Linear motion module accurate control method and system based on AI vision

Abstract

The invention relates to the technical field of image recognition control, in particular to a linear motion module accurate control method and system based on AI vision, and the method comprises the following steps: obtaining an interference fringe image and extracting a gray path, analyzing central point displacement to generate a coordinate sequence, repairing an interference feature position, and aligning the image and a control time sequence to generate a synchronization time table. The mapping structure response forms a set of control output configurations. According to the method, a peak path is constructed through interference fringe gray scale information in an image frame, central point pixel displacement is extracted, a coordinate migration sequence is constructed, fine-grained capture and tracking of an execution state are realized, an external interference area is identified according to image gradient change, vector repair is carried out, and image data stability and feature reliability are guaranteed. And aligning the repaired image response with a control instruction time sequence, constructing an impulse response lag statistical mechanism, realizing accurate mapping adjustment of the control frequency, and improving the control precision and response consistency in high-frequency displacement execution.

Description

Technical Field

[0001] This invention relates to the field of image recognition and control technology, and in particular to a precise control method and system for linear motion modules based on AI vision. Background Technology

[0002] Image recognition control technology involves acquiring image information of the external environment or target objects through image sensors, analyzing and recognizing the image content using image recognition algorithms, and using the recognition results as the basis for control commands to achieve closed-loop control of the system or equipment. Core aspects of this technology include image acquisition and preprocessing, feature extraction and recognition, target localization and tracking, and visual feedback control. Image recognition control is widely used in industrial automation, robot navigation, intelligent manufacturing, traffic monitoring, and other scenarios. It possesses high-precision perception and dynamic response capabilities and is an interdisciplinary technology system integrating artificial intelligence, image processing, and control engineering, playing a crucial role, especially in tasks requiring high-precision control based on visual feedback. Traditional linear motion module precision control methods involve acquiring motion state information through position detection devices such as encoders and limit switches, and adjusting the output of the drive motor in conjunction with a preset control program to control linear motion components such as slides to move along a specified trajectory. In this process, position detection mainly relies on hardware sensors, and the control accuracy is limited by the sensor resolution and the precision of the mechanical structure. Traditional methods use PID control to adjust the motor speed or step frequency in real time based on the position signal returned by the sensor to achieve the target displacement. However, this method has problems such as control response lag, susceptibility to interference, and lack of environmental adaptability.

[0003] Existing technologies rely on hardware sensors such as encoders and limit switches to obtain position status. The sensing accuracy is limited by the sensor resolution and the fit of the mechanical structure, resulting in insufficient detection capability in high-frequency response or small displacement scenarios. The position information update frequency is limited, making it difficult to achieve high-precision control of fast-moving parts. The control method relies on PID feedback regulation, which is prone to signal lag and error accumulation in dynamic response. It cannot respond to sudden changes in the target state in a timely manner, and its adaptability to external disturbances in strong interference environments is weak. It lacks the means to utilize multi-dimensional image information during the motion process, making it difficult to achieve real-time accurate identification and adjustment of the actuators, thus restricting the closed-loop control performance of linear motion devices. Summary of the Invention

[0004] To address the technical problems existing in the prior art, embodiments of the present invention provide a precise control method for linear motion modules based on AI vision, comprising the following steps: S1: Acquire the interference fringe image irradiated on the surface of the high-speed slide actuator, acquire continuous image frames in the image processing unit, extract the gray-scale change region of the interference fringes in the image frame, identify the central axis of the interference fringes in the image, and obtain the interference peak gray-scale path set; S2: Call the interference peak grayscale path set, extract the horizontal and vertical pixel positions of the interference center point in the image coordinate system of the continuous frame, and make a difference judgment based on the horizontal change distance of the center point coordinates between the real frame and the previous frame and the sampling time interval to obtain the sliding table execution coordinate migration sequence. S3: Based on the coordinate migration sequence of the slide, identify the gray gradient direction value corresponding to the edge of the high-speed slide structure in the image frame and the brightness distribution of the boundary region, calculate the consistency of the gray gradient direction of pixels in consecutive frames at the boundary, and obtain the feature location set after interference repair. S4: Using the feature position set after interference repair, align the repaired position difference with the timing of the control pulse signal. Based on the time difference between the command effective time and the real-time image response delay, statistically analyze the response lag change trend in adjacent control cycles to obtain the control pulse synchronization timetable.

[0005] As a further embodiment of the present invention, the interference peak grayscale path set includes grayscale distribution peak position, fringe path pixel trajectory, and inter-frame fringe morphology change; the slider execution coordinate migration sequence includes coordinate change trajectory, inter-frame displacement change value, and pixel motion trend; the interference-corrected feature position set includes edge vector offset value, feature point brightness correction parameter, and gradient direction correction index; and the control pulse synchronization time table includes control cycle identifier, time delay change curve, and synchronization alignment difference.

[0006] As a further aspect of the present invention, the specific steps of S1 are as follows: S101: Acquire the interference fringe image irradiated on the surface of the high-speed slide actuator, call the continuously acquired image frame sequence in the image processing unit, perform linear traversal on the gray values ​​of the pixel columns in the image frame, perform curve fitting on the gray value difference sequence of adjacent pixels, extract the position index of the gray value change area, and generate a set of gray value change regions of interference fringes. S102: Based on the set of gray-scale abrupt change regions of the interference fringes, the coordinates of the center points between adjacent abrupt change regions in each frame image are aggregated by mean, and the position sequence of the interference center line in the two-dimensional matrix of the image frame is constructed. Then, linear fitting is performed on the longitudinal coordinate points of the center line in the image frame to generate the axial path sequence of the center of the interference fringes. S103: Call the central axis path sequence of the interference fringes, extract the gray-scale peak points along the central axis path coordinates of each frame in the continuous image frames, and aggregate the gray-scale peak points into multiple sets of homologous sequences according to the image frame acquisition time sequence index. Number each set of sequences by path to obtain the interference peak gray-scale path set.

[0007] As a further aspect of the present invention, the specific steps of S2 are as follows: S201: Call the interference peak grayscale path set, extract the horizontal and vertical pixel indices of the grayscale peak points corresponding to consecutive frames on each path in the image coordinate system, arrange the peak point coordinates of consecutive frames in the same path in a time series, and perform structured processing on the arranged coordinate sequence to generate the inter-frame interference center point pixel position sequence. S202: Based on the inter-frame interference center point pixel position sequence, the ratio of the distance change value of the center point in the horizontal pixel coordinate direction in adjacent frames to the image sampling time interval is calculated to obtain the pixel displacement increment per unit time. According to the set horizontal displacement judgment threshold of two pixel units, the points in the continuous frames whose displacement increment exceeds the threshold are indexed and recorded to obtain the continuous horizontal pixel mutation point sequence. S203: Call the continuous horizontal pixel mutation point sequence, extract the image horizontal and vertical coordinate values ​​corresponding to multiple points, and map them to the execution space coordinate system of the high-speed slide table. Arrange the mapped coordinates in time series to generate the slide table execution coordinate migration sequence.

[0008] As a further aspect of the present invention, the specific steps of S3 are as follows: S301: Call the slider to execute the coordinate migration sequence, extract the slider edge region image block in the image frame corresponding to the execution coordinate, perform convolution processing on the pixels in the slider edge region image block, obtain the gray-level gradient component values ​​in the horizontal and vertical directions, and traverse the brightness value of the edge region pixel by pixel to obtain the gray-level gradient direction value and brightness distribution set of the edge region. S302: Based on the gray gradient direction value and brightness distribution set of the edge region, calculate the angle difference of the gray gradient direction of the boundary region at the same position in consecutive image frames, and count the proportion of pixels whose absolute value of the angle difference is less than the preset angle consistency threshold. Perform density statistics on pixels whose gray change is greater than the brightness change detection threshold to obtain the edge interference feature point region index set. S303: Call the edge interference feature point region index set, perform vector translation operation on the region image coordinates, correct the coordinates based on the center position difference of adjacent non-interference regions, update the position of the coordinate correction points, identify the position distribution after continuous correction, and generate the interference repair feature position set.

[0009] As a further aspect of the present invention, the specific steps of S4 are as follows: S401: Call the interference-corrected feature position set, extract the position difference of the same feature point in consecutive image frames within the control cycle, arrange the position difference sequence according to the image acquisition timestamp, obtain the timing mark of the control pulse signal issued by the controller, match the corrected position change with the control pulse time axis, and generate a position response and control command alignment sequence. S402: Based on the alignment sequence of the position response and control command, extract the command issuance time and the acquisition time of the corresponding position change response frame within the control cycle, calculate the difference between their timestamps as the image response lag time, and perform control cycle serialization processing on the lag time to generate a control cycle response lag change trend set. S403: Call the control cycle response lag change trend set, perform window smoothing on the time difference value in the lag change sequence, and establish an equally spaced pulse number index sequence in combination with the original control command time axis. Match and record the response delay of the control cycle with the pulse index to obtain the control pulse synchronization time table.

[0010] As a further aspect of the present invention, the process of performing window smoothing on the time difference value in the hysteresis change sequence specifically involves calculating the average value of the image response hysteresis time within three adjacent control cycles in the control cycle response hysteresis change trend set, and obtaining the corresponding smoothed time difference value sequence. The process of establishing an equally spaced pulse number index sequence by combining the original control command time axis is specifically as follows: taking the timestamp of the timing mark of the control pulse signal issued by the controller as the starting reference point, and numbering and indexing are performed according to a fixed time interval. The process of matching and recording the response delay of the control cycle with the pulse index specifically involves recording the smoothed time difference value at the time axis position corresponding to the numbered index.

[0011] As a further aspect of the present invention, the method further includes step S5: S5: Call the timing configuration corresponding to the key control cycle of the high-speed slide table in the control pulse synchronization time table, match the load peak change trend in the real-time slide table actuator operation record, combine the deformation range and limit threshold of the high-speed slide table body material, determine the structural response range within the corresponding control cycle, map the structural response range to the controller output pulse frequency curve, and obtain the slide table control output configuration set. The slide control output configuration set includes a control frequency parameter group, a structural response mapping table, and a key cycle control scheme.

[0012] As a further aspect of the present invention, the specific steps of S5 are as follows: S501: Call the control pulse synchronization time table, extract the pulse timing configuration interval marked as the key control cycle of the high-speed slide table, sort and calibrate the control pulse number and time point within the pulse timing configuration interval, extract the load peak change trend data within the corresponding time period from the operation record of the slide table actuator, perform interval correspondence after normalization of the time axis, and generate the load trend alignment sequence of the key control cycle. S502: Based on the load trend alignment sequence of the key control cycle, extract the load peak time period in the corresponding control cycle, compare the load peak time period with the deformation range limit threshold of the high-speed slide table body material, filter the continuous time periods where the load peak exceeds the deformation range limit threshold, and aggregate them into active structural response intervals according to time order to obtain a set of structural response interval ranges. S503: Call the set of structural response intervals, map the response intervals to the corresponding time intervals on the controller pulse output frequency curve, and perform statistics and interval labeling on the pulse frequency change gradient within the mapped segment to generate a slide control output configuration set.

[0013] The AI ​​vision-based linear motion module precision control system includes: The image acquisition module illuminates the interference fringe image on the surface of the high-speed slide actuator, detects the change range of gray-level distribution area in the image frame, identifies the central axis of the fringe formed by the continuity of gray-level differences in the image, locates the coordinates of the gray-level peak point on the same interference fringe in consecutive frames, and generates a set of interference peak gray-level paths. The pixel trajectory module extracts the horizontal and vertical pixel positions of the interference center point in the image coordinate system based on the interference peak grayscale path set, obtains the pixel offset value of the center point in the horizontal axis and the frame sampling time interval between the real-time frame and the previous frame, classifies and calibrates the sliding direction, and generates a sliding table execution coordinate migration sequence. The coordinate migration module performs a coordinate migration sequence based on the slide, extracts the gray distribution curve of the high-speed slide structure edge contour in the image frame, obtains the gray gradient direction value of continuous pixels at the edge position, performs consistency judgment on the gray gradient direction of the inter-frame boundary region, records the coordinate range of edge feature points affected by illumination disturbance and structural deformation, and generates a feature position set after interference repair. The feature repair module calls the interference-repaired feature location set, obtains the matching items between the location point sequence and the timestamp of the control pulse signal issued by the controller, calculates the time difference between the instruction effective time and the position response frame in the image within three consecutive frames, identifies the response lag change information in the control cycle, constructs the correspondence between the lag change coefficient and the control cycle, and generates a control pulse synchronization timetable. Based on the control pulse synchronization schedule, the timing configuration module extracts the control pulse signal sequence corresponding to the key control cycle, calls the load peak change curve of the corresponding cycle in the high-speed slide actuator operation record, compares the load peak change curve with the deformation threshold range of the slide body material, and obtains the slide control output configuration set.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, a peak path is constructed by using the grayscale information of interference fringes in an image frame, the displacement of the center point pixel is extracted and a coordinate migration sequence is constructed, thereby achieving fine-grained capture and tracking of the execution state. External interference areas are identified and vector repair is performed based on image gradient changes to ensure image data stability and feature reliability. The timing of the repaired image response is aligned with the control command, and a pulse response hysteresis statistical mechanism is constructed to effectively correct the time deviation between the control rhythm and visual feedback. The execution structure response interval is established by combining the load peak trend and the material deformation range, thereby achieving precise mapping and adjustment of the control frequency and improving the control accuracy and response consistency in high-frequency displacement execution. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention; Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] Please see Figure 1 This invention provides a precise control method for a linear motion module based on AI vision, comprising the following steps: S1: Acquire the interference fringe image irradiated on the surface of the high-speed slide actuator, acquire continuous image frames in the image processing unit, extract the gray-scale change area of ​​the interference fringes in the image frame, identify the central axis of the interference fringes in the image, arrange the corresponding gray-scale peak points of the same fringe in the continuous frames in frame order, and obtain the interference peak gray-scale path set. S2: Call the interference peak grayscale path set, extract the horizontal and vertical pixel positions of the interference center point in the image coordinate system of the continuous frame, judge the difference based on the horizontal change distance of the center point coordinate between the real frame and the previous frame and the sampling time interval, filter the continuous point sequence with displacement change exceeding two pixel units, calculate the execution coordinates corresponding to the high-speed slide table based on the projection of the movement trajectory formed by the center point in the image, and obtain the slide table execution coordinate migration sequence. S3: Based on the coordinate migration sequence of the slide, identify the gray gradient direction value corresponding to the edge of the high-speed slide structure in the image frame and the brightness distribution of the boundary region, calculate the consistency of the gray gradient direction of pixels in consecutive frames at the boundary, and perform a mutation density judgment on gray abrupt points, identify the feature point region affected by illumination or edge contour deformation interference, and perform vector translation repair operation on the position in the image coordinate to obtain the feature position set after interference repair. S4: By using the feature position set after interference repair, the position difference after repair is aligned with the timing of the control pulse signal issued by the controller. Based on the time difference between the command effective time and the real-time image response delay, the response lag change trend in adjacent control cycles is statistically analyzed to obtain the control pulse synchronization time table. S5: Call the timing configuration corresponding to the key control cycle of the high-speed slide table in the control pulse synchronization time table, match the load peak change trend in the real-time slide table actuator operation record, combine the deformation range and limit threshold of the high-speed slide table body material, determine the structural response range within the corresponding control cycle, map the structural response range to the controller output pulse frequency curve, and obtain the slide table control output configuration set. The interference peak grayscale path set includes the grayscale distribution peak position, fringe path pixel trajectory, and inter-frame fringe morphology changes. The slider execution coordinate migration sequence includes the coordinate change trajectory, inter-frame displacement change value, and pixel motion trend. The interference-corrected feature position set includes the edge vector offset value, feature point brightness correction parameter, and gradient direction correction index. The control pulse synchronization time table includes the control cycle identifier, time delay change curve, and synchronization alignment difference. The slider control output configuration set includes the control frequency parameter group, structural response mapping table, and key cycle control scheme.

[0023] Please see Figure 2 The specific steps of S1 are as follows: S101: Acquire the interference fringe image irradiated on the surface of the high-speed slide actuator, call the continuously acquired image frame sequence in the image processing unit, perform linear traversal on the gray values ​​of the pixel columns in the image frame, perform curve fitting on the gray value difference sequence of adjacent pixels, extract the position index of the gray value change area, and generate a set of gray value change regions of interference fringes. This is achieved by linking an interferometer with a high-speed image acquisition system, such as by installing a Michelson interferometer. A laser source is projected onto the surface of a sliding stage, forming a fringe image through reflection and interference. A high-speed camera continuously captures images of the region at fixed time intervals, generating a continuous sequence of image frames. These image frames are processed in real time. In the image processing flow, for each image frame, a vertical column of pixels is selected for grayscale traversal, that is, the grayscale value of each pixel is read point by point along the vertical direction, forming a continuous grayscale sequence of the entire column of pixel values. Then, the grayscale difference between each pair of adjacent pixels is calculated. The resulting gray-level difference sequence exhibits a certain degree of fluctuation. To identify regions with significant gray-level jumps, a fitting method can be used to smooth the difference curve, eliminating interference fluctuations during image acquisition. By statistically analyzing the abrupt change locations in the fitted curve—that is, points where the gray-level value jump is greater than a certain amplitude—the positional features of the interference fringes can be identified. In practical production applications, such as performing interferometric measurements on a high-speed photolithography slide, the brightness variation bands formed by the interference light in the image can be extracted. This allows for accurate marking of the image row index where the gray-level abrupt changes in the fringes occur, generating a set of gray-level abrupt change regions for the interference fringes.

[0024] S102: Based on the set of gray-scale abrupt change regions of interference fringes, the coordinates of the center points between adjacent abrupt change regions in each frame of the image are aggregated by mean, and the position sequence of the interference center line in the two-dimensional matrix of the image frame is constructed. Then, linear fitting is performed on the longitudinal coordinate points of the center line in the image frame to generate the axial path sequence of the center of the interference fringes. The pixel center points between adjacent abrupt change regions in each frame of the image are identified. The midpoint position can be calculated by pairwise pairing according to the abrupt change point index to obtain the coordinates of the center pixel located between two gray-level abrupt change regions. The midpoint reflects the local center position of the fringe. The center coordinates obtained in each frame of the image are arranged in order to obtain the interference fringe center line of the current frame image. Several frames of images can correspond to multiple center line points. A two-dimensional matrix is ​​constructed frame by frame to record the position change trend of the center line over time. The coordinates of the center point in the image space are continuously linearly fitted to form an approximate axis of the center line. In the image processing flow, the center point coordinates are read frame by frame and the fitting operation is performed. This is suitable for continuous acquisition scenarios of interference images during high-speed sliding table operation. For example, in a set of high-speed acquired images, if the interference fringes are slightly displaced with the movement of the sliding table, the dynamic trajectory of the interference center axis can be depicted by center point aggregation and fitting, generating the axial path sequence of the interference fringe center axis.

[0025] S103: Call the central axis path sequence of the interference fringes, extract the gray value peak points along the central axis path coordinates of each frame in the continuous image frames, and aggregate the gray value peak points into multiple sets of homologous sequences according to the image frame acquisition time sequence index. Number each set of sequences to obtain the interference peak gray value path set. In each frame of the image, gray-scale peak points are extracted along the central axis path coordinates. That is, using the fitted central path as a reference, the gray-scale values ​​at the positions traversed by the path are searched frame by frame, and the point with the largest gray-scale value is recorded as the peak feature of the image frame. Then, combined with the time series index of the image frame, the gray-scale peak points in the image frame are classified, and peak points with similar spatial positions are grouped into the same path group. In the process, peak aggregation judgment is required, that is, to determine whether the peaks of adjacent frames appear in almost the same position. If so, they are grouped together. This is suitable for analyzing the dynamic evolution of interference fringes in the time dimension. For example, in the interferometric imaging of a high-speed sliding stage, the fringe peaks in several frames of images appear in the same position area near the image center. Through time series aggregation, a complete interference peak path can be formed. The path is then numbered and labeled to obtain the interference peak gray-scale path set.

[0026] Please see Figure 3 The specific steps of S2 are as follows: S201: Call the interference peak grayscale path set, extract the horizontal and vertical pixel indices of the grayscale peak points corresponding to consecutive frames on each path in the image coordinate system, arrange the peak point coordinates of consecutive frames in the same path in a time series, and perform structured processing on the arranged coordinate sequence to generate the inter-frame interference center point pixel position sequence. It is necessary to traverse the continuous image frames contained in each path and extract the specific positions of the gray-level peak points in each frame in the image coordinate system, including the pixel indices in both the horizontal and vertical directions. In practice, the peak points of each frame are represented as two-dimensional coordinate points (e.g., the coordinates of the peak point in frame t are xt, yt). The peak points of the image frames are arranged in chronological order, that is, the coordinate points corresponding to t are organized into a time series. In this way, the peak points on each interference path can form a pixel position trajectory that progresses over time. In the image processing device, the sequence is stored in the form of structured data, which can be organized and managed using a two-dimensional array or data table. For example, an index can be established to correspond one-to-one with the pixel coordinates of each time point under the path number. In a high-speed sliding table position measurement task, if the interferometer obtains a specific interference fringe peak path, the image frame number is from t=1 to t=100, and the corresponding extracted peak coordinates are (120, 250), (121, 250), (122, 250)..., generating an inter-frame interference center point pixel position sequence.

[0027] S202: Based on the inter-frame interference center point pixel position sequence, the ratio of the distance change value of the center point in the horizontal pixel coordinate direction in adjacent frames to the image sampling time interval is calculated to obtain the pixel displacement increment per unit time. According to the set horizontal displacement judgment threshold of two pixel units, the points in the continuous frames whose displacement increment exceeds the threshold are indexed and recorded to obtain the continuous horizontal pixel mutation point sequence. The horizontal coordinate difference is calculated for the center point coordinates of any two adjacent image frames, i.e., the difference operation is performed on the horizontal pixel index to obtain the horizontal pixel displacement between frames. Each horizontal displacement value is divided by the image sampling time interval to calculate the pixel displacement increment per unit time. The image sampling time interval is set according to the image acquisition device. For example, if the frame rate is 100fps, the time interval is 0.01 seconds. In the high-speed motion detection task of the sliding table, if the horizontal coordinate change between frame 1 and frame 2 in a certain path is 5 pixels, then the displacement increment per unit time is 5 / 0.01 = 500 pixels / second. According to the preset horizontal displacement change judgment threshold of 2 pixels, the calculated inter-frame displacement increment is judged. If the displacement difference between a pair of frames exceeds the threshold, it is judged as a change point, and the index of the frame where the change point is located is recorded. The entire path sequence is scanned in this way to obtain a continuous horizontal pixel change point sequence.

[0028] S203: Call the continuous horizontal pixel mutation point sequence, extract the corresponding horizontal and vertical coordinate values ​​of the image of multiple points, and map them to the execution space coordinate system of the high-speed slide table. Arrange the mapped coordinates in time series to generate the slide table execution coordinate migration sequence. The horizontal and vertical pixel coordinates corresponding to the abrupt change points are extracted from the image. The image coordinate points are then transformed and mapped to the execution space coordinate system used by the slide table. The image coordinate system is in pixels, while the slide table space coordinate system uses physical units such as millimeters or micrometers. Therefore, the calibration parameters of the image acquisition device need to be referenced during the mapping process, such as the physical distance corresponding to a single pixel (e.g., 0.005 mm / pixel). Multiplying the image coordinates by this scaling factor yields the actual position coordinates on the slide table. The physical coordinate points are arranged in a time sequence according to the image frame index to construct the position change trajectory during the slide table execution process. For example, if the abrupt change points in frames 5 to 10 are mapped to coordinates of 2.1 mm, 2.3 mm, 2.8 mm, 3.0 mm, 3.6 mm, and 4.0 mm on the x-axis of the slide table, respectively, the abrupt change points are sequentially combined by frame to reflect the non-uniform change process of the slide table's position within a certain time period. The coordinate migration sequence is of great reference value for analyzing the acceleration and deceleration, abrupt change points, and execution accuracy in the slide table movement, thus generating the slide table execution coordinate migration sequence.

[0029] Please see Figure 4 The specific steps of S3 are as follows: S301: Call the slider to execute the coordinate migration sequence, extract the slider edge region image block in the image frame corresponding to the execution coordinate, perform convolution processing on the pixels in the slider edge region image block, obtain the gray-level gradient component values ​​in the horizontal and vertical directions, and traverse the brightness values ​​of the edge region pixel by pixel to obtain the gray-level gradient direction value and brightness distribution set of the edge region. For each set of physical coordinates, the image frame corresponding to the image frame is precisely located. Image blocks containing the edge of the slide are extracted from the image. Each image block is a small area of ​​fixed size. For example, a 40×40 pixel edge sub-image is cropped from the image frame with the actual physical coordinates of the slide edge as the center, forming a set of edge image blocks. Each image block is processed at the pixel level. Convolution operation is used to obtain the directional information of gray-level changes in the image. Conventional edge detection convolution kernels, such as the Sobel operator, are selected during the processing. Convolution operations are applied to the horizontal and vertical directions of the image block to extract the gray-level gradient components of each pixel in the two directions. By traversing the convolution results of the pixels, the strength and directional features of gray-level changes in the entire image are obtained. At the same time, the brightness value in the image block is extracted separately. The gray-level brightness of each point is scanned pixel by pixel, and the position, value and corresponding directional gradient are recorded to obtain the gray-level gradient direction value and brightness distribution set of the edge region.

[0030] S302: Based on the gray gradient direction value and brightness distribution set of the edge region, calculate the angle difference of the gray gradient direction of the boundary region at the same position in consecutive image frames, and count the proportion of pixels whose absolute value of the angle difference is less than the preset angle consistency threshold. Perform density statistics on pixels whose gray change is greater than the brightness change detection threshold to obtain the edge interference feature point region index set. Angle consistency analysis is performed on pixels at the same position on the slider edge in adjacent image frames. Specifically, for edge pixels at known positions in each frame, the grayscale gradient direction angle is calculated, and the gradient direction difference is calculated for the same position across consecutive frames. By comparing the directional change angle of each pixel point by point, it is determined whether a drastic directional jump has occurred, and the stability of the boundary contour is analyzed. During this process, a preset angle consistency threshold, such as 10 degrees, is set. Pixels with an absolute angle difference less than the threshold are counted, and then divided by the total number of pixels to obtain the proportion of pixels with consistent grayscale direction. This proportion reflects the consistency of edge changes. The brightness values ​​of pixels are iterated, and pixels with grayscale value changes greater than a set brightness abrupt change detection threshold are selected, for example, a threshold of 30 grayscale units. If the brightness difference at the same pixel position in two consecutive frames exceeds the threshold, it is recorded as a sudden change point. Density statistics are performed on these sudden change points. If the number of abrupt change pixels in a unit area exceeds a set density threshold, the area is marked as having edge interference, and an index set of edge interference feature point regions is obtained.

[0031] S303: Call the edge interference feature point region index set, perform vector translation operation on the region image coordinates, correct the coordinates based on the center position difference of adjacent non-interference regions, update the position of the coordinate correction points, identify the position distribution after continuous correction, and generate the interference repair feature position set. The coordinate correction operation is performed on the image coordinate points within the marked area to identify the position coordinates of each disturbed area. The center point coordinates are extracted from the undisturbed image blocks in the adjacent areas. The difference between the center point positions of the two areas is compared to determine the correction direction and distance of the vector translation. For example, if the center of the disturbed area is located at coordinates (220, 340) in the image, while the center of the adjacent undisturbed area is (225, 345), the coordinates of the current disturbed area pixel are corrected by the translation vector (+5, +5). The corrected point will update its position in the image coordinate system, and the change magnitude is identified by comparing it with the original position. After performing continuous frame-by-frame processing on the corrected point, the feature position distribution of the edge area in the slide image after interference compensation is represented. For example, in a certain image frame, vector correction is performed on five disturbed areas respectively to generate a set of feature positions after interference compensation.

[0032] Please see Figure 5 The specific steps of S4 are as follows: S401: Call the feature position set after interference repair, extract the position difference of the same feature point in consecutive image frames within the control cycle, arrange the position difference sequence according to the image acquisition timestamp, obtain the timing mark of the control pulse signal issued by the controller, match the position change after repair with the control pulse time axis, and generate a position response and control command alignment sequence. In an image sequence, the positional changes of the same feature point across consecutive image frames are tracked. A range of image frames within a control cycle is selected, for example, within a control cycle T, image frames numbered t1 to tN. The image coordinates of the feature point are extracted in each frame, and the displacement difference between the real-time frame and the previous frame is calculated, resulting in a continuous positional difference sequence per frame. This sequence reflects the motion changes of the feature point within the control cycle. This difference sequence is arranged sequentially according to the timestamp of each image frame acquisition to ensure time consistency. Simultaneously, the control pulse signal information recorded in the control device is retrieved, and the timestamps of the pulse commands issued by the controller within the control cycle are compared. Through time matching, a correspondence is established between the displacement changes of the feature point and the specific pulse signals issued by the controller. In a high-speed sliding table positioning system, if the controller continuously issues 5 control pulses within a 10-millisecond interval while capturing 10 images during this period, the correspondence between the positional changes of the feature points in each frame and the pulse issuance time needs to be compared to generate a position response and control command alignment sequence.

[0033] S402: Based on the alignment sequence of position response and control command, extract the command issuance time and the acquisition time of the corresponding position change response frame within the control cycle, calculate the difference between their timestamps as the image response lag time, and perform control cycle serialization processing on the lag time to generate a control cycle response lag change trend set. The time difference between the feature point response time and the controller command time is analyzed. For each matching record, the timestamp difference between the control pulse issuance time and the image frame acquisition time is extracted and defined as the image response lag time. The difference represents the delay in the response of the image processing device to the control command. This operation is repeated to process the pulse commands within a complete control cycle and calculate the corresponding image response lag time one by one. The series of lag times are arranged in pulse order. In some high-speed control scenarios, such as a control cycle of 50 milliseconds and an image acquisition interval of 5 milliseconds, if the response lag time after each control pulse fluctuates between 3 milliseconds and 7 milliseconds, the regularity and instability of the response time evolution with the control cycle can be intuitively displayed by constructing a change sequence. This helps to adjust the time coordination relationship between the image and the control device and generate a set of control cycle response lag change trends.

[0034] S403: Call the control cycle response lag change trend set, perform window smoothing on the time difference value in the lag change sequence, and establish an equally spaced pulse number index sequence in combination with the original control command time axis. Match and record the response delay of the control cycle with the pulse index to obtain the control pulse synchronization time table. Window smoothing is required for each time difference in the lag time series. This operation involves setting a fixed number of time windows to perform a moving average on the lag values, eliminating interference caused by abnormal fluctuations in individual frames. For example, a smoothing window of 5 pulses is set, and the average value of each group of lag times is taken to form a more stable response characteristic sequence. After smoothing the lag sequence, a new set of equally spaced control pulse number index sequences is established based on the time axis of the original control commands, ensuring that each pulse command has a unique number index on the time axis. For example, a pulse every 10 milliseconds can be labeled as P1, P2, P3, etc. Each smoothed response delay time is matched one by one with the corresponding pulse number, which can clearly identify the response lag level of each control command and obtain the control pulse synchronization time table during the high-speed operation of the slide.

[0035] Please see Figure 6 The specific steps of S5 are as follows: S501: Call the control pulse synchronization schedule, extract the pulse timing configuration interval marked as the key control cycle of the high-speed slide table, sort and calibrate the control pulse number and time point within the pulse timing configuration interval, extract the load peak change trend data within the corresponding time period from the slide table actuator's operation record, normalize the time axis and perform interval correspondence, and generate the key control cycle load trend alignment sequence. Identify and extract pulse timing configuration intervals marked as critical control cycles for the high-speed slide table. A critical cycle refers to the time window used by the control equipment when performing high-precision motion or high-speed segments. After selection, sort and label the control pulse numbers and corresponding timestamps within the interval to ensure a one-to-one correspondence between control command times and numbers, forming a clear timing sequence. Extract load peak change trend data over multiple time periods from the slide table actuator's operation log based on the time axis range, recording the actual load fluctuations experienced by the slide table during command execution. To facilitate comparison of the time consistency of different data types, it is necessary to synchronize the control pulse time axis and the load change trend data. The axes are simultaneously normalized to align the intervals on a unified time scale. The alignment method can be to convert the timestamp into a standardized proportional position value, such as between 0 and 1, and then establish a connection between the standardized time point corresponding to each control pulse and the load value with the same normalized time in the load change trend through matching. In a high-speed lithography slide operation record, if the control pulse numbers P10 to P30 constitute the critical interval, corresponding to the time from t=120ms to t=180ms, after normalization, the load trend in this interval can be proportionally mapped to the control pulse sequence to generate a critical control cycle load trend alignment sequence.

[0036] S502: Based on the load trend alignment sequence of the key control cycle, extract the load peak time period in the corresponding control cycle, compare the load peak time period with the deformation range limit threshold of the high-speed slide table body material, filter the continuous time period when the load peak exceeds the deformation range limit threshold, and aggregate them into active structural response intervals according to the time order to obtain the set of structural response interval ranges. Analyzing the load value change trend and extracting the peak time period, i.e., identifying continuous time segments where the load value is significantly higher than the remaining time period, serves as the time window when the slide structure is subjected to greater stress. After extracting the peak time period, it is compared with the deformation range limit threshold of the main material of the high-speed slide. The limit threshold is given by the material performance data. For example, the maximum instantaneous load of a certain aerospace-grade aluminum alloy material should be 25N. If the load peak value of three consecutive sampling points in the load trend sequence exceeds this value, it is judged that the material is in the critical state of plastic deformation. The over-limit segment is screened out and marked as the response abnormal segment. According to the time sequence, the continuous over-limit time segments are aggregated to form the response interval. The aggregated result constitutes the active structural response interval, which indicates the significant structural response of the slide in the critical control cycle due to the superposition of control signals and load response. In actual high-speed slide motion control applications, for example, if the load exceeds 28N in a critical task segment from 130ms to 135ms, and if the over-limit value occurs continuously in this period, the interval is determined as the structural response interval, and the set of structural response interval ranges is obtained.

[0037] S503: Call the set of structural response interval ranges, map the response intervals to the corresponding time intervals on the controller pulse output frequency curve, and perform statistics and interval labeling on the pulse frequency change gradient within the mapped segment to generate the slide control output configuration set. The response intervals need to be mapped to the corresponding time periods in the controller's pulse output frequency curve. By matching the time range, the specific time point of each response interval is located on the controller's pulse output frequency change curve. During this process, the controller's pulse output frequency data needs to be extracted. The data reflects the frequency change of the pulses issued by the controller at different time points. For example, the frequency increases from 200Hz to 500Hz, which corresponds to the high-speed acceleration stage of the slide. After mapping the structural response intervals to frequency segments, the frequency change gradient within each mapped segment is statistically processed to determine whether the pulse frequency changes rapidly within the response interval. If the frequency rises by more than 100Hz within 5ms, it can be marked as a high-frequency surge segment, and the change trend can be marked within the interval. In practical engineering applications, if the structural response peak and the pulse frequency surge occur simultaneously in the interval between 150ms and 160ms, a slide control output configuration set is generated.

[0038] Please see Figure 7 A precision control system for linear motion modules based on AI vision, including: The image acquisition module illuminates the interference fringe image on the surface of the high-speed slide actuator, detects the change range of gray-level distribution area in the image frame, identifies the central axis of the fringe formed by the continuity of gray-level differences in the image, locates the coordinates of the gray-level peak point on the same interference fringe in consecutive frames, and generates a set of interference peak gray-level paths. The pixel trajectory module extracts the horizontal and vertical pixel positions of the interference center point in the image coordinate system based on the interference peak grayscale path set, obtains the pixel offset value of the center point in the horizontal axis and the frame sampling time interval between the real frame and the previous frame, classifies and calibrates the sliding direction, and generates the sliding table execution coordinate migration sequence. The coordinate migration module performs coordinate migration according to the slide, extracts the gray distribution curve of the high-speed slide structure edge contour in the image frame, obtains the gray gradient direction value of continuous pixels at the edge position, performs consistency judgment on the gray gradient direction of the inter-frame boundary region, records the coordinate range of edge feature points affected by illumination disturbance and structural deformation, and generates a feature position set after interference repair. The feature repair module calls the feature location set after interference repair, obtains the matching items between the location point sequence and the timestamp of the control pulse signal issued by the controller, calculates the time difference between the effective time of the command and the position response frame in the image within three consecutive frames, identifies the response lag change information in the control cycle, constructs the correspondence between the lag change coefficient and the control cycle, and generates a control pulse synchronization timetable. The timing configuration module extracts the control pulse signal sequence corresponding to the key control cycle based on the control pulse synchronization time table, calls the load peak change curve of the corresponding cycle in the high-speed slide actuator operation record, compares the load peak change curve with the deformation threshold range of the slide body material, and obtains the slide control output configuration set.

[0039] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A precise control method for a linear motion module based on AI vision, characterized in that, Includes the following steps: S1: Acquire the interference fringe image irradiated on the surface of the high-speed slide actuator, acquire continuous image frames in the image processing unit, extract the gray-scale change region of the interference fringes in the image frame, identify the central axis of the interference fringes in the image, and obtain the interference peak gray-scale path set; S2: Call the interference peak grayscale path set, extract the horizontal and vertical pixel positions of the interference center point in the image coordinate system of the continuous frame, and make a difference judgment based on the horizontal change distance of the center point coordinates between the real frame and the previous frame and the sampling time interval to obtain the sliding table execution coordinate migration sequence. S3: Based on the coordinate migration sequence of the slide, identify the gray gradient direction value corresponding to the edge of the high-speed slide structure in the image frame and the brightness distribution of the boundary region, calculate the consistency of the gray gradient direction of pixels in consecutive frames at the boundary, and obtain the feature location set after interference repair. S4: Using the feature position set after interference repair, align the repaired position difference with the timing of the control pulse signal. Based on the time difference between the command effective time and the real-time image response delay, statistically analyze the response lag change trend in adjacent control cycles to obtain the control pulse synchronization timetable.

2. The precise control method for linear motion modules based on AI vision according to claim 1, characterized in that, The interference peak grayscale path set includes grayscale distribution peak position, fringe path pixel trajectory, and inter-frame fringe morphology change; the slider execution coordinate migration sequence includes coordinate change trajectory, inter-frame displacement change value, and pixel motion trend; the interference-corrected feature position set includes edge vector offset value, feature point brightness correction parameter, and gradient direction correction index; and the control pulse synchronization time table includes control cycle identifier, time delay change curve, and synchronization alignment difference.

3. The precise control method for linear motion modules based on AI vision according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Acquire the interference fringe image irradiated on the surface of the high-speed slide actuator, call the continuously acquired image frame sequence in the image processing unit, perform linear traversal on the gray values ​​of the pixel columns in the image frame, perform curve fitting on the gray value difference sequence of adjacent pixels, extract the position index of the gray value change area, and generate a set of gray value change regions of interference fringes. S102: Based on the set of gray-scale abrupt change regions of the interference fringes, the coordinates of the center points between adjacent abrupt change regions in each frame image are aggregated by mean, and the position sequence of the interference center line in the two-dimensional matrix of the image frame is constructed. Then, linear fitting is performed on the longitudinal coordinate points of the center line in the image frame to generate the axial path sequence of the center of the interference fringes. S103: Call the central axis path sequence of the interference fringes, extract the gray-scale peak points along the central axis path coordinates of each frame in the continuous image frames, and aggregate the gray-scale peak points into multiple sets of homologous sequences according to the image frame acquisition time sequence index. Number each set of sequences by path to obtain the interference peak gray-scale path set.

4. The precise control method for linear motion modules based on AI vision according to claim 3, characterized in that, The specific steps of S2 are as follows: S201: Call the interference peak grayscale path set, extract the horizontal and vertical pixel indices of the grayscale peak points corresponding to consecutive frames on each path in the image coordinate system, arrange the peak point coordinates of consecutive frames in the same path in a time series, and perform structured processing on the arranged coordinate sequence to generate the inter-frame interference center point pixel position sequence. S202: Based on the inter-frame interference center point pixel position sequence, the ratio of the distance change value of the center point in the horizontal pixel coordinate direction in adjacent frames to the image sampling time interval is calculated to obtain the pixel displacement increment per unit time. According to the set horizontal displacement judgment threshold of two pixel units, the points in the continuous frames whose displacement increment exceeds the threshold are indexed and recorded to obtain the continuous horizontal pixel mutation point sequence. S203: Call the continuous horizontal pixel mutation point sequence, extract the image horizontal and vertical coordinate values ​​corresponding to multiple points, and map them to the execution space coordinate system of the high-speed slide table. Arrange the mapped coordinates in time series to generate the slide table execution coordinate migration sequence.

5. The precise control method for linear motion modules based on AI vision according to claim 4, characterized in that, The specific steps for S3 are as follows: S301: Call the slider to execute the coordinate migration sequence, extract the slider edge region image block in the image frame corresponding to the execution coordinate, perform convolution processing on the pixels in the slider edge region image block, obtain the gray-level gradient component values ​​in the horizontal and vertical directions, and traverse the brightness value of the edge region pixel by pixel to obtain the gray-level gradient direction value and brightness distribution set of the edge region. S302: Based on the gray gradient direction value and brightness distribution set of the edge region, calculate the angle difference of the gray gradient direction of the boundary region at the same position in consecutive image frames, and count the proportion of pixels whose absolute value of the angle difference is less than the preset angle consistency threshold. Perform density statistics on pixels whose gray change is greater than the brightness change detection threshold to obtain the edge interference feature point region index set. S303: Call the edge interference feature point region index set, perform vector translation operation on the region image coordinates, correct the coordinates based on the center position difference of adjacent non-interference regions, update the position of the coordinate correction points, identify the position distribution after continuous correction, and generate the interference repair feature position set.

6. The precise control method for a linear motion module based on AI vision according to claim 5, characterized in that, The specific steps of S4 are as follows: S401: Call the interference-corrected feature position set, extract the position difference of the same feature point in consecutive image frames within the control cycle, arrange the position difference sequence according to the image acquisition timestamp, obtain the timing mark of the control pulse signal issued by the controller, match the corrected position change with the control pulse time axis, and generate a position response and control command alignment sequence. S402: Based on the alignment sequence of the position response and control command, extract the command issuance time and the acquisition time of the corresponding position change response frame within the control cycle, calculate the difference between their timestamps as the image response lag time, and perform control cycle serialization processing on the lag time to generate a control cycle response lag change trend set. S403: Call the control cycle response lag change trend set, perform window smoothing on the time difference value in the lag change sequence, and establish an equally spaced pulse number index sequence in combination with the original control command time axis. Match and record the response delay of the control cycle with the pulse index to obtain the control pulse synchronization time table.

7. The precise control method for linear motion modules based on AI vision according to claim 6, characterized in that, The process of performing window smoothing on the time difference value in the lag change sequence is specifically as follows: the average value of the image response lag time in three adjacent control cycles within the control cycle response lag change trend set is calculated to obtain the corresponding smoothed time difference value sequence. The process of establishing an equally spaced pulse number index sequence by combining the original control command time axis is specifically as follows: taking the timestamp of the timing mark of the control pulse signal issued by the controller as the starting reference point, and numbering and indexing are performed according to a fixed time interval. The process of matching and recording the response delay of the control cycle with the pulse index specifically involves recording the smoothed time difference value at the time axis position corresponding to the numbered index.

8. The precise control method for a linear motion module based on AI vision according to claim 1, characterized in that, The method further includes step S5: S5: Call the timing configuration corresponding to the key control cycle of the high-speed slide table in the control pulse synchronization time table, match the load peak change trend in the real-time slide table actuator operation record, combine the deformation range and limit threshold of the high-speed slide table body material, determine the structural response range within the corresponding control cycle, map the structural response range to the controller output pulse frequency curve, and obtain the slide table control output configuration set. The slide control output configuration set includes a control frequency parameter group, a structural response mapping table, and a key cycle control scheme.

9. The precise control method for a linear motion module based on AI vision according to claim 8, characterized in that, The specific steps of S5 are as follows: S501: Call the control pulse synchronization time table, extract the pulse timing configuration interval marked as the key control cycle of the high-speed slide table, sort and calibrate the control pulse number and time point within the pulse timing configuration interval, extract the load peak change trend data within the corresponding time period from the operation record of the slide table actuator, perform interval correspondence after normalization of the time axis, and generate the load trend alignment sequence of the key control cycle. S502: Based on the load trend alignment sequence of the key control cycle, extract the load peak time period in the corresponding control cycle, compare the load peak time period with the deformation range limit threshold of the high-speed slide table body material, filter the continuous time periods where the load peak exceeds the deformation range limit threshold, and aggregate them into active structural response intervals according to time order to obtain a set of structural response interval ranges. S503: Call the set of structural response intervals, map the response intervals to the corresponding time intervals on the controller pulse output frequency curve, and perform statistics and interval labeling on the pulse frequency change gradient within the mapped segment to generate a slide control output configuration set.

10. A precision control system for linear motion modules based on AI vision, characterized in that, The system is used to implement the AI ​​vision-based precise control method for linear motion modules as described in any one of claims 1-9, and the system comprises: The image acquisition module illuminates the interference fringe image on the surface of the high-speed slide actuator, detects the change range of gray-level distribution area in the image frame, identifies the central axis of the fringe formed by the continuity of gray-level differences in the image, locates the coordinates of the gray-level peak point on the same interference fringe in consecutive frames, and generates a set of interference peak gray-level paths. The pixel trajectory module extracts the horizontal and vertical pixel positions of the interference center point in the image coordinate system based on the interference peak grayscale path set, obtains the pixel offset value of the center point in the horizontal axis and the frame sampling time interval between the real-time frame and the previous frame, classifies and calibrates the sliding direction, and generates a sliding table execution coordinate migration sequence. The coordinate migration module performs a coordinate migration sequence based on the slide, extracts the gray distribution curve of the high-speed slide structure edge contour in the image frame, obtains the gray gradient direction value of continuous pixels at the edge position, performs consistency judgment on the gray gradient direction of the inter-frame boundary region, records the coordinate range of edge feature points affected by illumination disturbance and structural deformation, and generates a feature position set after interference repair. The feature repair module calls the interference-repaired feature location set, obtains the matching items between the location point sequence and the timestamp of the control pulse signal issued by the controller, calculates the time difference between the instruction effective time and the position response frame in the image within three consecutive frames, identifies the response lag change information in the control cycle, constructs the correspondence between the lag change coefficient and the control cycle, and generates a control pulse synchronization timetable. Based on the control pulse synchronization schedule, the timing configuration module extracts the control pulse signal sequence corresponding to the key control cycle, calls the load peak change curve of the corresponding cycle in the high-speed slide actuator operation record, compares the load peak change curve with the deformation threshold range of the slide body material, and obtains the slide control output configuration set.

Images