Dynamic vision-based puncture deviation correction method, device and equipment and storage medium

Abstract

The application relates to a dynamic vision-based puncture deviation correction method, device, equipment and storage medium, which comprises collecting image data and spatial position data of a to-be-punctured object, then extracting target positioning information from the image data and constructing a corresponding three-dimensional deformation model; determining corresponding offset data according to the three-dimensional deformation model, fusing the target positioning information and the offset data to generate a deviation vector; based on the deviation vector, performing fuzzy control rule calculation and combining historical control data to generate a control output; by analyzing the change trend of the deviation vector at multiple time points, predicting the offset direction and amplitude of a preset time period to generate a feedforward compensation amount, and then generating a compensation control instruction; generating three-order trajectory information of a puncture path according to the compensation control instruction to complete dynamic deviation correction control according to the three-order trajectory information. The application has the effect of improving the precision of an automatic puncture system.

Description

Technical Field

[0001] This application relates to the technical field of medical automation control, and in particular to a puncture correction method, device, equipment and storage medium based on dynamic vision. Background Technology

[0002] Currently, automated puncture systems are widely used in medical infusion, drug extraction and experimental automation, and their core objective is to perform high-precision and safe puncture operations on flexible containers in dynamic environments.

[0003] Existing automated puncture technologies typically obtain the initial position of the target container through image recognition or tactile positioning, then combine this with a preset path to drive the puncture actuator to complete the needle insertion operation, or integrate laser sensors to monitor changes in target displacement and adjust the path using conventional PID control. However, these methods often lack effective correction mechanisms and adaptive control capabilities when dealing with target deformation, positioning drift, and vibration interference in dynamically changing environments. This can easily lead to needle deviation, puncture failure, or container damage, especially when dealing with nonlinear interference and sudden path changes, resulting in insufficient control stability.

[0004] The existing technical solutions mentioned above have the following drawbacks: existing automatic puncture systems typically use a single image positioning or sensor displacement feedback method for path control, which results in a lack of accurate correction capability when the target is rapidly deformed or its position shifts, thus leaving room for improvement. Summary of the Invention

[0005] To improve the accuracy of automated puncture systems, this application provides a puncture correction method, apparatus, device, and storage medium based on dynamic vision.

[0006] The above-mentioned objective of this application is achieved through the following technical solution:

[0007] A puncture correction method based on dynamic vision, the method comprising:

[0008] Image data and spatial location data of the object to be punctured are collected, and then subpixel-level target positioning information is extracted from the image data, and a corresponding three-dimensional deformation model is constructed based on the spatial location data.

[0009] The corresponding offset data is determined based on the three-dimensional deformation model, and the target positioning information is fused with the offset data to generate a deviation vector under a unified reference coordinate system.

[0010] Based on the deviation vector, fuzzy control rules are calculated, and parameters are adaptively adjusted in combination with historical control data to generate control output.

[0011] By analyzing the changing trend of the deviation vector at multiple time points, the offset direction and magnitude of the preset time period are predicted, and a feedforward compensation amount is generated. Then, the feedforward compensation amount is superimposed with the control output amount to generate a compensation control command.

[0012] The third-order trajectory information of the puncture path is generated according to the compensation control command, so as to complete the dynamic correction control based on the third-order trajectory information.

[0013] By adopting the above technical solutions, and by acquiring image data and spatial location data, extracting sub-pixel-level target positioning information and constructing a three-dimensional deformation model, high-precision modeling of the spatial state of the object to be punctured can be achieved, thereby improving the initial accuracy and model adaptability of the subsequent correction process. By fusing target positioning information and offset data to generate a deviation vector, a unified error expression form can be achieved under a unified coordinate system, thus facilitating standardized processing in subsequent control links. By performing fuzzy control calculations based on the deviation vector and combining historical data for adaptive parameter adjustment, the response adaptability of the control system to different error states can be improved, thereby achieving a more stable path correction effect. By analyzing the deviation change trend to generate feedforward compensation and superimposing it on the control output, error offset caused by control delay or sudden disturbances can be effectively compensated, thereby improving the real-time performance and dynamic accuracy of puncture path tracking. By generating third-order trajectory information to execute path control, the smooth continuity of the correction path in terms of position, velocity, and acceleration can be ensured, thereby improving the safety and mechanical stability of the puncture motion.

[0014] In one example, this application can be further configured as follows: extracting sub-pixel level target positioning information from image data and constructing a corresponding three-dimensional deformation model based on spatial location data specifically includes:

[0015] Image enhancement and edge refinement are performed on the image data to extract the contour features of the target region. Elliptical shape constraints are introduced during the edge fitting process to improve the stability and robustness of sub-pixel localization.

[0016] Multiple spatial location sensor sequences are collected simultaneously, and a time-layered three-dimensional deformation dataset is constructed based on the position changes at different time points. Then, by fusing the data structures of the spatial layer and the time layer, the three-dimensional deformation model is generated.

[0017] By adopting the above technical solutions, and by performing image enhancement and edge refinement to extract target contour features and introducing elliptical shape constraints, the stability and anti-interference ability of edge detection in complex structural regions can be enhanced, thereby ensuring the reliability of sub-pixel-level positioning results in dynamic scenes. By collecting multiple spatial position sensing sequences and constructing a time-layered three-dimensional deformation dataset, the gradual and abrupt changes of the target object in the time process can be reflected more realistically, thereby improving the simulation accuracy of the three-dimensional model for the deformation of liquid bag-like soft tissues.

[0018] In one example, this application can be further configured as follows: fusing the target positioning information with the offset data to generate a deviation vector in a unified reference coordinate system specifically includes:

[0019] Based on the two-dimensional coordinate data in the target positioning information, and combined with the imaging parameters and installation pose information of the image acquisition device, a projection mapping operation is performed to convert the two-dimensional coordinates into three-dimensional reference points.

[0020] Under the unified reference coordinate system, the three-dimensional reference points are registered with the offset data, and the registered three-dimensional modeling error is dynamically adjusted through a preset weighted compensation strategy to obtain the deviation vector.

[0021] By adopting the above technical solution, and by performing a two-dimensional to three-dimensional projection mapping operation and fusing imaging parameters and pose information to generate three-dimensional reference points, the image recognition results can be reliably mapped to actual spatial coordinates, improving the alignment between data and thus laying an accurate foundation for subsequent coordinate registration. By performing coordinate registration under a unified reference coordinate system and dynamically adjusting the three-dimensional modeling error, the alignment offset caused by modeling deviation or pose error can be effectively reduced, thereby improving the accuracy of the final generated deviation vector and providing a reliable basis for control decisions.

[0022] In one example, this application can be further configured as follows: the step of dynamically adjusting the registered 3D modeling error using a preset weighted compensation strategy to obtain the deviation vector specifically includes:

[0023] Based on the historical positioning error statistics, the confidence interval of the 3D modeling error is calculated, and the weight of the unified reference coordinate system is adjusted according to the confidence distribution of the confidence interval.

[0024] By adopting the above technical solution, the confidence interval of the 3D modeling error is calculated based on the statistical results of historical positioning errors, and the coordinate weights are dynamically adjusted according to the confidence distribution. This enables quantitative feedback and hierarchical processing of the data credibility, thereby prioritizing the use of highly reliable feature information during the registration process and effectively enhancing the consistency and anti-interference capability of the fused data.

[0025] In one example, this application can be further configured as follows: the step of performing fuzzy control rule calculation based on the deviation vector, and adaptively adjusting parameters in combination with historical control data to generate control output, specifically includes:

[0026] The magnitude and rate of change of the deviation vector are used as input variables, corresponding to the error quantity and error rate of change input dimensions in the fuzzy controller, respectively, and the preset fuzzy rule base is called to output the control level.

[0027] Based on the control residuals and actual response results over multiple consecutive historical control cycles, the proportional factor, integral factor, and derivative factor of the fuzzy controller are dynamically adjusted, and different parameter tuning intervals are selected according to the residual convergence rate, thereby calculating the control output.

[0028] By adopting the above technical solution, and using the magnitude and rate of change of the deviation vector as fuzzy control inputs, combined with preset rules to output control levels, a flexible response to nonlinear error fluctuations can be achieved, thereby adapting to various sudden and non-stationary disturbance scenarios. By dynamically adjusting the proportional, integral, and derivative factors based on historical control residuals, and switching the parameter tuning strategy according to the residual convergence rate, the control system can maintain optimal control performance in different disturbance environments, thereby effectively balancing response speed and stability.

[0029] In one example, this application can be further configured such that: the dynamic adjustment of the scaling factor, integral factor, and derivative factor of the fuzzy controller specifically includes:

[0030] Construct a sliding sequence that records the residuals of multiple consecutive control cycles with a fixed window length, and calculate the mean residual and rate of change index within the current window;

[0031] The residual change rate is compared with multiple preset convergence rate thresholds, and the corresponding parameter tuning strategy is selected according to the rate range. When the residual converges slowly, the scaling factor is increased and the integration time is shortened. When the residual oscillates, the scaling factor is decreased and the integration time is extended.

[0032] By adopting the above technical solution, constructing a residual sliding sequence and extracting the residual change rate, and comparing it with a preset convergence rate threshold to select a parameter tuning strategy, the control performance change trend can be reflected in real time. Based on this, the strength of the control factor and the time constant can be dynamically adjusted, thereby improving the control algorithm's adaptability to external disturbances and model uncertainties, and ensuring the efficient and stable operation of the control process.

[0033] In one example, this application can be further configured as follows: by analyzing the changing trend of the deviation vector at multiple time points, predicting the offset direction and magnitude of a preset time period, generating a feedforward compensation amount, and then superimposing the feedforward compensation amount with the control output amount to generate a compensation control command, specifically including:

[0034] A time series window containing deviation vectors from multiple consecutive historical moments is constructed based on a preset sampling period, and the first derivative of the time series in each axis is calculated to obtain the current offset trend.

[0035] Based on the current offset trend and the preset control response delay parameters, the potential offset increment within the target response period is predicted, and then a feedforward compensation amount containing a three-axis compensation component is generated.

[0036] The feedforward compensation amount and the control output amount are weighted and superimposed in vector space to generate the compensation control command.

[0037] By adopting the above technical solutions, and by constructing a time series window containing multiple historical deviation vectors and calculating the first derivative, the instantaneous development trend of system error can be obtained, thereby enabling the prediction of future short-term offsets. By combining control response delay parameters to predict the offset increment within the target response period and generating triaxial compensation components, dynamic errors caused by control link lag can be actively compensated. By performing weighted superposition of control output and feedforward compensation in vector space, feedback regulation and predictive look-ahead control can be integrated to improve the integrity and dynamic accuracy of command output, thereby effectively improving path tracking capabilities in complex deformation scenarios.

[0038] The second objective of this invention is achieved through the following technical solution:

[0039] A puncture correction device based on dynamic vision, the device comprising:

[0040] The image localization and modeling module is used to collect image data and spatial location data of the object to be punctured, and then extract subpixel-level target localization information from the image data, and construct a corresponding three-dimensional deformation model based on the spatial location data.

[0041] The deviation fusion module is used to determine the corresponding offset data based on the three-dimensional deformation model, and fuse the target positioning information with the offset data to generate a deviation vector in a unified reference coordinate system.

[0042] The fuzzy control module is used to perform fuzzy control rule calculation based on the deviation vector, and to adaptively adjust the parameters by combining historical control data to generate control output.

[0043] The prediction and compensation module is used to predict the offset direction and magnitude of the deviation vector at multiple time points by analyzing the changing trend of the deviation vector at multiple time points, generate a feedforward compensation amount, and then superimpose the feedforward compensation amount with the control output amount to generate a compensation control command.

[0044] The trajectory planning module is used to generate third-order trajectory information of the puncture path according to the compensation control command, so as to complete dynamic correction control based on the third-order trajectory information.

[0045] By adopting the above technical solutions, and by acquiring image data and spatial location data, extracting sub-pixel-level target positioning information and constructing a three-dimensional deformation model, high-precision modeling of the spatial state of the object to be punctured can be achieved, thereby improving the initial accuracy and model adaptability of the subsequent correction process. By fusing target positioning information and offset data to generate a deviation vector, a unified error expression form can be achieved under a unified coordinate system, thus facilitating standardized processing in subsequent control links. By performing fuzzy control calculations based on the deviation vector and combining historical data for adaptive parameter adjustment, the response adaptability of the control system to different error states can be improved, thereby achieving a more stable path correction effect. By analyzing the deviation change trend to generate feedforward compensation and superimposing it on the control output, error offset caused by control delay or sudden disturbances can be effectively compensated, thereby improving the real-time performance and dynamic accuracy of puncture path tracking. By generating third-order trajectory information to execute path control, the smooth continuity of the correction path in terms of position, velocity, and acceleration can be ensured, thereby improving the safety and mechanical stability of the puncture motion.

[0046] The above-mentioned objective three of this application is achieved through the following technical solution:

[0047] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described dynamic vision-based puncture correction method.

[0048] The fourth objective of this application is achieved through the following technical solution:

[0049] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described dynamic vision-based puncture correction method.

[0050] In summary, this application includes the following beneficial technical effects:

[0051] 1. By acquiring image and spatial location data, extracting sub-pixel-level target positioning information, and constructing a 3D deformation model, high-precision modeling of the spatial state of the object to be punctured can be achieved, thereby improving the initial accuracy and model adaptability of the subsequent correction process. By fusing target positioning information and offset data to generate a deviation vector, a unified error expression form can be achieved under a unified coordinate system, facilitating standardized processing in subsequent control stages. By performing fuzzy control calculations based on the deviation vector and combining historical data for adaptive parameter adjustment, the response adaptability of the control system to different error states can be improved, resulting in a more stable path correction effect. By analyzing the deviation change trend to generate feedforward compensation and superimposing it on the control output, error offset caused by control delay or sudden disturbances can be effectively compensated, thereby improving the real-time performance and dynamic accuracy of puncture path tracking. By generating third-order trajectory information to execute path control, the smooth continuity of the correction path in terms of position, velocity, and acceleration can be ensured, thereby improving the safety and mechanical stability of the puncture motion.

[0052] 2. By performing image enhancement and edge refinement to extract target contour features and introducing elliptical shape constraints, the stability and anti-interference ability of edge detection in complex structural regions can be enhanced, thereby ensuring the reliability of sub-pixel-level localization results in dynamic scenes; by collecting multiple spatial position sensing sequences and constructing a time-layered three-dimensional deformation dataset, the gradual and abrupt changes of the target object in the time process can be reflected more realistically, thereby improving the simulation accuracy of the three-dimensional model for the deformation of liquid bag-like soft tissues;

[0053] 3. By performing a 2D-to-3D projection mapping operation and fusing imaging parameters and pose information to generate 3D reference points, the image recognition results can be reliably mapped to actual spatial coordinates, improving the alignment between data and thus laying an accurate foundation for subsequent coordinate registration. By performing coordinate registration under a unified reference coordinate system and dynamically adjusting 3D modeling errors, the alignment offset caused by modeling deviations or pose errors can be effectively reduced, thereby improving the accuracy of the final generated deviation vector and providing a reliable basis for control decisions. Attached Figure Description

[0054] Figure 1 This is a flowchart of a puncture correction method based on dynamic vision in one embodiment of this application;

[0055] Figure 2 This is a principle block diagram of a puncture correction device based on dynamic vision in one embodiment of this application;

[0056] Figure 3 This is a schematic diagram of a device according to one embodiment of this application. Detailed Implementation

[0057] The present application will be further described in detail below with reference to the accompanying drawings.

[0058] In one embodiment, such as Figure 1 As shown, this application discloses a puncture correction method based on dynamic vision, which specifically includes the following steps:

[0059] S10: Collect image data and spatial location data of the object to be punctured, then extract sub-pixel level target positioning information from the image data, and construct the corresponding three-dimensional deformation model based on the spatial location data.

[0060] Specifically, when acquiring image data, a fixed frame rate image acquisition command is used to control the imaging process. Based on the set sampling area, the target area is sampled to obtain an image frame sequence. Then, the edge extraction and morphology fitting process is performed on the acquired image frames. During the edge recognition process, sub-pixel level coordinate fitting logic is called to obtain the precise position of the contour points in the image. At the same time, the position data stream input from the external sensing interface is received synchronously. The position samples at different times are sorted by time label and a multi-frame fusion structure is constructed for subsequent modeling. Finally, the target localization results extracted from the image and the position information at multiple time points are organized into a unified data stream as the initial input for spatial modeling processing.

[0061] S20: Determine the corresponding offset data based on the three-dimensional deformation model, and fuse the target positioning information with the offset data to generate a deviation vector under a unified reference coordinate system.

[0062] Specifically, the spatial modeling module is invoked to input and map the previously processed position information and image positioning results. Based on the distance difference between the target initial reference point and the current modeling result in the three axes of space, spatial offset data of the current frame is generated. The image recognition coordinates and offset data are converted to a consistent reference system through the coordinate transformation interface. Data alignment is performed and the position difference vector is calculated as a deviation result vector representing the current target offset state, which is used to drive the deviation feedback entry in the subsequent control process.

[0063] S30: Based on the deviation vector, perform fuzzy control rule calculation, and combine historical control data to adaptively adjust parameters and generate control output.

[0064] Specifically, firstly, the magnitude of the current deviation vector and the change during adjacent control cycles are extracted as control input variables and passed to the fuzzy rule base to find the control magnitude label and obtain the response instruction for the current deviation level. While obtaining the fuzzy output, the historical control response records of multiple rounds are read from the control cache to calculate the changing trend of the current control residual. Based on this trend, the configuration values ​​of the proportional, integral, and derivative parameters are adjusted in the current control logic to form a set of adaptively corrected control parameters for the current operating condition. Finally, the fuzzy control results and the corrected parameters are input together into the control quantity generation function to output a more accurate control execution instruction for path control.

[0065] S40: By analyzing the changing trend of the deviation vector at multiple time points, the offset direction and magnitude of the preset time period are predicted, and a feedforward compensation amount is generated. Then, the feedforward compensation amount is superimposed with the control output amount to generate a compensation control command.

[0066] Specifically, the trend analysis module is invoked to construct a sequence of deviation vectors recorded with a fixed sampling period. The trend of the deviation vector is approximated by solving the slope of its change between the most recent time points. Then, the expected offset state at the end of the target control cycle is calculated by combining the current control response delay parameter, thereby generating a predictive compensation vector for advance correction. In the control synthesis stage, the predictive compensation vector is weighted and merged one-to-one with the feedback control output vector to generate a composite control command that includes both feedback response and predictive compensation, which is used to achieve rapid following and offset correction in dynamic environments.

[0067] S50: Generates third-order trajectory information of the puncture path according to the compensation control command, so as to complete dynamic correction control based on the third-order trajectory information.

[0068] Specifically, the target point information and current execution state in the composite control command are analyzed, the target displacement and control time window boundary conditions are calculated, the initial state of the third-order interpolation equation is constructed in the trajectory planning process, and the trajectory function parameters are solved using velocity and acceleration constraints as control conditions. Then, the trajectory function is sampled at a set time step to generate a continuous and differentiable spatial path point sequence, and this sequence is used as the input of the path control module. During the puncture control process, the execution target is dynamically updated to achieve smooth connection and continuous trajectory puncture action control.

[0069] In one embodiment, step S10, which involves extracting sub-pixel-level target positioning information from image data and constructing a corresponding three-dimensional deformation model based on spatial location data, specifically includes:

[0070] S11: Image enhancement and edge refinement are performed on the image data to extract the contour features of the target region. Elliptical shape constraints are introduced during the edge fitting process to improve the stability and robustness of sub-pixel localization.

[0071] Specifically, in the image processing flow, grayscale normalization and high-pass filtering are first performed on the original image to enhance the contrast between the target area and the background. Then, edge detection operators such as the Canny algorithm or Sobel gradient operation are called to extract the initial boundary of the contour. Based on the extracted boundary point set, edge connection and contour closure strategies are used to form a closed shape. During the contour fitting process, elliptical geometric constraints are applied to make the fitted shape conform to the shape characteristics of the target area in the major and minor axis directions. In the fitting result, the center point and the main axis direction are selected as sub-pixel level positioning outputs. The center position of the ellipse can be obtained by solving the least squares fitting equation.

[0072] S12: Simultaneously acquire multiple spatial position sensor sequences, and construct a time-layered three-dimensional deformation dataset based on position changes at different time points. Then, by fusing the data structures of the spatial and temporal layers, a three-dimensional deformation model is generated.

[0073] Specifically, while performing image acquisition, a time synchronization mechanism is invoked to receive output data from external position sensors. At the receiving end, the data from each sensor channel is organized and aligned according to the timestamp to construct a spatial position frame sequence covering the X, Y, and Z directions. Each frame of acquired data is divided into its own time layer index according to its sampling time. Then, multiple points in the same time layer are combined into a spatial layer grid to form a spatial + temporal dual index structure. The spatial grid sequence of each time layer is input into the three-dimensional interpolation reconstruction function to generate a spatially continuous voxel model or point cloud surface model, and the local deformation trend at each time node is retained to support subsequent modeling adjustment and predictive analysis.

[0074] In one embodiment, step S20 involves fusing the target positioning information with the offset data to generate a deviation vector in a unified reference coordinate system, specifically including:

[0075] S21: Based on the two-dimensional coordinate data in the target positioning information, and combined with the imaging parameters and installation pose information of the image acquisition device, perform a projection mapping operation to convert the two-dimensional coordinates into three-dimensional reference points.

[0076] Specifically, the two-dimensional coordinate points extracted from the image are input into the camera calibration model. This model contains the internal parameter matrix and external pose description of the image acquisition device. The intrinsic parameters are used to complete the inverse transformation from the pixel plane to the normalized camera coordinates, and the extrinsic parameters are used to represent the rotation matrix and translation vector between the camera and the world coordinate system. Then, a back projection operation is performed. Under the condition that the target depth is estimated or the depth is inferred through the three-dimensional deformation model, the corresponding three-dimensional spatial reference point is calculated. This point is output in the form of world coordinates for subsequent alignment with the modeling offset.

[0077] S22: Under a unified reference coordinate system, the coordinates of the three-dimensional reference points and the offset data are registered, and the error of the registered three-dimensional modeling is dynamically adjusted by a preset weighted compensation strategy to obtain the deviation vector.

[0078] Specifically, firstly, in a unified reference coordinate system, the current offset vector output by the 3D deformation model is aligned with the corresponding position of the aforementioned reference point. The spatial vector difference between the theoretical reference position and the current offset position is calculated to form a preliminary original deviation result. Based on this, an error compensation strategy is introduced, and the weight evaluation function is called to dynamically weight and correct the deformation disturbance error that may exist in the modeling. The corresponding confidence adjustment coefficient is multiplied in each direction to form a set of corrected three-axis offsets. Finally, this set of corrected vectors is provided as the deviation vector for subsequent control logic.

[0079] In one embodiment, step S22, namely, dynamically adjusting the registered 3D modeling error using a preset weighted compensation strategy to obtain a deviation vector, specifically includes:

[0080] S221: Based on the historical positioning error statistics, calculate the confidence interval of the 3D modeling error, and adjust the weight of the unified reference coordinate system according to the confidence distribution of the confidence interval.

[0081] Specifically, the historical data module is called to read error samples of similar objects from several puncture tasks. The mean and standard deviation of the error are calculated for each spatial axis, and an error probability distribution function is constructed. The upper and lower bound intervals are calculated according to the preset confidence level. The confidence level is determined at the error landing point between the current registration point and the model output point. The corresponding weight value is selected by looking up the table according to the level. The displacement in the three axes of the current reference coordinate system is multiplied by the weight factor to complete the dynamic confidence adjustment based on statistical evaluation, thereby enhancing the fault tolerance capability of the low confidence model output.

[0082] In one embodiment, step S30, which involves calculating fuzzy control rules based on the deviation vector and adaptively adjusting parameters using historical control data to generate a control output, specifically includes:

[0083] S31: The magnitude and rate of change of the deviation vector are used as input variables, corresponding to the error quantity and error rate of change input dimensions in the fuzzy controller, respectively, and the preset fuzzy rule base is called to output the control level.

[0084] Specifically, the deviation vector of the current cycle is read and its magnitude is calculated as the error amplitude. At the same time, the difference between the deviation value of the previous cycle and the current value is read and divided by the cycle length as the error change rate. Fuzzification operation is performed on the two input variables respectively, and they are fuzzily mapped to linguistic variable levels, such as "large", "medium", "small" or "fast increase" and "slow decrease". Then, these two fuzzy inputs are sent to the fuzzy rule matching module, which calls the preset fuzzy control rule table, such as the IF-THEN type rule set, for condition judgment. Finally, the control level label is output, such as "emphasis mode", "fine adjustment" or "hold", for subsequent control quantity generation.

[0085] S32: Based on the control residuals and actual response results over multiple consecutive historical control cycles, the proportional factor, integral factor, and derivative factor of the fuzzy controller are dynamically adjusted, and different parameter tuning intervals are selected according to the residual convergence rate, thereby calculating the control output.

[0086] Specifically, a fixed-length historical control window is established. After each control execution, the difference between the corresponding control output and the system response is recorded to form a control residual sequence. In the current period, this sequence is read to calculate its mean and rate of change to represent the convergence trend. Multiple parameter tuning levels are set in the fuzzy controller structure, such as high response, weak response, and oscillation suppression. The optimal level is selected based on the current residual trend, and the weight coefficients of the three factors, proportional (P), integral (I), and derivative (D), are adjusted. Then, the control level output by the fuzzy rules is used as the target quantity, and the control function is called in combination with the parameter tuning results to calculate the actual control output value, which is used to guide path correction.

[0087] In one embodiment, step S32, namely dynamically adjusting the scaling factor, integral factor, and derivative factor of the fuzzy controller, specifically includes:

[0088] S321: Construct a sliding sequence that records the residuals of multiple consecutive control cycles with a fixed window length, and calculate the mean residual and rate of change index within the current window.

[0089] Specifically, in the control loop, a window length parameter such as N frames is set, and the control residual value of each frame is continuously recorded and formed into a sliding array. When a new cycle arrives, the sliding array is advanced by one frame, that is, the oldest data is removed and the latest residual value is added. Then, the average value of the array is calculated to obtain the current residual level. At the same time, the difference between the first and last residuals is divided by the time window length to calculate the rate of change index. These two indices are used together to evaluate whether the current system control error is within a controllable range and whether there is an aggravated fluctuation phenomenon.

[0090] S322: Compare the residual change rate with multiple preset convergence rate thresholds, and select the corresponding parameter tuning strategy according to the rate range. When the residual converges slowly, increase the scaling factor and shorten the integration time. When the residual oscillates, decrease the scaling factor and extend the integration time.

[0091] Specifically, the calculated rate of change is input to the parameter adjustment judgment module and compared with the set low-speed, medium-speed, and high-speed convergence thresholds. If the residual convergence rate is slow, the current control is considered to be lagging, the P parameter is increased to improve the response strength, and the I parameter integral period is shortened to accelerate the error accumulation correction. If the residual is found to fluctuate continuously and change drastically, it is considered to have the risk of oscillation, the P value is reduced to suppress the control gain, and the I integral period is extended to slow down the error response rhythm, thereby realizing the dynamic optimization selection of control parameters and improving system stability.

[0092] In one embodiment, in step S40, by analyzing the changing trend of the deviation vector at multiple time points, the offset direction and magnitude of the preset time period are predicted, a feedforward compensation amount is generated, and then the feedforward compensation amount is superimposed with the control output amount to generate a compensation control command, specifically including:

[0093] S41: Construct a time series window containing deviation vectors of multiple consecutive historical moments based on a preset sampling period, and calculate the first derivative of the time series on each axis to obtain the current offset trend.

[0094] Specifically, a fixed time step is defined as the sampling period. In each period, the current three-axis deviation vector value is recorded and sequentially written into the deviation sequence window. The window length is kept constant. When the window data is full, the rolling update begins. By performing a first-order difference operation on the deviation value in each direction, the rate of change of the offset of each axis in time is calculated, thereby obtaining the offset trend vector. This trend vector is used to describe the inertial offset direction and intensity characteristics of the target object in the next time period.

[0095] S42: Based on the current offset trend and the preset control response delay parameters, predict the potential offset increment within the target response period, and then generate a feedforward compensation amount containing three-axis compensation components.

[0096] Specifically, the trend vector and response time product model is invoked. Based on the multiplication of the current offset trend vector of each axis with the set control delay period, the displacement increment that the target may generate during the control execution lag period is calculated. Predicted values ​​are generated for the X, Y, and Z axes respectively. The prediction results of these three directions are combined to form a three-dimensional feedforward compensation vector, which serves as the predicted output of the upcoming offset state and provides a predictive reference for the generation of compensation control commands.

[0097] S43: Perform component weighted superposition of the feedforward compensation quantity and the control output quantity in the vector space to generate a compensation control command.

[0098] Specifically, the control fusion module receives the feedforward vector and feedback control vector as input, assigns corresponding weighting coefficients according to the reliability index of each vector component, and performs weighted superposition operations in each spatial direction. For example, in the Z-axis direction, it combines 70% feedforward and 30% feedback, and in the Y-axis direction, it combines 50% to 50% to complete the synthesis of three-dimensional spatial control commands. The composite vector is used as the final path control input signal for the next cycle trajectory generation and execution adjustment.

[0099] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0100] In one embodiment, a dynamic vision-based puncture correction device is provided, which corresponds one-to-one with the dynamic vision-based puncture correction method described in the above embodiments. For example... Figure 2 As shown, the puncture correction device based on dynamic vision includes an image localization and modeling module, a deviation fusion module, a fuzzy control module, a prediction compensation module, and a trajectory planning module. Detailed descriptions of each functional module are as follows:

[0101] The image localization and modeling module is used to collect image data and spatial location data of the object to be punctured, and then extract subpixel-level target localization information from the image data and construct the corresponding three-dimensional deformation model based on the spatial location data.

[0102] The deviation fusion module is used to determine the corresponding offset data based on the 3D deformation model, and fuse the target positioning information with the offset data to generate a deviation vector in a unified reference coordinate system.

[0103] The fuzzy control module is used to perform fuzzy control rule calculations based on the deviation vector, and to adaptively adjust parameters by combining historical control data to generate control output quantities.

[0104] The prediction compensation module is used to predict the offset direction and magnitude of the preset time period by analyzing the changing trend of the deviation vector at multiple time points, generate the feedforward compensation amount, and then superimpose the feedforward compensation amount with the control output amount to generate the compensation control command.

[0105] The trajectory planning module is used to generate third-order trajectory information of the puncture path based on the compensation control command, so as to complete dynamic correction control based on the third-order trajectory information.

[0106] Optionally, the image localization and modeling module specifically includes:

[0107] The image processing submodule is used to enhance and refine the image data, thereby extracting the contour features of the target region, and introducing elliptical shape constraints during the edge fitting process to improve the stability and robustness of subpixel localization.

[0108] The deformation modeling submodule is used to simultaneously acquire multiple spatial location sensor sequences and construct a time-layered 3D deformation dataset based on the position changes at different time points. Then, by fusing the data structures of the spatial and temporal layers, a 3D deformation model is generated.

[0109] Optionally, the deviation fusion module specifically includes:

[0110] The coordinate transformation submodule is used to perform projection mapping operations based on the two-dimensional coordinate data in the target positioning information, combined with the imaging parameters and installation pose information of the image acquisition device, to convert the two-dimensional coordinates into three-dimensional reference points.

[0111] The coordinate fusion submodule is used to register the 3D reference points with the offset data in a unified reference coordinate system, and dynamically adjust the 3D modeling error after registration through a preset weighted compensation strategy to obtain the deviation vector.

[0112] Optionally, the coordinate fusion submodule specifically includes:

[0113] The confidence-weighted unit is used to calculate the confidence interval of the 3D modeling error based on the historical positioning error statistics, and to adjust the weight of the unified reference coordinate system according to the confidence distribution of the confidence interval.

[0114] Optionally, the fuzzy control module specifically includes:

[0115] The fuzzy inference submodule is used to take the magnitude and rate of change of the deviation vector as input variables, which correspond to the error quantity and error rate of change input dimensions in the fuzzy controller, respectively, and call the preset fuzzy rule library to output the control level.

[0116] The adaptive parameter tuning submodule is used to dynamically adjust the proportional factor, integral factor, and derivative factor of the fuzzy controller based on the control residuals and actual response results over multiple consecutive historical control cycles, and select different parameter tuning intervals according to the residual convergence rate, thereby calculating the control output.

[0117] Optionally, the adaptive parameter tuning submodule includes:

[0118] The residual analysis unit is used to construct a sliding sequence that records the residuals of multiple consecutive control cycles with a fixed window length, and to calculate the mean residual and rate of change index within the current window.

[0119] The strategy selection unit is used to compare the residual change rate with multiple preset convergence rate thresholds and select the corresponding parameter tuning strategy according to the rate range. When the residual converges slowly, the scaling factor is increased and the integration time is shortened. When the residual oscillates, the scaling factor is decreased and the integration time is extended.

[0120] Optionally, the prediction compensation module specifically includes:

[0121] The trend extraction submodule is used to construct a time series window containing deviation vectors of multiple consecutive historical moments based on a preset sampling period, and to calculate the first derivative of the time series on each axis to obtain the current offset trend.

[0122] The feedforward calculation submodule is used to predict the potential offset increment within the target response period based on the current offset trend and the preset control response delay parameters, and then generate the feedforward compensation amount containing the three-axis compensation component.

[0123] The control fusion submodule is used to perform component weighted superposition of the feedforward compensation quantity and the control output quantity in the vector space to generate compensation control commands.

[0124] Specific limitations regarding the dynamic vision-based puncture correction device can be found in the limitations of the dynamic vision-based puncture correction method described above, and will not be repeated here. Each module in the aforementioned dynamic vision-based puncture correction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0125] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a dynamic vision-based puncture correction method.

[0126] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0127] Image data and spatial location data of the object to be punctured are collected, and then subpixel-level target positioning information is extracted from the image data, and a corresponding three-dimensional deformation model is constructed based on the spatial location data.

[0128] The corresponding offset data is determined based on the three-dimensional deformation model, and the target positioning information is fused with the offset data to generate a deviation vector under a unified reference coordinate system.

[0129] Based on the deviation vector, fuzzy control rules are calculated, and parameters are adaptively adjusted by combining historical control data to generate control output.

[0130] By analyzing the changing trend of the deviation vector at multiple time points, the offset direction and magnitude of the preset time period are predicted, and a feedforward compensation amount is generated. Then, the feedforward compensation amount is superimposed with the control output amount to generate a compensation control command.

[0131] The third-order trajectory information of the puncture path is generated according to the compensation control command, so as to complete the dynamic correction control based on the third-order trajectory information.

[0132] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0133] Image data and spatial location data of the object to be punctured are collected, and then subpixel-level target positioning information is extracted from the image data, and a corresponding three-dimensional deformation model is constructed based on the spatial location data.

[0134] The corresponding offset data is determined based on the three-dimensional deformation model, and the target positioning information is fused with the offset data to generate a deviation vector under a unified reference coordinate system.

[0135] Based on the deviation vector, fuzzy control rules are calculated, and parameters are adaptively adjusted by combining historical control data to generate control output.

[0136] By analyzing the changing trend of the deviation vector at multiple time points, the offset direction and magnitude of the preset time period are predicted, and a feedforward compensation amount is generated. Then, the feedforward compensation amount is superimposed with the control output amount to generate a compensation control command.

[0137] The third-order trajectory information of the puncture path is generated according to the compensation control command, so as to complete the dynamic correction control based on the third-order trajectory information.

[0138] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0139] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0140] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A puncture correction method based on dynamic vision, characterized in that, The method includes: Image data and spatial location data of the object to be punctured are collected, and then subpixel-level target positioning information is extracted from the image data, and a corresponding three-dimensional deformation model is constructed based on the spatial location data. The corresponding offset data is determined based on the three-dimensional deformation model, and the target positioning information is fused with the offset data to generate a deviation vector under a unified reference coordinate system. Based on the deviation vector, fuzzy control rules are calculated, and parameters are adaptively adjusted in combination with historical control data to generate control output. By analyzing the changing trend of the deviation vector at multiple time points, the offset direction and magnitude of the preset time period are predicted, and a feedforward compensation amount is generated. Then, the feedforward compensation amount is superimposed with the control output amount to generate a compensation control command. The third-order trajectory information of the puncture path is generated according to the compensation control command, so as to complete the dynamic correction control based on the third-order trajectory information; The step of fusing the target positioning information with the offset data to generate a deviation vector in a unified reference coordinate system specifically includes: Based on the two-dimensional coordinate data in the target positioning information, and combined with the imaging parameters and installation pose information of the image acquisition device, a projection mapping operation is performed to convert the two-dimensional coordinates into three-dimensional reference points. Under the unified reference coordinate system, the three-dimensional reference points are registered with the offset data, and the registered three-dimensional modeling error is dynamically adjusted through a preset weighted compensation strategy to obtain the deviation vector. The step of performing fuzzy control rule calculation based on the deviation vector, and adaptively adjusting parameters in conjunction with historical control data to generate control output quantities specifically includes: The magnitude and rate of change of the deviation vector are used as input variables, corresponding to the error quantity and error rate of change input dimensions in the fuzzy controller, respectively, and the preset fuzzy rule base is called to output the control level. Based on the control residuals and actual response results over multiple consecutive historical control cycles, the proportional factor, integral factor, and derivative factor of the fuzzy controller are dynamically adjusted, and different parameter tuning intervals are selected according to the residual convergence rate, thereby calculating the control output.

2. The puncture correction method based on dynamic vision according to claim 1, characterized in that, The extraction of sub-pixel-level target positioning information from image data and the construction of a corresponding three-dimensional deformation model based on spatial location data specifically include: Image enhancement and edge refinement are performed on the image data to extract the contour features of the target region. Elliptical shape constraints are introduced during the edge fitting process to improve the stability and robustness of sub-pixel localization. Multiple spatial location sensor sequences are collected simultaneously, and a time-layered three-dimensional deformation dataset is constructed based on the position changes at different time points. Then, by fusing the data structures of the spatial layer and the time layer, the three-dimensional deformation model is generated.

3. The puncture correction method based on dynamic vision according to claim 1, characterized in that, The step of dynamically adjusting the registered 3D modeling error using a preset weighted compensation strategy to obtain the deviation vector specifically includes: Based on the historical positioning error statistics, the confidence interval of the 3D modeling error is calculated, and the weight of the unified reference coordinate system is adjusted according to the confidence distribution of the confidence interval.

4. The puncture correction method based on dynamic vision according to claim 1, characterized in that, The dynamic adjustment of the scaling factor, integral factor, and derivative factor of the fuzzy controller specifically includes: Construct a sliding sequence that records the residuals of multiple consecutive control cycles with a fixed window length, and calculate the mean residual and rate of change index within the current window; The residual change rate is compared with multiple preset convergence rate thresholds, and the corresponding parameter tuning strategy is selected according to the rate range. When the residual converges slowly, the scaling factor is increased and the integration time is shortened. When the residual oscillates, the scaling factor is decreased and the integration time is extended.

5. The puncture correction method based on dynamic vision according to claim 1, characterized in that, The process involves analyzing the changing trend of the deviation vector at multiple time points to predict the offset direction and magnitude within a preset time period, generating a feedforward compensation amount, and then superimposing the feedforward compensation amount with the control output to generate a compensation control command. Specifically, this includes: A time series window containing deviation vectors from multiple consecutive historical moments is constructed based on a preset sampling period, and the first derivative of the time series in each axis is calculated to obtain the current offset trend. Based on the current offset trend and the preset control response delay parameters, the potential offset increment within the target response period is predicted, and then a feedforward compensation amount containing three-axis compensation components is generated. The feedforward compensation amount and the control output amount are weighted and superimposed in vector space to generate the compensation control command.

6. A puncture correction device based on dynamic vision, characterized in that, The device includes: The image localization and modeling module is used to collect image data and spatial location data of the object to be punctured, and then extract subpixel-level target localization information from the image data, and construct a corresponding three-dimensional deformation model based on the spatial location data. The deviation fusion module is used to determine the corresponding offset data based on the three-dimensional deformation model, and fuse the target positioning information with the offset data to generate a deviation vector in a unified reference coordinate system. The fuzzy control module is used to perform fuzzy control rule calculation based on the deviation vector, and to adaptively adjust the parameters by combining historical control data to generate control output. The prediction compensation module is used to predict the offset direction and magnitude of the deviation vector at multiple time points by analyzing the changing trend of the deviation vector at multiple time points, generate a feedforward compensation amount, and then superimpose the feedforward compensation amount with the control output amount to generate a compensation control command. The trajectory planning module is used to generate third-order trajectory information of the puncture path according to the compensation control command, so as to complete dynamic correction control based on the third-order trajectory information. The step of fusing the target positioning information with the offset data to generate a deviation vector in a unified reference coordinate system specifically includes: Based on the two-dimensional coordinate data in the target positioning information, and combined with the imaging parameters and installation pose information of the image acquisition device, a projection mapping operation is performed to convert the two-dimensional coordinates into three-dimensional reference points. Under the unified reference coordinate system, the three-dimensional reference points are registered with the offset data, and the registered three-dimensional modeling error is dynamically adjusted through a preset weighted compensation strategy to obtain the deviation vector. The step of performing fuzzy control rule calculation based on the deviation vector, and adaptively adjusting parameters in conjunction with historical control data to generate control output quantities specifically includes: The magnitude and rate of change of the deviation vector are used as input variables, corresponding to the error quantity and error rate of change input dimensions in the fuzzy controller, respectively, and the preset fuzzy rule base is called to output the control level. Based on the control residuals and actual response results over multiple consecutive historical control cycles, the proportional factor, integral factor, and derivative factor of the fuzzy controller are dynamically adjusted, and different parameter tuning intervals are selected according to the residual convergence rate, thereby calculating the control output.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the puncture correction method based on dynamic vision as described in any one of claims 1 to 6.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the puncture correction method based on dynamic vision as described in any one of claims 1 to 4.

Images