Machine Vision-Based PCB Tool Motion Control Method and System

Abstract

This invention belongs to the field of industrial data processing technology, specifically relating to a machine vision-based method and system for PCB tool motion control. The method includes: acquiring multivariate sequence data of the tool during processing, normalizing and performing first-order differencing to obtain a stationary sequence; using an improved local anomaly factor algorithm to detect anomalies in the stationary sequence and correcting detected anomalies to obtain a corrected stationary sequence; using an ARIMA model to predict the corrected stationary sequence to obtain a predicted stationary deviation value for the next time step; restoring the predicted stationary deviation value to the original predicted deviation value through inverse differencing, and calculating the tool compensation amount by combining a preset compensation rate and safety margin to adjust the tool's motion trajectory. This invention improves the accuracy of anomaly detection, achieves more precise tool motion control, and enhances the processing quality of PCBs.

Description

Technical Field

[0001] This invention relates to the field of industrial data processing technology. More specifically, this invention relates to a machine vision-based method and system for controlling the motion of PCB cutting tools. Background Technology

[0002] In modern PCB manufacturing, high-speed, large-scale array drilling and milling are key steps in achieving miniaturization and high performance of electronic products. Traditional tool motion control typically relies on pre-set CNC programs to execute machining tasks through open-loop or simple closed-loop control.

[0003] However, in actual industrial production, due to the progressive wear of cutting tools, uneven local hardness of PCB substrate materials, and the generation of small chips or thermal deformation during processing, small, cumulative deviations can occur in the actual machined hole diameter or milled edge. These deviations directly affect the electrical performance and reliability of the PCB.

[0004] To address these deviations, existing technologies have introduced machine vision-based detection methods. These methods involve using a high-precision industrial camera to capture microscopic images of the aperture or edge of the previously processed hole in real time, employing image processing techniques to accurately measure the actual deviation data, and using this deviation data as sequence data for prediction to calculate the tool compensation amount for the next processing step.

[0005] In the prediction process, the Autoregressive Integrated Moving Average (ARIMA) model is typically used. This model excels at capturing long-term trends and periodicity in time series data and can be used to predict the gradual increase in aperture caused by tool wear. The prediction process also incorporates a local anomaly factor algorithm to identify and remove sudden anomalies or transient noise from the data, making the data input into the ARIMA model more accurate and reliable.

[0006] However, throughout the tool's lifecycle, machining deviation data changes from a low-density stable period to a high-density rapid wear period and the scrap period. Local anomaly factor algorithms typically use a fixed neighborhood parameter k for anomaly detection. If the fixed neighborhood parameter k is too small, it may be overly sensitive during the tool's stable period, falsely reporting normal fluctuations; if the fixed neighborhood parameter k is too large, during the rapid wear period, due to the dispersed overall data, it may be slow to react to real critical anomalies, resulting in missed detections.

[0007] Therefore, a fixed neighborhood parameter k is usually used in anomaly detection, which is difficult to adapt to the changes in data density under different tool wear stages. This leads to unstable detection sensitivity of abnormal deviations, which greatly reduces the prediction accuracy and robustness of tool compensation, ultimately resulting in a decrease in PCB processing quality and production efficiency. Summary of the Invention

[0008] The purpose of this invention is to propose a PCB tool motion control method and system based on machine vision, in order to solve the problems of unstable anomaly detection sensitivity and decreased prediction accuracy caused by the fixed neighborhood parameter k in the local anomaly factor algorithm in the prior art; to this end, this invention provides solutions in the following two aspects.

[0009] In a first aspect, the present invention provides a machine vision-based PCB tool motion control method, comprising:

[0010] Multivariate sequence data of the cutting tool during the machining process are collected, and then normalized and subjected to first-order difference processing to obtain a stationary sequence;

[0011] An improved local anomaly factor algorithm is used to detect anomalies in stationary sequences, and the detected anomalies are corrected to obtain a corrected stationary sequence. The improved local anomaly factor algorithm includes an adaptive neighborhood parameter, which is negatively correlated with the wear cycle activity of each stationary data point in the stationary sequence. The wear cycle activity characterizes the distribution characteristics of each stationary data point.

[0012] The ARIMA model is used to predict the corrected stationary sequence and obtain the predicted stationary deviation value for the next time step.

[0013] The predicted stable deviation value is restored to the original deviation prediction value through inverse differential operation, and the tool compensation amount is calculated by combining the preset compensation rate and safety margin to adjust the tool's motion trajectory.

[0014] The above scheme constructs a wear cycle activity index and dynamically calculates adaptive neighborhood parameters, thereby realizing the adaptive adjustment of local anomaly factor algorithm parameters. It can intelligently adjust the detection sensitivity according to the different life cycle stages of the tool, solving the defects of fixed k value being prone to false alarms in the new tool period and prone to missed alarms in the scrap period. It realizes accurate identification and correction of anomalies, ensuring that the sequence input into the ARIMA model is pure trend data, and further calculates the tool compensation amount, realizing more precise tool motion control, and ensuring the final PCB processing quality and product yield.

[0015] Optionally, the wear cycle activity for:

[0016] ;

[0017] in, This represents the mean absolute deviation corresponding to the current processing step t. The reference value representing the average absolute deviation of the tool during the initial stabilization period. This represents the standard deviation corresponding to the current processing step t. It represents an exponential function with the natural constant e as the base; the mean absolute deviation and standard deviation are calculated by the stationary sequence within a preset sliding window, the sliding window is composed of a set number of processing steps, and the processing step corresponding to the end of the window is the current processing step t.

[0018] The above-mentioned wear cycle activity can accurately reflect the changes in tool condition during different processing stages.

[0019] Optionally, the adaptive neighborhood parameter is a rounded value of the neighborhood parameter. for: ;

[0020] in, The neighborhood parameters of the current processing step t, The wear cycle activity level for the current processing step t. The preset minimum neighborhood parameter, This is the preset maximum neighborhood parameter.

[0021] The aforementioned adaptive neighborhood parameters enable adaptive adjustment of the parameters of the local anomaly factor algorithm, solving the problem of poor robustness caused by the fixed k value in the traditional local anomaly factor algorithm, and ensuring the accuracy and reliability of anomaly detection.

[0022] Optionally, the process of obtaining the corrected stationary sequence is as follows:

[0023] The basic local outlier score is obtained by fitting a basic local anomaly factor model to the stationary sequence using the maximum neighborhood parameter.

[0024] The basic outlier score is dynamically corrected using the adaptive neighborhood parameter and the maximum neighborhood parameter to obtain the corrected outlier score;

[0025] The corrected outlier score is compared with a preset outlier threshold to identify outliers in the stationary sequence;

[0026] The outliers are corrected to obtain corrected values, and a corrected stationary sequence is obtained.

[0027] The process of identifying and correcting outliers described above ensures that the sequences input into the ARIMA model are pure trend data, enabling the ARIMA model's prediction results to focus more on capturing the essential trend of tool wear, thereby improving the long-term prediction stability and industrial generalization ability of the entire composite model.

[0028] Optionally, the corrected outlier score is:

[0029] ;

[0030] in, This represents the corrected outlier score for the current processing step t. Represents the basic outlier score. Represents the maximum neighborhood parameter. This represents the adaptive neighborhood parameter of the current processing step t. This indicates the preset amplification factor.

[0031] Optionally, the corrected value is obtained by weighting the outlier with the ARIMA prediction value of the previous time step.

[0032] Optionally, the multivariate sequence data includes data points from multiple processing steps, and each data point includes the corresponding aperture deviation and edge roughness.

[0033] Optionally, the aperture deviation is the difference between the actual aperture and the target aperture; the edge roughness is obtained by calculating the gradient magnitude or average grayscale dispersion of the edge pixels.

[0034] Optionally, the first-order difference processing involves subtracting the data points from the data points of the previous processing step from each data point in the current processing step.

[0035] The first-order differencing process described above can transform multivariate sequence data into stationary sequences, providing stable input data that meets the model requirements for subsequent adaptive local anomaly detection and ARIMA model prediction.

[0036] In the second aspect, the machine vision-based PCB tool motion control system includes:

[0037] processor;

[0038] The memory stores computer instructions for machine vision-based PCB tool motion control, which, when executed by the processor, cause the system to perform the aforementioned machine vision-based PCB tool motion control method.

[0039] The beneficial effects of this invention are as follows:

[0040] This invention obtains the wear cycle activity index and, based on this, achieves dynamic adaptive adjustment of the neighborhood parameters of the local anomaly factor algorithm. This enables anomaly detection to intelligently adjust its detection sensitivity according to different life cycle stages of the tool, thereby solving the problem of fixed parameters in existing technologies. The abnormal values ​​introduced by the false positives and false negatives have greatly improved the robustness and accuracy of anomaly detection.

[0041] Meanwhile, by accurately identifying and correcting outliers through the local anomaly factor algorithm, the trend of the data input into the ARIMA model is kept relatively stable, which enhances the long-term predictive stability and industrial generalization of the ARIMA model. This allows the final calculated tool compensation amount to combine the long-term trend prediction of progressive wear with the real-time correction of transient abnormal disturbances, thus achieving more precise and adaptive control of PCB tool movement and significantly improving the processing quality and product yield of PCB tools. Attached Figure Description

[0042] Figure 1 The flowchart illustrating the steps of the PCB tool motion control method based on machine vision in this embodiment is shown in the schematic diagram.

[0043] Figure 2 A schematic diagram illustrating the comparison between the adaptive local anomaly detection of the present invention and the prior art is shown.

[0044] Figure 3 The schematic diagram illustrates the structural block diagram of the machine vision-based PCB tool motion control system in this embodiment. Detailed Implementation

[0045] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0046] like Figure 1 As shown, the PCB tool motion control method based on machine vision in this embodiment includes the following steps:

[0047] S1: Collect multivariate sequence data of PCB cutting tools, and perform feature normalization and first-order difference processing to obtain stationary sequences.

[0048] In this embodiment, the machine vision system first acquires data from multiple dimensions of the PCB cutting tool during the processing, and then preprocesses this data to obtain a stable sequence.

[0049] The acquisition of multi-dimensional data involves using a high-resolution industrial camera and a matching optical system to capture microscopic images of the hole diameter or milled edge of the PCB tool after the previous processing is completed. Image processing algorithms, such as Canny edge detection and sub-pixel fitting, are then used to accurately measure the data of the PCB tool.

[0050] Specifically, the multi-dimensional data represents two key indicators: aperture deviation. and edge roughness .

[0051] Among them, the aperture deviation is the difference between the actual aperture and the target aperture; the edge roughness is obtained by calculating the gradient magnitude or average grayscale dispersion of the edge pixels.

[0052] Two key indicators from multiple consecutive processing steps are used to construct multivariate sequence data for storage. The data points at the current processing step t within the multivariate sequence data are used as the basis for this storage. For example, , This is the sequence number of the processing step, where T represents transpose, and this data point... It can be viewed as vector data.

[0053] It should be noted that the sequence number of the processing steps corresponds to the processing time, that is, one processing step corresponds to one processing time.

[0054] The preprocessing in this embodiment includes feature normalization and trend stabilization. Specifically, since aperture deviation and edge roughness have different dimensions, normalization is required first to normalize the data of each dimension in the multivariate sequence data. The interval is used to eliminate the impact of dimensional differences on subsequent model calculations.

[0055] Secondly, first-order difference operations are used to make each data point in the multivariate sequence data stationary, so as to eliminate the long-term linear or nonlinear growth trend caused by tool wear and obtain a stationary sequence.

[0056] Data points under the current processing step t For example, calculate data points Corresponding stationary data points Specifically:

[0057] ;

[0058] in, These are stationary data points under the current processing step t. These are the data points under the current processing step t. These are the data points under processing step t-1.

[0059] It should be noted that the above-mentioned stable data points are the deviation data between data points in two adjacent processing steps, which is a vector containing the deviation of data in two dimensions.

[0060] All stationary data points are used to form a stationary sequence, which includes the instantaneous changes in tool wear and transient environmental disturbances.

[0061] In this embodiment, feature normalization and first-order difference processing are performed on the multivariate data composed of the collected PCB tool data to transform the data into a stationary sequence that meets the input requirements of the time series model, thus laying the data foundation for subsequent anomaly correction and prediction.

[0062] S2: Based on the stationary sequence, calculate the wear cycle activity of the PCB tool, and determine the adaptive neighborhood parameters of the local anomaly factor algorithm according to the wear cycle activity.

[0063] The process for obtaining the wear cycle activity in this embodiment is as follows:

[0064] First, obtain the activity index under the current processing step.

[0065] In this embodiment, the stationary sequence is analyzed within a sliding window of a set length. The activity index is constructed using statistical characteristics within the range, including mean absolute deviation and standard deviation.

[0066] Specifically, a sliding window of a set length that precedes the current machining step t and includes the current machining step t. Within the sliding window, calculate the mean absolute deviation and standard deviation of all stationary data points.

[0067] Mean absolute deviation represents the most recent The standard deviation measures the average variation of all stationary data points within a single processing step.

[0068] The mean absolute deviation is:

[0069] ;

[0070] in, These are stationary data points under the current processing step t. The mean absolute deviation The length is the module length.

[0071] Secondly, based on the mean absolute deviation and standard deviation, the wear cycle activity of the corresponding stable data points is calculated.

[0072] The wear cycle activity is as follows:

[0073] ;

[0074] in, This indicates the wear cycle activity level under the current processing step t. This represents the reference mean absolute deviation during the initial stabilization period of the tool. This represents the standard deviation at the current processing step t. This represents an exponential function with the natural constant e as its base.

[0075] For example, suppose a tool has a reference average absolute deviation during the initial stabilization period. for During the stable period time, , The wear cycle activity During the period of rapid wear and tear , , The wear cycle activity .

[0076] It can be seen that the wear cycle activity can accurately reflect the transition of the tool condition from the stable period (2.06) to the rapid wear period (9.88).

[0077] In this embodiment, based on wear cycle activity The adaptive neighborhood parameters of the Local Outlier Factor (LOF) algorithm are determined. Specifically, adaptive neighborhood parameters This is obtained by rounding down the neighborhood parameters. for:

[0078] ;

[0079] in, This represents the neighborhood parameters under the current processing step t. This represents the wear cycle activity level under the current processing step t. This represents the preset minimum neighborhood parameter. This represents the preset maximum neighborhood parameter.

[0080] The above rounding can be done by rounding to the nearest integer, rounding up, or rounding down.

[0081] It should be noted that the adaptive neighborhood parameter in this embodiment is not 0. Also, since the wear cycle activity is always a value greater than 0, therefore... The value of is necessarily not equal to 0.

[0082] For example, suppose , Stable period hour, Rounding Rapid wear and tear period hour, Rounding .

[0083] The above embodiments achieve adaptive adjustment of the parameters of the local anomaly factor algorithm by constructing a wear cycle activity index and dynamically calculating adaptive neighborhood parameters. This solves the problem of poor robustness caused by the fixed k value of the traditional local anomaly factor algorithm, and ensures the accuracy and reliability of anomaly detection.

[0084] S3: Using the adaptive neighborhood parameters, anomaly detection and correction are performed on the stationary sequence to obtain the corrected stationary sequence. Based on the ARIMA model, the corrected stationary sequence is predicted to obtain the predicted stationary deviation value at the next time step.

[0085] The process of obtaining the corrected stationary sequence is as follows:

[0086] First, obtain the base outlier score for each stationary data point.

[0087] Specifically, the maximum neighborhood parameter is used for the entire stationary sequence. A basic local outlier factor model is fitted to obtain the basic outlier scores for each stationary data point.

[0088] Secondly, using adaptive neighborhood parameters The basic outlier scores are dynamically adjusted to obtain the adjusted outlier scores. :

[0089] ;

[0090] in, This represents the corrected outlier score for the current processing step t. Represents the basic outlier score. Represents the maximum neighborhood parameter. This represents the adaptive neighborhood parameter of the current processing step t. This indicates the preset amplification factor.

[0091] The preset amplification factor is set to 1.2. By introducing the amplification factor, the abnormal situation of the data during the rapid wear period can be further amplified to prevent missed reports.

[0092] Then, the outlier score will be corrected. Compared with the preset abnormal threshold Compare, if Then it is believed If an outlier is identified, it is corrected to obtain a corrected value; otherwise, no correction is performed, and a corrected stationary sequence is obtained.

[0093] The corrected value was obtained by taking a weighted average of the anomaly and the predicted value at the previous time point.

[0094] Among them, the corrected value for:

[0095] ;

[0096] in, This indicates the corrected value. Indicates the adjusted weights. This represents the predicted stationary deviation value after the ARIMA model prediction at the previous time step. This represents a stationary data point under the current processing step t.

[0097] The aforementioned corrected weight is set to 0.7.

[0098] The above corrections can prevent outliers from erroneously affecting the predicted trend of ARIMA.

[0099] In this embodiment, the corrected stationary sequence is input into the ARIMA model to predict the stationary bias at the next time step. .

[0100] The above-mentioned stationary deviation prediction values ​​represent the trend deviation of tool wear.

[0101] In this embodiment, by dynamically adjusting the neighborhood parameters of anomaly detection, anomaly points are accurately identified and corrected, ensuring the purity of the data input to the ARIMA model and improving the stability and accuracy of the ARIMA model in predicting tool wear trends.

[0102] S4: Based on the predicted stable deviation value, it is restored to the original predicted deviation value through inverse differential operation, and combined with the preset compensation rate and safety margin, the final tool compensation amount is calculated to adjust the tool's motion trajectory.

[0103] The process of obtaining the tool compensation amount is as follows:

[0104] First, the predicted stationary deviation is inversely differentiald (restored) to obtain the original predicted deviation value.

[0105] Specifically, the predicted stationary deviation value for the next time step as predicted by the ARIMA model. Perform an inverse difference operation to obtain the original bias prediction value for the next time step. Specifically:

[0106] ;

[0107] in, This represents the original deviation prediction value at the next time step. For the data points in the current processing step t, This represents the predicted stationary deviation value at the next moment.

[0108] Next, calculate the tool compensation amount.

[0109] Specifically, based on the original deviation prediction value Calculate the tool compensation amount required for the tool motion system. Specifically:

[0110] ;

[0111] in, For the compensation rate, For safety margin.

[0112] The range of values ​​for the above compensation rate is as follows: The safety margin is a very small fixed offset used to prevent overcutting.

[0113] The above tool compensation amount The acquisition of this data integrates information on the long-term trend prediction of tool progressive wear and the real-time correction of transient abnormal disturbances.

[0114] For example, assume the original deviation is measured. The predicted stationary deviation value of the ARIMA model Then the predicted value of the original deviation at the next moment .

[0115] When the compensation rate Safety margin At that time, tool compensation amount At this point, the tool compensation amount is sent to the CNC motion control system in real time to adjust the tool's motion trajectory.

[0116] The above embodiments restore the predicted stationary deviation value of the ARIMA model to the original deviation in the actual physical space, and introduce compensation rate and safety margin for fine adjustment, thereby obtaining the tool compensation amount that can simultaneously cope with long-term trends and instantaneous disturbances, and realizing precise adaptive control of the movement of the machining tool.

[0117] Figure 2 This diagram illustrates the comparison between the adaptive local anomaly detection of the present invention and existing technologies. The diagram shows the measured hole diameter deviation values ​​under different machining steps and indicates the stable wear period, rapid wear period, and scrap period of the tool. Existing technologies, when detecting anomalies, may falsely report hollow dots during the stable wear period and miss critical anomalies during the scrap period due to insufficient sensitivity. However, the solution of the present invention accurately captures a very small number of transient high peaks of interest during the stable wear period using coarse cross-shaped dots, eliminating false alarms from existing technologies. During the rapid wear period and scrap period, it captures all high-amplitude deviation points requiring correction and accurately identifies the highest peaks deviating from the local average level.

[0118] The solution of this invention can intelligently adjust the detection sensitivity according to the different processing cycle stages of the tool, which solves the defects of false alarms during the new tool period and false alarms during the scrap period when the fixed k value is easy to detect. It realizes accurate identification and correction of abnormal points and ensures that the sequence input into the ARIMA model is pure trend data, which provides accurate tool compensation amount for subsequent tool motion control, and ensures the final PCB processing quality and product yield.

[0119] This invention also provides a PCB tool motion control system based on machine vision. For example... Figure 3 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the machine vision-based PCB tool motion control method according to the present invention.

[0120] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and therefore will not be described in detail here.

[0121] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented by computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0122] In the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

[0123] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A machine vision based PCB tool motion control method, characterized by, The method comprises the following steps: Collecting multivariate sequence data of a tool in a machining process, normalizing and first-order differentiating the data to obtain a stationary sequence; The improved local anomaly factor algorithm is used for anomaly detection on the stationary sequence, and the detected anomaly points are corrected to obtain a corrected stationary sequence; wherein the improved local anomaly factor algorithm comprises an adaptive neighborhood parameter, the adaptive neighborhood parameter is negatively correlated with a wear cycle activity degree of each stationary data point in the stationary sequence; the wear cycle activity degree represents a distribution characteristic of each stationary data point; and the wear cycle activity degree is calculated according to the following formula: is: ； an average absolute deviation corresponding to the current machining step sequence t, a reference value of the average absolute deviation of the tool in the initial stable period, a standard deviation corresponding to the current machining step sequence t, an exponential function with the natural constant e as the base; the average absolute deviation and the standard deviation are calculated by taking the stationary sequence in the preset sliding window, the sliding window is composed of a set number of machining step sequences, and the machining step sequence corresponding to the end of the window is the current machining step sequence t; The adaptive neighborhood parameter is a rounded value of the neighborhood parameter is: ; a neighborhood parameter for the current machining step sequence t, a wear cycle activity for the current machining step sequence t, a preset minimum neighborhood parameter, a preset maximum neighborhood parameter; Using an ARIMA model to predict the modified stationary sequence to obtain a stationary deviation prediction value at the next moment; the process of obtaining the modified stationary sequence is as follows: Using a maximum neighborhood parameter to fit a basic local anomaly factor model to the stationary sequence to obtain a basic outlier score; Using an adaptive neighborhood parameter and a maximum neighborhood parameter to dynamically correct the basic outlier score to obtain a corrected outlier score; Comparing the corrected outlier score with a preset anomaly threshold to identify an abnormal point in the stationary sequence; Correcting the abnormal point to obtain a corrected value and a modified stationary sequence; Restoring the stationary deviation prediction value to an original deviation prediction value through an inverse difference operation, and calculating a tool compensation amount by combining a preset compensation rate and a safety margin to adjust the motion trajectory of the tool.

2. The machine vision-based PCB tool motion control method of claim 1, wherein, The corrected outlier score is: ； wherein, denotes the modified outlier score for the current processing step t, denotes the base outlier score, denotes the maximum neighborhood parameter, denotes the adaptive neighborhood parameter for the current processing step t, denotes a preset amplification factor.

3. The machine vision-based PCB tool motion control method of claim 1, wherein, The corrected value is obtained by weighted averaging the abnormal point and an ARIMA prediction value at the previous moment.

4. The machine vision-based PCB tool motion control method of claim 1, wherein, The multivariate sequence data comprises data points at multiple machining step sequences, and each data point comprises a corresponding aperture deviation and edge roughness.

5. The machine vision-based PCB tool motion control method of claim 4, wherein, The aperture deviation is the difference between an actual aperture and a target aperture; and the edge roughness is obtained by calculating the gradient amplitude or average gray level dispersion of edge pixel points.

6. The machine vision-based PCB tool motion control method of claim 1, wherein, The first-order differentiation processing is to subtract each data in a data point at a current machining step sequence from each data in a data point at a previous machining step sequence.

7. A machine vision based PCB tool motion control system characterized by, The system comprises: A processor; A memory storing computer instructions for machine vision-based PCB tool motion control, which, when executed by the processor, causes the system to perform the machine vision-based PCB tool motion control method according to any one of claims 1-6.

Images