is judged to occur; and a shutdown instruction is triggered before rolling is finished, so that chip offset real-time monitoring and active protection in the whole rolling process are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tire manufacturing quality control, more particularly, the present application relates to a visual-based real-time detection method and system for position deviation of tire chip installation. BACKGROUND

[0002] In the rubber tire manufacturing, in order to realize the automatic installation of the built-in chip, the existing technology generally adopts the following process: the chip is grabbed to the pre-marked rubber blank position by a mechanical hand, and after a short visual positioning confirmation, the chip is fixed by rolling, and the position of the chip is detected once before or after rolling. This scheme relies on the pre-set static calibration coordinate system, and the detection node is usually independent of the rolling process.

[0003] However, due to the viscoelasticity of the tire rubber blank, the chip is prone to micro-displacement during the pressure process of the rolling roller, and the existing detection link and the rolling process are mutually cut off in time sequence and space: when the rolling mechanism acts, its physical structure completely blocks the chip area, causing the displacement process to become a visual monitoring blind area. This process chain design defect makes the existing method unable to capture the real-time position deviation in the rolling stage, causing the contradiction of "qualified detection result but actual chip has deviated", and further causing functional defects of the vulcanized tire. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the present application provides a visual-based real-time detection method and system for position deviation of tire chip installation to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: A visual-based real-time detection method for position deviation of tire chip installation, comprising the following steps: S1, continuously acquiring real-time images of the chip area when the rolling mechanism starts to act; S2, extracting the contour feature points of the chip and the pre-set mark points on the surface of the rubber blank in the real-time images, and generating a dynamic displacement vector of the chip in the rolling process based on the relative positions of the contour feature points and the pre-set mark points; S3, tracking the displacement trajectory of the pre-set tread pattern unit around the chip under the action of the dynamic displacement vector, and generating the relative strain gradient between the tread pattern unit and the chip; S4, decomposing the dynamic displacement vector into a normal displacement component and a tangential displacement component, extracting the flow angle of the natural texture of the rubber blank surface in the chip edge contact area, and calculating the real-time deviation angle between the flow angle and the movement direction of the rolling roller, and simultaneously correcting the real-time deviation angle in association with the relative strain gradient; S5. When the corrected real-time deviation angle exceeds a preset critical angle, or the ratio of the tangential displacement component to the relative strain gradient exceeds a preset ratio threshold, it is determined that the chip has irreversible offset; S6. Before the rolling mechanism completes its operation, if irreversible deviation is detected, a stop command is immediately triggered.

[0006] Furthermore, when the crushing mechanism starts to operate, real-time images of the chip area are continuously collected, including: An optical lens is fixedly installed in the axial through hole of the rolling roller so that the optical axis of the lens is vertically aligned with the chip mounting area; Based on the downward displacement of the rolling roller, the coaxial light source is turned on and off in real time to make the irradiation direction of the light source parallel to the optical axis; Dynamically adjust the image acquisition frame rate according to the linear speed of the rolling roller, and perform motion blur correction on the acquired images; Output the corrected image sequence as a live image.

[0007] Furthermore, the contour feature points of the chip and the preset marking points on the surface of the rubber blank in the real-time image are extracted, and the dynamic displacement vector of the chip during the rolling process is generated based on the relative positions of the contour feature points and the preset marking points, including: Receive multiple frames of real-time images, perform edge enhancement on each frame, and extract the chip outline pixel set and the pre-set marker pixel set on the surface of the rubber blank; Identify corner features in the chip contour pixel set as contour feature points, and locate centroid coordinates in the preset marker pixel set as marker point positions; The relative offset between the contour feature points and the marker points between adjacent frames is tracked, and a dynamic displacement vector is generated based on the vector superposition of the relative offset in the rolling time sequence.

[0008] Furthermore, the vector superposition based on the relative offset on the rolling time sequence includes: Establish a displacement accumulation coordinate system with the start time of rolling as the time zero; Convert the relative offset between the contour feature points and the marker points between adjacent frames into a two-dimensional vector in the displacement accumulation coordinate system; Perform vector addition operation on the two-dimensional vectors of consecutive frames according to the rolling time sequence, and output the accumulated vector as the dynamic displacement vector.

[0009] Furthermore, the displacement trajectory of the preset tread pattern unit around the chip under the action of the dynamic displacement vector is tracked to generate the relative strain gradient between the tread pattern unit and the chip, including: Select at least three tread pattern units at equal angles on the periphery of the chip as monitoring points, and obtain the initial spatial coordinates of the monitoring points; Based on the dynamic displacement vector, the coordinates of the monitoring points are compensated for the rigid displacement and the non-rigid deformation of the monitoring points is separated; Calculate the real-time change rate of the distance vector from each monitoring point to the center of the chip, and take the maximum value of the real-time change rate as the local strain representation value of the current frame; Fit the spatial distribution curve of the local strain representation value along the movement direction of the rolling roller, and calculate the slope of the curve as the relative strain gradient.

[0010] Further, fitting the spatial distribution curve of the local strain representation value along the movement direction of the rolling roller and calculating the slope of the curve comprises: Establish a spatial rectangular coordinate system with the center of the chip as the origin and the movement direction of the rolling roller as the X-axis; Map the local strain representation value of each monitoring point to the corresponding coordinate position of the X-axis; Least squares linear fitting is performed on the coordinate position and the strain value; Extract the first term coefficient of the fitted straight line equation as the relative strain gradient.

[0011] Further, the dynamic displacement vector is decomposed into a normal displacement component and a tangential displacement component, the flow direction angle of the natural texture of the rubber surface in the contact area of the chip edge is extracted, and the real-time deviation angle between the flow direction angle and the movement direction of the rolling roller is calculated. At the same time, the real-time deviation angle is corrected in association with the relative strain gradient, comprising: Establish a local coordinate system with the movement direction of the rolling roller as the reference, project the dynamic displacement vector onto the normal axis and the tangential axis to generate the normal displacement component and the tangential displacement component; Divide a rectangular analysis area in the contact area of the chip edge, and extract the main direction angle of the natural texture of the rubber surface in the rectangular analysis area as the flow direction angle; Calculate the absolute value of the included angle between the flow direction angle and the movement direction of the rolling roller as the real-time deviation angle; Linearly compensate the real-time deviation angle according to the relative strain gradient: multiply the relative strain gradient by a compensation coefficient, and then subtract the real-time deviation angle to output the corrected real-time deviation angle.

[0012] Further, when the corrected real-time deviation angle exceeds a preset critical angle, or the ratio of the tangential displacement component to the relative strain gradient exceeds a preset ratio threshold, it is determined that the chip has occurred irreversible deviation, comprising: Receive the corrected real-time deviation angle, the tangential displacement component and the relative strain gradient; Real-time compare the corrected real-time deviation angle with the preset critical angle; Synchronously calculate the quotient value of the tangential displacement component and the relative strain gradient, and real-time compare the quotient value with the preset ratio threshold; When the corrected real-time deviation angle is greater than the preset critical angle, or the quotient value is greater than the preset ratio threshold, an irreversible deviation flag is generated.

[0013] Further, if irreversible deviation is detected before the action of the rolling mechanism ends, a stop command is triggered immediately, including: Real-time monitoring of the rolling mechanism movement position, and activating a response enable when the movement position is in a preset safety intervention interval; Receiving an irreversible deviation flag, and if the irreversible deviation flag is true under the effective state of the response enable, sending an emergency stop command to the rolling mechanism servo driver; The emergency stop command triggers the rolling roller to retreat to the top dead center position; Clearing all detection state flags in the current rolling period.

[0014] On the other hand, the present application provides a real-time visual-based tire chip installation position deviation detection system, including the following modules: An image acquisition module for continuously acquiring real-time images of the chip area when the rolling mechanism starts to act; A displacement calculation module for extracting the contour feature points of the chip and the preset marker points on the rubber surface in the real-time images, and generating a dynamic displacement vector of the chip in the rolling process based on the relative positions of the contour feature points and the preset marker points; A strain analysis module for tracking the displacement trajectory of the preset tread pattern unit around the chip under the action of the dynamic displacement vector, and generating the relative strain gradient between the tread pattern unit and the chip; A deviation correction module for decomposing the dynamic displacement vector into a normal displacement component and a tangential displacement component, extracting the flow angle of the natural texture of the rubber surface in the chip edge contact area, and calculating the real-time deviation angle between the flow angle and the rolling roller movement direction, while correlating the relative strain gradient to correct the real-time deviation angle; An offset determination module for determining that the chip has irreversible deviation when the corrected real-time deviation angle exceeds a preset critical angle, or the ratio of the tangential displacement component to the relative strain gradient exceeds a preset ratio threshold; A safety control module for triggering a stop command immediately if irreversible deviation is detected before the action of the rolling mechanism ends.

[0015] Compared with the prior art, the present application has the following beneficial effects: 1. Through continuous image acquisition and analysis of the chip area during the action of the rolling mechanism, the visual monitoring blind area problem caused by the rolling mechanism is overcome; a real-time visual feedback chain covering the entire rolling process is established, and the micro-displacement of the chip on the viscoelastic rubber is dynamically captured, fundamentally solving the contradiction between "static detection qualification" and "dynamic actual deviation" caused by process fragmentation, ensuring the spatiotemporal consistency of the detection node and the rolling action, and significantly improving the authenticity and timeliness of the chip position detection.

[0016] 2. A multi-dimensional displacement and strain collaborative analysis is constructed, the overall movement of the chip is described by a dynamic displacement vector, the local deformation of the glue body is reflected by the relative strain gradient, the material rheological influence is compensated by the texture flow direction angle deviation correction technology, the position, strain, texture flow direction and other multi-source information are fused, the recoverable elastic displacement and irreversible displacement of the chip can be accurately distinguished, and finally the safety control is triggered based on the rigorous multi-threshold judgment logic, the functional defects of the tire caused by the displacement of the chip are effectively prevented, and the product quality and yield are greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A flowchart of a real-time detection method for a tire chip installation position deviation based on vision according to the present application; Figure 2 A structural schematic diagram of a real-time detection system for a tire chip installation position deviation based on vision according to the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0019] Embodiment 1: Figure 1 A real-time detection method for a tire chip installation position deviation based on vision according to the present application is given, which includes the following steps: S1. When the rolling mechanism starts to act, real-time images of the chip area are continuously collected; S2. The contour feature points of the chip and the preset marker points on the glue body surface in the real-time images are extracted, and the dynamic displacement vector of the chip in the rolling process is generated based on the relative positions of the contour feature points and the preset marker points; S3. The displacement trajectory of the preset tread pattern unit around the chip under the action of the dynamic displacement vector is tracked, and the relative strain gradient between the tread pattern unit and the chip is generated; S4. The dynamic displacement vector is decomposed into a normal displacement component and a tangential displacement component, the flow direction angle of the natural texture on the glue body surface of the chip edge contact area is extracted, the real-time deviation angle of the flow direction angle and the rolling roller movement direction is calculated, and the real-time deviation angle is corrected in association with the relative strain gradient; S5. When the corrected real-time deviation angle exceeds a preset critical angle, or the ratio of the tangential displacement component to the relative strain gradient exceeds a preset ratio threshold, it is determined that the chip has irreversible displacement; S6. Before the rolling mechanism stops acting, if irreversible displacement is detected, a stop command is immediately triggered.

[0020] S1, when the rolling mechanism starts to act, the real-time image of the chip area is continuously collected, specifically: When the rolling mechanism starts to act, the real-time image of the chip area is continuously collected by the penetrating vision acquisition unit. The specific implementation process includes the following operations: a flange is used to fix and install an optical lens in the axial through hole of the rolling roller, which is a through-type machined hole with a diameter of 20 mm for example. When installing, a laser calibrator is used to adjust the posture of the lens, so that the optical axis of the lens forms a 90-degree vertical angle with the rolling roller axis, and the optical axis center line is accurately aligned with the geometric center position of the chip mounting area, with a position deviation controlled within a range of plus or minus 0.1 mm for example. The depth of field range of the optical lens is set to 15-25 mm for example, covering the full size in the thickness direction of the chip, and an adjustable aperture mechanism is configured to adapt to the change of the reflectivity of the tire rubber surface. The aperture opening is dynamically adjusted according to the ambient light intensity, with an adjustment step of 0.1 unit increment of the aperture coefficient for example.

[0021] The operation of controlling the on-off of the coaxial light source based on the real-time displacement of the rolling roller is implemented as follows: a high-precision magnetic grid displacement sensor is embedded in the side wall of the rolling roller, which monitors the axial displacement of the rolling roller relative to the rack in real time. The displacement is increased by 1 mm for example, which triggers the light source control signal once. The control logic is to turn on the light source when the displacement is within the range of 5-95 mm, and to turn off the light source when the displacement is less than 5 mm or greater than 95 mm. The coaxial light source uses an infrared LED array with a wavelength of 850 nm for example, which refracts the light beam into a parallel light path through a beam splitter prism, ensuring that the parallelism error between the light source irradiation direction and the lens optical axis is less than 0.5 degrees. The light source intensity is dynamically adjusted according to the gray histogram of the collected image, and when the average gray value of the image is lower than 80 for example, the light source power is increased by 10%, and when the average gray value is higher than 180 for example, the power is reduced by 15%, with a response time of no more than 5 ms.

[0022] The implementation process of adjusting the image acquisition frame rate according to the dynamic roller linear speed includes: obtaining the real-time linear speed of the roller through the encoder of the servo motor, and the linear speed unit is millimeters per second. The corresponding relationship between the image acquisition frame rate and the linear speed is established as follows: when the linear speed is, for example, 50 millimeters per second, the reference frame rate is set to 100 frames per second, and the frame rate is increased by 20 frames per second for each increase of 10 millimeters per second. The dynamic adjustment is realized through the following relationship: the acquisition frame rate is equal to the reference frame rate multiplied by the real-time linear speed and then divided by the reference linear speed, wherein the reference linear speed is set to 50 millimeters per second. In the acceleration stage of the roller, the frame rate adjustment lag time is not more than 2 acquisition cycles. When performing motion blur correction on the collected images, a bidirectional recursive filtering algorithm is used to first calculate the pixel displacement of the feature points in the adjacent three frames of images, and then to perform reverse pixel compensation on the current frame according to the displacement. The compensation amount calculation relationship is: the compensation pixel number is equal to the feature point displacement multiplied by the exposure time and then divided by the frame interval time, wherein the exposure time is fixed at 1 millisecond, and the frame interval time is dynamically calculated according to the actual frame rate.

[0023] The specific operation of outputting the corrected image sequence as a real-time image is as follows: performing median filtering and noise reduction processing on a single frame of image after motion blur correction, and the filter window size is, for example, 3 pixels by 3 pixels. Subsequently, twenty consecutive frames of images are packaged in time sequence as image sequence data packets, and the time stamp information is embedded in the data packet header, and the time stamp accuracy is 0.1 millisecond. The image sequence is transmitted to the processing unit through a gigabit Ethernet interface, and the transmission protocol adopts the industrial standard image transmission protocol, and the transmission bandwidth is stabilized at 900 megabits per second. Each frame of image is attached with metadata tags, including lens focal length value such as 85 millimeters, aperture coefficient such as 8, light source intensity value such as 300 lumens, and other acquisition parameters, to ensure that the subsequent processing steps can trace the original acquisition conditions.

[0024] The flange and the axial through hole of the roller shaft are assembled in a transition fit manner, and the fitting tolerance is controlled within the standard tolerance level. The flange end face is processed with a positioning pin hole, and the circumferential position is locked by, for example, a 3 millimeter diameter hardened steel pin. A high-temperature-resistant optical glass protective cover is installed at the front end of the lens, which can withstand, for example, a high temperature environment of 150 degrees Celsius. The illumination uniformity of the chip mounting area is optimized by a diffuse reflection plate, and the illuminance fluctuation range measured at a distance of, for example, 50 millimeters from the surface of the chip is less than 10%.

[0025] The resolution of the magnetic grid displacement sensor reaches 0.001 millimeter, and the sampling frequency is 5 kilohertz. The on-off action of the light source is executed through a solid-state relay, and the switching response time is less than 1 millisecond. The light source driving circuit adopts shielded twisted pair wiring, and the shielding layer ground resistance is less than 4 ohms. The light source power regulation adopts pulse width modulation technology, and the duty cycle regulation accuracy is 0.1%.

[0026] The servo motor encoder signal is transmitted to the image acquisition card through a standard interface, and the signal transmission delay compensation value is 2 milliseconds. When a line speed mutation exceeding 20 mm / s2 is detected, a frame rate prediction algorithm is enabled: according to the speed change rate of the previous five frames, the line speed at the next acquisition time is predicted, and the frame rate parameter is adjusted in advance. Motion blur correction uses pixel-level displacement compensation, the compensation direction is opposite to the movement direction of the rolling mill, and the compensation amount is calculated considering the decomposition of the drum angular velocity vector.

[0027] In the image sequence output stage, the median filter processing retains the original image bit depth of 10 bits, and the filtered image is converted to an 8-bit grayscale image. The data packet encapsulation format follows the standard time format, and the time zone information synchronously records the offset of Coordinated Universal Time. The transmission link sets a double-buffer mechanism, and when the network delay exceeds 10 milliseconds, the local cache is automatically enabled, and the cache capacity can store 200 frames of images. All acquisition parameters are written into a standard format log file, and the file header contains the device serial number and calibration certificate number.

[0028] S2, extract the contour feature points of the chip and the preset marker points on the rubber surface in the real-time image, and generate a dynamic displacement vector of the chip in the rolling process based on the relative positions of the contour feature points and the preset marker points, specifically: After receiving the continuous multiple frames of real-time image sequence output from step S1, the image sequence is an 8-bit grayscale image after motion blur correction. The specific operation of edge enhancement processing for each real-time image is as follows: first, a Laplacian operator is used for convolution operation, for example, a 3x3 matrix template, the center pixel weight is for example negative 4, and the adjacent pixel weight is for example positive 1. When performing convolution calculation, each pixel point of the image is traversed, and the output value is obtained through the following relationship: multiply the center pixel gray value by negative 4, add the gray values of the adjacent four direction pixels by positive 1, and then divide by the total weight for normalization. When the absolute value of the output value is greater than, for example, 50, it is determined as an edge pixel, and the gray value of such pixel is mapped to 255, and the gray values of the remaining pixels are compressed to the interval of 0 to 50, forming a binary edge image. This processing enhances the contrast of the chip and the rubber surface, especially suitable for the reflective characteristics of the tire rubber surface.

[0029] When extracting the chip contour pixel set and the preset marker point pixel set on the adhesive surface respectively, the operation is based on the edge-enhanced binary image. The extraction logic of the chip contour pixel set is: through the region growing algorithm, the seed point is the center of the image, and the diffusion condition is that the gray value of the adjacent pixel is equal to 255 and the distance from the seed point is not more than, for example, 100 pixels. The extraction of the preset marker point pixel set on the adhesive surface adopts the morphological opening operation, first uses, for example, a circular structural element with a diameter of 5 pixels to erode the image to remove noise points, and then uses the same structural element to dilate to restore the shape of the marker point. The preset marker point is a high-contrast circular marker on the adhesive surface, with a diameter of, for example, 2 mm, which is represented as a white area with an area of, for example, 50 to 200 pixels in the image.

[0030] The implementation process of identifying the corner point feature in the chip contour pixel set as the contour feature point includes: traversing the boundary chain code of the contour pixel set, and calculating the curvature change rate of each boundary point. The curvature change rate is calculated by the following relationship: taking the chord length formed by, for example, 5 pixels before and after the current point, calculating the cosine value of the included angle of the two chords. When the cosine value is less than, for example, 0.2, it is determined as a corner point candidate position, and the final corner point is screened through non-maximum suppression, and the suppression window is, for example, 7 by 7 pixels. Each chip contour retains, for example, 4 corner point features, respectively corresponding to the four vertex positions of the chip rectangular contour. When positioning the centroid coordinates as the marker point position in the preset marker point pixel set, the pixel coordinates of each marker point connected domain are counted, the horizontal coordinate of the centroid is equal to the sum of the horizontal coordinates of all pixels in the connected domain divided by the total number of pixels, and the vertical coordinate is calculated in the same way. Each marker point outputs the centroid coordinates in the form of, for example, a floating point number, with a precision of 0.01 pixels.

[0031] When tracking the relative offset of the contour feature point and the marker point position between adjacent frames, the inter-frame matching relationship is established. For the Nth frame and the N+1th frame image, the feature points are matched through bidirectional nearest neighbor search: taking a corner point in the Nth frame as a reference, searching for the nearest corner point in the N+1th frame in terms of Euclidean distance, and then verifying the uniqueness of the match in the reverse direction. The offset is calculated as the coordinate difference of the matched point pair, for example, the horizontal coordinate difference of a certain corner point in two consecutive frames is Δx, and the vertical coordinate difference is Δy. The marker point position offset is calculated in the same way, but the average displacement of multiple marker points needs to be considered. Finally, each feature point outputs, for example, a two-dimensional offset (Δx, Δy) with a numerical unit of pixels.

[0032] The operation of generating a dynamic displacement vector based on the vector superposition of the relative offset amounts on the rolling timing includes establishing a displacement cumulative coordinate system. The coordinate system takes the instant when the rolling roller contacts the rubber blanket as the time zero point, and determines the absolute time reference through the encoder pulse signal. The coordinate system origin is set at the center position of the chip in the first frame image, the X-axis is parallel to the rolling roller movement direction, and the Y-axis is perpendicular to the rubber blanket surface. Each offset amount is converted into a two-dimensional vector in the coordinate system: multiply the pixel offset amount by a physical calibration coefficient, for example, each pixel corresponds to an actual size of 0.05 mm, then the vector X component is Δx multiplied by 0.05, and the Y component is Δy multiplied by 0.05, with the unit of millimeter.

[0033] When performing vector addition operation on the two-dimensional vectors of the continuous frames according to the rolling timing, the vectors are sequentially accumulated from the time zero point. Assuming that the offset vector of the i-th frame is (ΔXi, ΔYi), then the cumulative vector at the k-th frame is the algebraic sum of all i equal to 1 to k frame vectors, that is, the X component is equal to ΔX1 plus ΔX2 plus... plus ΔXk, and the Y component is calculated in the same way. The accumulation process uses double-precision floating-point number calculation, and the result of each frame is updated and stored to a ring buffer. When the rolling time exceeds, for example, 1 second, the early data is automatically emptied, and only the vector sequence of the recent, for example, 100 frames is retained. The final output dynamic displacement vector is the cumulative vector (Xk, Yk) of the current frame, which is transmitted to the S3 processing unit through gigabit Ethernet.

[0034] In the corner point identification link, the specific implementation of non-maximum suppression is: among the candidate points with a curvature change rate greater than a threshold value, only the point with the maximum change rate in the local window is retained. The window size is set according to the chip size, for example, the actual size of 10 mm of the chip corresponds to 100 pixels in the image, then the suppression window is set to 10 by 10 pixels. If the distance between two candidate points is less than the window size, the point with the smaller change rate is eliminated.

[0035] The centroid coordinate calculation adopts a sub-pixel accuracy algorithm: a quadratic surface fitting is performed on the connected domain of the marker point to obtain the sub-pixel level centroid position by solving the extreme point of the fitting surface. Specifically, 9 points including the pixel with the highest gray value in the connected domain and its eight neighbors are taken to establish a bivariate quadratic equation system to solve the maximum value point coordinates, and the accuracy can reach 0.001 pixels.

[0036] An abnormal value elimination mechanism is set in the vector accumulation process: when a certain frame offset amount exceeds, for example, 3 times the standard deviation of the historical average value, it is determined as a matching error, and the frame vector does not participate in the accumulation. The standard deviation calculation is based on the previous, for example, 10 frame offset amount sequence, and the threshold boundary is dynamically updated. The accumulation result is additionally marked with a time stamp every frame, and the time stamp is derived from the hardware clock of the image acquisition card, which is strictly synchronized with the rolling timing.

[0037] S3, the displacement trajectory of the preset tread pattern unit around the chip under the action of the dynamic displacement vector is generated, and the relative strain gradient between the tread pattern unit and the chip is generated, specifically: At least three tread pattern units are selected as monitoring points at equal angles around the chip, and the specific operation is as follows: taking the geometric center point of the chip as the reference point, the monitoring point positions are evenly distributed along the circumferential direction around the reference point. The distribution method adopts a fixed angular interval, for example, one monitoring point is set every 120 degrees, and a total of three monitoring points are set. Each monitoring point corresponds to the center position of the top of the tire tread pattern unit, which is a raised structure pre-formed on the rubber blank surface. When obtaining the initial spatial coordinates of the monitoring points, the three-dimensional position data of each monitoring point relative to the center of the chip is recorded by a non-contact three-dimensional measuring device. The position data includes three directional components: a component parallel to the movement direction of the rolling mill, a component perpendicular to the rubber blank surface, and a component along the lateral direction of the chip. The measurement accuracy is controlled at, for example, 0.01 millimeter level, and the initial coordinate data is stored in a storage unit.

[0038] The implementation process of rigid displacement compensation for monitoring point coordinates based on dynamic displacement vector includes: receiving the dynamic displacement vector output from the previous step, which includes two displacement components parallel to the rolling direction and perpendicular to the rubber blank surface. Reverse displacement correction is performed on the current coordinates of each monitoring point: subtract the first component value of the dynamic displacement vector in the parallel rolling direction component, subtract the second component value in the perpendicular rubber blank surface component, and the lateral component remains unchanged. The compensated coordinate data eliminates the position offset caused by the overall movement of the chip. When separating the non-rigid deformation variable of the monitoring point, calculate the spatial straight line distance change between the compensated coordinates and the initial coordinates: the deformation variable is equal to the difference between the straight line distance of the compensated coordinates and the initial coordinates in three-dimensional space, and the absolute value is taken as the deformation variable value.

[0039] When calculating the real-time change rate of the distance vector of each monitoring point to the center of the chip, first calculate the actual spatial distance of the monitoring point to the center of the chip at the current time. The distance is obtained by calculating the square of the three-dimensional coordinate components: add the square values of the three directional coordinate components and take the square root. The change rate calculation adopts the following steps: first step, calculate the difference between the current distance and the initial distance, second step, divide the difference by the initial distance to get the relative change, third step, divide the relative change by the time interval of adjacent two frames of images. The time interval is determined by the image acquisition frame rate, for example, when one frame of image is collected every 0.01 seconds, the time interval is taken as 0.01 seconds. Take the positive value of the change rate value in all monitoring points as the local strain representation value of the current frame, which reflects the maximum tensile deformation degree of the gel.

[0040] A spatial rectangular coordinate system is established when fitting the spatial distribution curve of the local strain characteristic value along the movement direction of the rolling roller: the origin of the coordinate system is fixed at the center point of the chip, and the horizontal coordinate axis direction is completely coincident with the movement direction of the rolling roller. The spatial position of each monitoring point is projected onto the horizontal coordinate axis, and the coordinate value of the projection point is taken as the horizontal axis position data. The local strain characteristic value corresponding to each monitoring point is taken as the vertical axis data to form a set of horizontal and vertical coordinate data pairs. For example, three monitoring points generate three data points, and each data point contains the horizontal coordinate position and the corresponding strain value.

[0041] The specific operation of least squares linear fitting of the coordinate position and the strain value includes: assuming that there is a straight line, so that the perpendicular distance square sum of all data points to the straight line is minimum. Two parameters of the straight line are solved by mathematical operation: the first parameter is the slope of the straight line, which represents the change rate of strain with the spatial position; the second parameter is the intercept, which represents the strain value at the origin position. The calculation method of the slope parameter is: first, calculate the deviation of all horizontal coordinates from the average value, and the deviation of all strain values from the average value; then multiply each pair of deviations and sum them up; finally, divide by the square sum of all horizontal coordinate deviations. The slope parameter is the relative strain gradient.

[0042] The relative strain gradient value is extracted as the output result, which represents the strain change degree per unit length along the rolling direction. A positive value indicates that the strain increases with the advancement of the rolling direction, and a negative value indicates that the strain decreases. The final output result is transmitted to the subsequent processing step, and the goodness of fit evaluation value is also output for judging the data reliability.

[0043] In the monitoring point arrangement link, the number of angle intervals is dynamically adjusted according to the tire size: three monitoring points are used when the tire size is less than a certain size (for example, seventeen-inch hub diameter), and four monitoring points are used when the size is greater than the certain size, and the interval is adjusted to ninety degrees. The distance between the monitoring point and the center of the chip is set to a fixed multiple of the characteristic size of the chip, for example, 1.5 times the side length of the chip.

[0044] Rigid displacement compensation increases the rotation component correction: when the chip rotates obviously, an angle correction term is added in the coordinate compensation. The rotation angle is obtained by calculating the position change of the feature points of the chip, and the rotation correction is implemented on the monitoring point coordinates around the vertical axis. The rotation correction amount is equal to the detected rotation angle, and the direction is opposite to the rotation direction of the chip.

[0045] Data filtering is set for strain change rate calculation: the median filtering process is performed on the strain change rate sequence of continuous multiple frames, and the processed median value is taken for maximum value comparison. When there are multiple monitoring points, the local strain characteristic value is taken as the arithmetic mean of the data of the two monitoring points with the largest change rate, so as to avoid the influence of single point anomaly.

[0046] The least square fitting introduces a distance weight mechanism: the weight coefficient is assigned according to the vertical distance between the monitoring point and the axis of the roller movement direction. The monitoring point with a distance less than a certain threshold (for example, two millimeters) is given a higher weight, and the one with a larger distance is given a lower weight. When weighted fitting, the contribution of each data point in the slope parameter calculation is multiplied by its weight coefficient.

[0047] The relative strain gradient outputs a pre-implementation credibility check: when the goodness-of-fit evaluation value is lower than the set threshold (for example, 0.6), it is determined that the current fitting result is unreliable, and the output value remains the previous valid result. When the gradient value jumps sharply, a historical data smoothing algorithm is used: the current calculation value is multiplied by a weight coefficient of 0.7, and the previous output value is multiplied by a weight coefficient of 0.3, and the final output value is obtained.

[0048] S4, decompose the dynamic displacement vector into normal displacement component and tangential displacement component, extract the flow direction angle of the natural texture of the die edge contact area adhesive surface, and calculate the real-time deviation angle of the flow direction angle and the roller movement direction, and correlate the relative strain gradient to correct the real-time deviation angle, specifically: The dynamic displacement vector and the relative strain gradient data from the previous step are received. The specific operation of establishing a local coordinate system with the roller movement direction as the reference is as follows: a space rectangular coordinate system is established at the center point of the chip, the positive direction of the horizontal axis of the coordinate system is completely coincided with the tangent direction of the actual movement path of the roller, and the positive direction of the vertical axis is perpendicular to the adhesive surface and points to the outside. Decompose the dynamic displacement vector into two coordinate axis directions: the vertical axis direction component is defined as the normal displacement component, and the value of this component is equal to the projection length of the dynamic displacement vector on the vertical axis; the horizontal axis direction component is defined as the tangential displacement component, and the value of this component is equal to the projection length of the dynamic displacement vector on the horizontal axis. The projection length is calculated by multiplying the vector module length and the cosine value of the direction angle, and the calculation result is kept to three decimal places, and the unit is unified to millimeters.

[0049] When the rectangular analysis region is defined at the chip edge contact area, a fixed width is extended from the chip outer contour line to the rubber region. The specific region parameters are: a 2mm width is extended along the normal direction of the chip contour, and a 100% range of the chip side length is covered along the contour tangent direction. This region appears as a closed belt area surrounding the chip in the image. When extracting the main direction angle of the natural texture of the rubber surface in the region, the block gradient statistical algorithm is used: first, the region is divided into 0.5mm by 0.5mm square sub-regions, and the following operations are performed in each sub-region: calculate the gray gradient direction of each pixel point, which is obtained by taking the inverse tangent of the horizontal gray difference divided by the vertical gray difference, and the angle range covers 0 to 180 degrees. The histogram distribution of all pixel gradient directions is counted, and the median value of the angle interval with the highest frequency is taken as the main direction of the sub-region. The main direction angle (flow direction angle) of the final rectangular region is the weighted average value of all sub-region main directions, and the weight is set according to the distance from the center of the sub-region to the chip edge, and the weight decreases by 10% for every 1mm increase in distance.

[0050] When calculating the real-time deviation angle between the flow direction angle and the movement direction of the rolling roller, the reference direction is determined by physical calibration: during the linear motion stage of the rolling roller, the displacement direction of a fixed reference point in the image is taken as the reference direction angle. The deviation angle is calculated as the absolute value of the difference between the flow direction angle and the reference direction angle, and when the difference is greater than 90 degrees, 180 degrees is subtracted from the difference and then the absolute value is taken. The calculation result precision is 0.1 degrees, and the angle value always remains in the 0 to 90 degree interval.

[0051] The operation of correcting the real-time deviation angle according to the relative strain gradient is implemented in steps: first, determine the compensation coefficient, which is calibrated by rubber deformation experiment to be a fixed value in the range of 0.25 to 0.35, preferably 0.3. Then calculate the compensation amount, multiply the relative strain gradient value by the compensation coefficient, and keep 3 decimal places. Finally, perform the compensation operation: when the relative strain gradient is positive, the corrected deviation angle is equal to the original deviation angle minus the compensation amount; when it is negative, it is equal to the original deviation angle plus the absolute value of the compensation amount. When the correction result is less than 0, it is automatically set to zero, and when it is greater than 90 degrees, it is set to 90 degrees. The corrected real-time deviation angle is output to the subsequent processing unit, and the compensation operation log is recorded at the same time.

[0052] The width setting of the rectangular analysis region introduces a material parameter adaptive mechanism: when the rubber Shore hardness is less than 50HA, the region width is expanded to 3mm; when the hardness is between 50HA and 70HA, it is 2mm; when it is higher than 70HA, it is reduced to 1.5mm. The region position is updated in real time with the movement of the chip, and the update delay does not exceed 2ms.

[0053] The texture main direction extraction adopts a multi-level screening strategy: in the sub-region main direction statistics stage, only the pixels with a gradient amplitude greater than 50 gray levels are reserved to participate in the calculation. In the final flow direction angle calculation, the data of the sub-regions deviating from the overall average direction by more than 15 degrees are discarded. The direction histogram statistical interval width is set to 5 degrees, and the peak value detection requires that the frequencies of adjacent three intervals decrease monotonously.

[0054] The real-time deviation angle calculation increases dynamic reference correction: when it is detected that the curvature radius of the movement track of the rolling roller is less than 5 meters, the reference direction is adjusted in real time according to the curvature. The adjustment amount is the negative value of the direction angle of the curvature center and the chip connecting line, and the curvature is obtained by fitting a circle through the positions of the rolling roller in three continuous frames.

[0055] Multiple protection mechanisms are set in the compensation link: the compensation coefficient is linearly adjusted according to the temperature of the rubber blank, and the coefficient decreases by 0.02 for each 10-degree Celsius increase in temperature. The maximum compensation amount is limited to 30% of the original deviation angle. When the absolute value of the relative strain gradient is less than 0.01, the compensation link is skipped, and the original deviation angle is directly output. The compensation result is subjected to sliding average filtering: the arithmetic average value of the current frame and the previous four frames of data is taken as the final output, and when the single-frame fluctuation exceeds 5 degrees, the median filtering is used.

[0056] S5, when the corrected real-time deviation angle exceeds a preset critical angle, or the ratio of the tangential displacement component to the relative strain gradient exceeds a preset ratio threshold, it is determined that the chip has undergone irreversible deviation, specifically: The corrected real-time deviation angle, the tangential displacement component and the relative strain gradient data from the previous processing link are received. The corrected real-time deviation angle is in units of degrees, the tangential displacement component is in units of millimeters, and the relative strain gradient is a dimensionless value. The data is input through a double-buffering zone transmission mechanism, and the sampling rate is synchronized with the image acquisition frame rate, for example, the data set is updated every 0.01 seconds. The data validity verification includes range check and continuity check: the deviation angle exceeds the range of 0-90 degrees, the absolute value of the tangential displacement component is greater than 10 millimeters, or the absolute value of the strain gradient is greater than 5, which triggers the data invalidity flag.

[0057] The operation of comparing the corrected real-time deviation angle with the preset critical angle in real time is specifically implemented as follows: the preset critical angle is determined through a rubber creep experiment, and the experimental method is to measure the minimum deviation angle when the chip undergoes permanent displacement under typical working conditions, and the typical value is set to 5 degrees. The comparison process uses a window comparator logic: when the real-time deviation angle exceeds the preset critical angle for 3 frames in a row, a first trigger signal is generated. The comparison result is marked with a time stamp, and the time synchronization accuracy is 0.1 milliseconds.

[0058] When calculating the quotient of the tangential displacement component and the corresponding strain gradient, data preprocessing is performed: when the absolute value of the corresponding strain gradient is less than 0.01, it is replaced by 0.01 to avoid division by zero error. The quotient calculation formula is expressed as: the numerical value of the tangential displacement component is divided by the numerical value of the corresponding strain gradient, and the calculation result is kept to three decimal places. The physical meaning of the quotient represents the tangential displacement amount caused by a unit strain gradient, with a unit of millimeters per strain unit. The calculation process uses a floating-point operation unit to accelerate, and the time consumption of a single calculation is less than 0.5 milliseconds.

[0059] When comparing the quotient with the preset ratio threshold in real time, the preset ratio threshold is calibrated by material mechanical property test. The calibration method is: measuring the ratio of shear strength and elastic modulus of the rubber compound on a universal testing machine, multiplying by a safety factor of 0.8, and the typical threshold range is 0.5-1.2 millimeters per strain unit. The comparison logic is: when the quotient of two consecutive frames exceeds the threshold, a second trigger signal is generated. The threshold is dynamically compensated according to the temperature of the rubber compound: for every 10 degrees Celsius increase in temperature, the threshold is reduced by 5%.

[0060] When any of the following conditions is met, an irreversible deviation flag is generated: the first trigger signal is in an active state, or the second trigger signal is in an active state. After the flag is generated, a delay confirmation mechanism is started: when the flag state is maintained for 5 frames without jumping, the irreversible deviation is finally confirmed to be established. The flag output uses binary encoding, 0 represents normal state, and 1 represents irreversible deviation state, which is transmitted to the actuator through an optoelectronic isolation circuit.

[0061] In the critical angle setting link, a temperature-angle correspondence table is established: when the temperature of the rubber blank is in the range of 70-90 degrees Celsius, the critical angle decreases by 0.3 degrees per 5 degrees Celsius increase; when the temperature is lower than 70 degrees Celsius, a fixed value of 5 degrees is used. The critical angle update period is 1 second, which is updated in real time through the temperature sensor.

[0062] The quotient calculation adds filtering processing: the quotient sequence of the last 5 frames is subjected to amplitude limiting and average filtering, and after removing abnormal values exceeding ±30% of the average value, the arithmetic mean is taken. When the sign of the strain gradient changes is detected, the current quotient does not participate in the comparison, and it is restored after the data is stable.

[0063] The double-channel trigger logic implements interlocking protection: when the first trigger signal and the second trigger signal are active at the same time, an emergency priority processing is started, and the flag is directly output without delay confirmation. The trigger signal state is reset every frame, and it needs to continuously meet the conditions to maintain the active state.

[0064] After the flag is generated, the associated position information is stored: the rolling position coordinates, chip temperature and pressure value at the triggering time are recorded. The storage format is a structure data packet, which includes time stamp (millisecond level), position coordinates (0.01 millimeter precision), temperature value (0.1 degree Celsius precision) and pressure value (0.1 kilo Newton precision).

[0065] Output link set hardware watchdog protection: automatic reset when the flag signal is continuously output for more than 500 milliseconds, preventing system deadlock from causing false downtime. All decision processes are attached with running log records, which include original input data, intermediate calculation results, and decision logic paths.

[0066] S6, if irreversible deviation is detected before the action of the rolling mechanism ends, immediately trigger the stop command, specifically: Real-time monitoring of the rolling mechanism movement position is achieved through multi-sensor fusion: an incremental rotary encoder is installed on the rolling roller shaft, with a resolution of 10,000 pulses per revolution; a magnetic grating ruler is arranged on the vertical guide rail of the machine frame, with a measurement accuracy of 0.001 millimeters. The movement position is calculated as a three-dimensional coordinate value: the horizontal position is obtained by multiplying the number of encoder pulses by the roller circumference, the vertical position is directly read by the magnetic grating ruler, and the axial position is measured by a laser range finder. The coordinate update frequency is 1000 Hz, and the position data is transmitted in a structure form, including four fields: time stamp (millisecond level), X coordinate (horizontal displacement), Y coordinate (vertical displacement), and Z coordinate (axial displacement).

[0067] The determination of the preset safety intervention interval is based on the process safety specification: defining a specific percentage range of the vertical direction travel as the effective intervention window. The specific implementation is: when the vertical position is in the 85% to 98% range of the total down pressure travel, the response enable is activated, and this range is determined through calibration experiments. The calibration method is: simulate the influence of different interruption positions on tire quality on the safety test bench to determine the latest intervention point that does not affect the integrity of the product. The interval boundary values are stored in the non-volatile memory and can be adjusted by the human-machine interface according to the tire specifications, for example, 85% to 95% for passenger car tires and 90% to 98% for truck tires. The response enable signal is a Boolean quantity, which is true when the position coordinate enters the interval and automatically resets when it leaves the interval.

[0068] The process of receiving the irreversible deviation flag and performing the decision contains a hardware cascade lock: the irreversible deviation flag is input through an optical coupling isolation circuit, and the signal high level lasting more than 2 milliseconds is determined as valid. When the response enable is in the valid state and the rising edge of the flag signal is detected, a 32-bit emergency stop instruction code is immediately sent to the motion controller. The instruction transmission uses the industrial real-time Ethernet protocol, and the transmission delay is compensated within 0.5 milliseconds. Before sending, three verifications are performed: verifying whether the duration of the flag signal is greater than 3 control periods; verifying whether the position coordinate is still in the safety interval; verifying whether the system clock synchronization state is normal.

[0069] Emergency stop command triggers the dynamics control of the emergency retraction of the rolling roller: after receiving the emergency stop command, the servo driver starts the preset S-shaped acceleration and deceleration curve. The retraction process is divided into three stages: in the first stage, the action is braked at an acceleration of 5 meters per second squared, and the duration is 20 milliseconds; in the second stage, the acceleration is reversed at 10 meters per second squared, and the speed rises to 0.5 meters per second; in the third stage, the deceleration is switched to 1 meter per second squared at a distance of 50 millimeters from the top dead center, and the final positioning accuracy is ±0.2 millimeters. The top dead center position is double-checked by a mechanical limit switch and an encoder zero point, and is automatically calibrated after each power-on.

[0070] Execution mechanism for clearing the detection state flag: when the rolling roller is 3 millimeters away from the surface of the rubber blank, the state reset operation is triggered. The reset range includes all intermediate process variables: the dynamic displacement vector register is reset to zero; the strain gradient history buffer is cleared; the offset flag is set to invalid. The reset signal is sent to all processing units simultaneously, and the state word is updated through the shared memory area. The original image data of the current rolling period is marked as overwritable, but the data 500 milliseconds before and after the abnormal trigger time is automatically transferred to an independent storage area.

[0071] Dynamic compensation of the safety intervention interval: when the detected rubber temperature is lower than 70 degrees Celsius, the lower limit of the intervention interval is automatically increased to 90%, preventing the low-temperature rubber from being interrupted prematurely. The interval parameters are automatically checked every 5 rolling periods, and the checking method is to simulate the trigger test to verify whether the emergency stop response time is less than 80 milliseconds.

[0072] Redundant channel setting for emergency stop command transmission: the main channel uses real-time Ethernet, and the backup channel uses hard-wired relays. When the main channel transmission delay exceeds 2 milliseconds, the backup channel is automatically switched. All command sending records are recorded in the black box memory, including command content, sending time, channel selection state, and other information.

[0073] Safety monitoring of the retraction process: real-time monitoring of the deviation of the motor current from the theoretical curve, triggering secondary protection when the deviation exceeds 15%. The secondary protection action is to cut off the main power supply of the servo system and enable the pneumatic backup system to complete the retraction. The retraction to position signal is confirmed by position and pressure dual criteria: vertical position error less than 0.5 millimeters and pressure sensor reading less than 5 Newton.

[0074] Abnormal handling of state clearing: when the reset operation fails (such as memory write protection), the forced clearing program is started. The program executes in sequence: sets the key state variables to the initial value; updates the circular buffer pointer to the starting address; writes the reset completion flag to the log file. After forced clearing, a diagnostic report is sent to the remote monitoring terminal.

[0075] Example 2: Figure 2The application provides a structure diagram of a tire chip installation position deviation real-time detection system based on vision. An image acquisition module is configured to continuously acquire real-time images of the chip region when the rolling mechanism starts to act. A displacement calculation module is configured to extract contour feature points of the chip and preset mark points on the rubber surface in the real-time images, and generate a dynamic displacement vector of the chip in the rolling process based on the relative positions of the contour feature points and the preset mark points. A strain analysis module is configured to track displacement trajectories of preset tread pattern units around the chip under the action of the dynamic displacement vector, and generate a relative strain gradient between the tread pattern units and the chip. A deviation correction module is configured to decompose the dynamic displacement vector into a normal displacement component and a tangential displacement component, extract a flow direction angle of a natural texture of the rubber surface in the chip edge contact area, calculate a real-time deviation angle between the flow direction angle and the movement direction of the rolling roller, and correct the real-time deviation angle in association with the relative strain gradient. An offset determination module is configured to determine that the chip has an irreversible offset when the corrected real-time deviation angle exceeds a preset critical angle or a ratio of the tangential displacement component to the relative strain gradient exceeds a preset ratio threshold. A safety control module is configured to trigger a stop command immediately if irreversible offset is detected before the rolling mechanism stops acting.

[0076] The calculations involved in the embodiments are all dimensionless numerical calculations, and preset parameters and threshold values in the calculations are set by those skilled in the art according to actual conditions.

[0077] The above embodiments can be realized wholly or partially by software, hardware, firmware or any combination thereof. When realized by software, the above embodiments can be realized in the form of a computer program product wholly or partially.

[0078] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application of the technical solutions and the constraints of the application. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0079] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically independently, or two or more modules can be integrated in one module.

[0080] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the division of the above-described device embodiment is only a logical function division, and there can be another division manner for actual implementation, for example, multiple devices or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different parts can be indirect couplings or communication connections through some interfaces, devices or modules, and can be in electrical, mechanical or other forms.

[0081] The above describes only the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any modification or replacement within the technical range disclosed by the present application can be easily thought by any person skilled in the art, and should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0082] Finally: the above described is only the preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A real-time detection method for tire chip installation position deviation based on vision, characterized in that: The steps include: S1. When the crushing mechanism starts to move, continuously collect real-time images of the chip area; S2. Extracting the contour feature points of the chip and the preset marking points on the surface of the rubber blank in the real-time image, and generating the dynamic displacement vector of the chip during the rolling process based on the relative positions of the contour feature points and the preset marking points; S3, tracking the displacement trajectory of the preset tread pattern unit around the chip under the action of the dynamic displacement vector, and generating a relative strain gradient between the tread pattern unit and the chip; S4. Decompose the dynamic displacement vector into normal displacement components and tangential displacement components, extract the flow angle of the natural texture of the rubber blank surface in the chip edge contact area, and calculate the real-time deviation angle between the flow angle and the movement direction of the rolling roller. At the same time, the real-time deviation angle is corrected by correlating it with the relative strain gradient. S5. When the corrected real-time deviation angle exceeds a preset critical angle, or the ratio of the tangential displacement component to the relative strain gradient exceeds a preset ratio threshold, it is determined that the chip has irreversible offset; S6. Before the rolling mechanism completes its operation, if irreversible deviation is detected, a stop command is immediately triggered.

2. The method for real-time detection of tire chip installation position deviation based on vision according to claim 1, characterized in that: When the crushing mechanism starts to move, real-time images of the chip area are continuously collected, including: An optical lens is fixedly installed in the axial through hole of the rolling roller so that the optical axis of the lens is vertically aligned with the chip mounting area; Based on the downward displacement of the rolling roller, the coaxial light source is turned on and off in real time to make the irradiation direction of the light source parallel to the optical axis; Dynamically adjust the image acquisition frame rate according to the linear speed of the rolling roller, and perform motion blur correction on the acquired images; Output the corrected image sequence as a live image.

3. The method for real-time detection of tire chip installation position deviation based on vision according to claim 1, characterized in that: Extract the contour feature points of the chip in the real-time image and the preset marking points on the surface of the rubber blank, and generate the dynamic displacement vector of the chip during the rolling process based on the relative positions of the contour feature points and the preset marking points, including: Receive multiple frames of real-time images, perform edge enhancement on each frame, and extract the chip outline pixel set and the pre-set marker pixel set on the surface of the rubber blank; Identify corner features in the chip contour pixel set as contour feature points, and locate centroid coordinates in the preset marker pixel set as marker point positions; The relative offset between the contour feature points and the marker points between adjacent frames is tracked, and a dynamic displacement vector is generated based on the vector superposition of the relative offset in the rolling time sequence.

4. The method for real-time detection of tire chip installation position deviation based on vision according to claim 3, characterized in that: Vector superposition based on relative offset on rolling time sequence includes: Establish a displacement accumulation coordinate system with the start time of rolling as the time zero; Convert the relative offset between the contour feature points and the marker points between adjacent frames into a two-dimensional vector in the displacement accumulation coordinate system; Perform vector addition operation on the two-dimensional vectors of consecutive frames according to the rolling time sequence, and output the accumulated vector as the dynamic displacement vector.

5. The method for real-time detection of tire chip installation position deviation based on vision according to claim 1, characterized in that: Track the displacement trajectory of the preset tread pattern unit around the chip under the action of the dynamic displacement vector, and generate the relative strain gradient between the tread pattern unit and the chip, including: Select at least three tread pattern units at equal angles on the periphery of the chip as monitoring points, and obtain the initial spatial coordinates of the monitoring points; Based on the dynamic displacement vector, the coordinates of the monitoring points are compensated for the rigid displacement and the non-rigid deformation of the monitoring points is separated; Calculate the real-time change rate of the distance vector between each monitoring point and the chip center, and take the maximum value of the real-time change rate as the local strain representation value of the current frame; The spatial distribution curve of the local strain characterization value is fitted along the moving direction of the rolling roller, and the slope of the curve is calculated as the relative strain gradient.

6. The method for real-time detection of tire chip installation position deviation based on vision according to claim 5, characterized in that: Fitting the spatial distribution curve of the local strain characterization value along the rolling roller movement direction and calculating the slope of the curve include: A spatial rectangular coordinate system is established with the chip center as the origin and the rolling roller movement direction as the X-axis; Map the local strain representation value of each monitoring point to the corresponding coordinate position on the X-axis; Perform least square linear fitting on the coordinate position and strain value; The linear coefficient of the fitted straight line equation is extracted as the relative strain gradient.

7. The method for real-time detection of tire chip installation position deviation based on vision according to claim 1, characterized in that: The dynamic displacement vector is decomposed into normal displacement components and tangential displacement components. The flow angle of the natural texture of the rubber surface in the chip edge contact area is extracted, and the real-time deviation angle between the flow angle and the rolling roller movement direction is calculated. At the same time, the real-time deviation angle is corrected by correlating it with the relative strain gradient, including: A local coordinate system is established based on the rolling roller's moving direction, and the dynamic displacement vector is projected onto the normal axis and tangential axis to generate the normal displacement component and the tangential displacement component. A rectangular analysis area is defined in the chip edge contact area, and the main direction angle of the natural texture on the surface of the rubber blank in the rectangular analysis area is extracted as the flow angle; Calculate the absolute value of the angle between the flow angle and the moving direction of the rolling roller as the real-time deviation angle; The real-time deviation angle is linearly compensated according to the relative strain gradient: the relative strain gradient is multiplied by the compensation coefficient and then the difference is made with the real-time deviation angle to output the corrected real-time deviation angle.

8. The method for real-time detection of tire chip installation position deviation based on vision according to claim 1, characterized in that: When the corrected real-time deviation angle exceeds the preset critical angle, or the ratio of the tangential displacement component to the relative strain gradient exceeds the preset ratio threshold, it is determined that the chip has irreversible deviation, including: Receive the corrected real-time deviation angle, tangential displacement component and relative strain gradient; Compare the corrected real-time deviation angle with the preset critical angle in real time; Synchronously calculate the quotient of the tangential displacement component and the relative strain gradient, and compare the quotient with a preset ratio threshold in real time; When the corrected real-time deviation angle is greater than a preset critical angle, or the quotient is greater than a preset ratio threshold, an irreversible offset flag is generated.

9. The method for real-time detection of tire chip installation position deviation based on vision according to claim 1, characterized in that: Before the rolling mechanism completes its action, if irreversible deviation is detected, a shutdown command will be triggered immediately, including: Real-time monitoring of the rolling mechanism's movement position, activating the response enable when the movement position is within the preset safety intervention range; Receive the irreversible offset flag, and when the response enable is valid, if the irreversible offset flag is true, send an emergency stop command to the servo driver of the rolling mechanism; The emergency stop command triggers the rolling roller to urgently retract to the top dead center position; Clear all detection status flags in the current rolling cycle.

10. A vision-based real-time detection system for tire chip installation position deviation, used to implement the vision-based real-time detection method for tire chip installation position deviation according to any one of claims 1 to 9, characterized in that: Includes the following modules: An image acquisition module is used to continuously acquire real-time images of the chip area when the rolling mechanism starts to move; The displacement calculation module is used to extract the contour feature points of the chip and the preset marking points on the surface of the rubber blank in the real-time image, and generate the dynamic displacement vector of the chip during the rolling process based on the relative positions of the contour feature points and the preset marking points; A strain analysis module is used to track the displacement trajectory of the preset tread pattern units around the chip under the action of the dynamic displacement vector and generate the relative strain gradient between the tread pattern units and the chip; Deviation correction module, used to decompose the dynamic displacement vector into normal displacement component and tangential displacement component, extract the flow angle of the natural texture of the rubber blank surface in the chip edge contact area, and calculate the real-time deviation angle between the flow angle and the movement direction of the rolling roller. At the same time, the real-time deviation angle is corrected by correlating it with the relative strain gradient. An offset determination module is used to determine that an irreversible offset has occurred in the chip when the corrected real-time deviation angle exceeds a preset critical angle, or when the ratio of the tangential displacement component to the relative strain gradient exceeds a preset ratio threshold; The safety control module is used to immediately trigger a shutdown command if irreversible deviation is detected before the rolling mechanism completes its action.

Images