Dimension measurement scoring system and method based on machine vision

Abstract

The invention relates to the technical field of machine vision and operation evaluation, and discloses a size measurement scoring system and method based on machine vision. The method includes parsing out a pause event and associating the pause event with a predefined standard measurement phase by synchronously capturing a continuous spatial position stream and a timestamp stream during a measurement process, forming a sequence of steps to check a logical sequence and generate a sequence deviation factor. Meanwhile, preset image features are dynamically detected, feature capture moments of the features are recorded, position point sets in time windows before and after the moments are extracted, and local trajectory complexity indexes are calculated. And integrating the total time consumption, the sequence deviation factor and the local trajectory complexity index, calculating an original operation efficiency value, carrying out normalized fusion on the original operation efficiency value and the reference size value, and finally outputting a comprehensive measurement score. According to the invention, automatic comprehensive evaluation of the standardization of the measurement operation flow and the stability of the operation process is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine vision and operation evaluation, in particular to a size measurement scoring system and method based on machine vision. BACKGROUND

[0002] In the field of industrial measurement, size measurement methods based on machine vision have been widely applied. Existing technical solutions mainly rely on discrete, trigger-based image acquisition and static analysis of measurement tools or measured objects. The system usually captures a single frame of image to identify feature positions and calculate sizes after the operator completes a measurement action and is stationary. The core of this technical path is to obtain and process isolated, state points representing measurement results, which is passively responsive to explicit completion actions of the operator or external instructions.

[0003] Such conventional technology has limitations. Its technical architecture cannot automatically analyze and logically understand continuous, natural operation processes. The execution order of steps and whether the transition is standardized are outside the system's perception range, resulting in a lack of evaluation means for the standardization of the operation process itself. At the same time, existing methods only focus on spatial coordinates in the final static position, completely discarding continuous dynamic trajectory information during the acquisition of the coordinates. The stability during the execution of key measurement actions and the precision of positioning trajectories, which reflect the quality of the operation, are ignored by the system, and these data are crucial for evaluating the rigor of the measurement and the repeatability of the results.

[0004] A technology is needed that can automatically analyze structured measurement step sequences from continuous operation data streams and implement dynamic quality evaluation of the execution process of key measurement actions. This requires the system to have the ability to identify key stages from continuous spatiotemporal data and correlate standard steps, as well as to quantitatively analyze local operation trajectory features around the measurement instant, thereby establishing a comprehensive evaluation system that integrates process quality and result accuracy. SUMMARY

[0005] The purpose of the present application is to provide a size measurement scoring system and method based on machine vision to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides a size measurement scoring method based on machine vision, which comprises:

[0007] establishing a multi-step measurement framework for the measured object, and synchronously capturing a continuous spatial position stream and a corresponding timestamp stream during the measurement process, identifying a pause event by analyzing the continuous spatial position stream, and correlating each pause event with a standard measurement stage in the multi-step measurement framework to generate a correlated step sequence;

[0008] According to the sequence of the association steps, it is checked whether the logical order of the standard measurement stages is followed, and when it is found that the logical order is out of place, a sequence deviation event is marked and a corresponding sequence deviation factor is generated;

[0009] In the continuous spatial position stream, it is dynamically detected whether an image region conforming to a preset morphological feature appears, and when the image region is detected, the time when the image region conforming to the preset morphological feature is detected is recorded as a feature capture time, and a position point set within a time window before and after the feature capture time is extracted, and the position point set is recorded as an effective measurement point set;

[0010] Based on the effective measurement point set, a local trajectory complexity index is calculated;

[0011] The total time consumption from the start of measurement to the end of measurement is obtained, and the sequence deviation factor and the local trajectory complexity index are combined to calculate an original operation efficiency value;

[0012] A pre-stored reference size value for the measured object is introduced, and the original operation efficiency value and the reference size value are normalized and fused to finally output a comprehensive measurement score.

[0013] Preferably, the step of identifying a pause event by analyzing the continuous spatial position stream comprises:

[0014] A minimum duration threshold and a maximum displacement tolerance threshold are set;

[0015] In the continuous spatial position stream, an analysis window of a fixed time length is slid in time sequence;

[0016] The standard deviation of the spatial coordinates of all position points in each analysis window is calculated;

[0017] When the standard deviation of the spatial coordinates is less than the maximum displacement tolerance threshold, and the duration of this state exceeds the minimum duration threshold, it is determined that a pause event occurs in the time period covered by the analysis window;

[0018] The average spatial coordinates of all position points in the time period are extracted as the representative position points of the pause event.

[0019] Preferably, the step of associating each pause event with a standard measurement stage in the multi-step measurement framework comprises:

[0020] The theoretical spatial region range of each standard measurement stage in the multi-step measurement framework is obtained;

[0021] The Euclidean distance between the representative position points of the pause event and the center points of each theoretical spatial region range is calculated.

[0022] selecting the standard measurement phase corresponding to the theoretical space region range with the minimum Euclidean distance as a candidate association phase;

[0023] verifying whether the representative position point of the pause event is located within the theoretical space region range of the candidate association phase;

[0024] if located within, successfully binding the current pause event with the candidate association phase, and generating a step binding record;

[0025] collecting all step binding records in chronological order to form the associated step sequence.

[0026] Preferably, the step of traversing to check whether the logical order of the standard measurement phases is followed comprises:

[0027] extracting, from the associated step sequence, the phase identifier of the standard measurement phase bound by each step binding record in sequence;

[0028] arranging the phase identifiers extracted in sequence in order to generate an actual execution sequence;

[0029] bit-by-bit comparing the actual execution sequence with the preset logical order in the multi-step measurement framework;

[0030] when a phase identifier is found in the actual execution sequence, the position of which in the preset logical order appears before the position of the phase identifier corresponding to the previous phase identifier, it is determined that a sequence misplacement occurs;

[0031] counting the total number of all sequence misplacements occurring in the entire actual execution sequence, and converting the total number into the sequence deviation factor according to a preset mapping relationship.

[0032] Preferably, the step of dynamically detecting whether an image region meeting the preset morphological characteristics appears comprises:

[0033] real-time acquiring a video frame under a measurement view angle collected by an image sensor on a measurement device;

[0034] applying a binaryzation segmentation operation to each video frame to obtain a corresponding binaryzation image;

[0035] finding all connected pixel regions in the binaryzation image through a region growing algorithm;

[0036] calculating a geometric feature descriptor of each connected pixel region, the geometric feature descriptor including area, perimeter and rectangularity;

[0037] The geometric feature descriptor of each connected pixel region is matched with feature descriptors in a pre-stored feature template library;

[0038] When the matching similarity of a connected pixel region exceeds a pre-set similarity threshold, it is determined that an image region conforming to the pre-set morphological feature is detected.

[0039] Preferably, the step of calculating the local trajectory complexity index based on the effective measurement point set comprises:

[0040] The spatial coordinates of each position point are extracted from the effective measurement point set;

[0041] The spatial coordinates of all position points are fitted into a smooth curve;

[0042] The variance value of the first derivative sequence of the smooth curve is calculated, and the variance value is recorded as the direction change intensity;

[0043] The average value of the curvature sequence of the smooth curve is calculated, and the average value is recorded as the path bending degree;

[0044] The direction change intensity and the path bending degree are weighted and summed, and the result is the local trajectory complexity index.

[0045] Preferably, the step of calculating the original operation performance value comprises:

[0046] A first timestamp corresponding to the measurement start time and a second timestamp corresponding to the measurement end time are obtained;

[0047] The difference between the second timestamp and the first timestamp is calculated to obtain the total time consumption;

[0048] An efficiency calculation model including a sequence deviation factor, a local trajectory complexity index, and a total energy consumption is established;

[0049] The sequence deviation factor is input into a first calculation channel of the efficiency calculation model to obtain a sequence penalty component;

[0050] The local trajectory complexity index is input into a second calculation channel of the efficiency calculation model to obtain a trajectory regularity component;

[0051] The total time consumption is input into a third calculation channel of the efficiency calculation model to obtain a time efficiency component;

[0052] The sequence penalty component, the trajectory regularity component, and the time efficiency component are combined to output the original operation performance value.

[0053] Preferably, the step of applying a binaryzation segmentation operation to each video frame comprises:

[0054] selecting a global threshold value on a specific color space channel of the video frame;

[0055] comparing the component value of each pixel in the video frame on the color space channel with the global threshold value;

[0056] setting the pixel with the component value greater than the global threshold value as a foreground pixel, and setting the pixel with the component value less than or equal to the global threshold value as a background pixel;

[0057] performing a morphological opening operation on the set image to remove small noise points in the foreground;

[0058] performing a morphological closing operation on the image after the morphological opening operation to fill small holes in the foreground, to obtain a final binaryzation image.

[0059] Preferably, the step of fitting the spatial coordinates of all position points into a smooth curve comprises:

[0060] using a non-uniform rational B-spline curve fitting method, taking the spatial coordinates of all position points in the effective measurement point set as control points;

[0061] assigning a weight value to each control point and determining a node vector;

[0062] generating a parametric equation of the smooth curve according to the control points, the weight values and the node vector;

[0063] calculating the coordinates of a series of densely sampled points on the curve through the parametric equation, for representing the fitted smooth curve.

[0064] Preferably, the present application further comprises a size measurement scoring system based on machine vision, which comprises a memory, a processor and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of the size measurement scoring method based on machine vision as described above when executing the computer program.

[0065] Compared with the prior art, the present application has the following advantages:

[0066] By synchronously capturing a continuous stream of timestamped spatial locations and automatically identifying and correlating pause events, a standardized sequence of steps can be reconstructed from unstructured, coherent operations. This enables the system to automatically verify the actual operational process according to a predefined logical order, generating a quantified sequence deviation factor when a step misalignment is detected. This method transforms the measurement of process standardization from a qualitative judgment into a calculable evaluation indicator, achieving an objective assessment of whether operators strictly adhere to standard operating procedures.

[0067] The system dynamically detects preset image features and records their occurrence times, then extracts a set of continuous position points within a time window before and after that moment to form an effective measurement point set. The local trajectory complexity index calculated based on this point set is essentially a quantitative description of the operator's fine motion control ability during the execution of key measurement actions. This index effectively reflects dynamic characteristics such as micro-jitter and unnecessary adjustments during the positioning process. Incorporating this index, which reflects process stability, into the evaluation system ensures that the final comprehensive score simultaneously reflects the accuracy of the measurement results and the quality of the operational process in obtaining those results, thus distinguishing operational performance with different levels of stability at the same result accuracy. Attached Figure Description

[0068] Figure 1 This is a schematic diagram illustrating the working principle of the machine vision-based size measurement and scoring method described in this invention.

[0069] Figure 2 A flowchart for identifying pause events;

[0070] Figure 3 A line graph showing the average Euclidean distance during the standard measurement phase;

[0071] Figure 4 A multi-dimensional polyline plot of the geometric features of connected pixel regions;

[0072] Figure 5 This is a flowchart for calculating the local trajectory complexity index. Detailed Implementation

[0073] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0074] Please see Figure 1The present application provides a machine vision-based dimension measurement scoring method, which comprises: establishing a multi-step measurement framework for the measured object, and synchronously capturing a continuous spatial position stream and a corresponding timestamp stream generated by a measurement device during the measurement process. The continuous spatial position stream is analyzed to identify pause events occurring therein, each identified pause event is associated with a predefined standard measurement phase in the multi-step measurement framework, thereby generating a time-sequentially arranged associated step sequence. According to the associated step sequence, it is checked whether the occurrence order of the standard measurement phases follows the logically preset order of the framework, and when a sequence deviation is found, a sequence deviation event is marked and a corresponding sequence deviation factor is generated. During the capturing of the continuous spatial position stream, it is dynamically detected whether an image region conforming to a preset morphological feature appears in the video stream collected by the image sensor, and once the image region is detected, the current time is immediately recorded as a feature capture time, and an effective measurement point set composed of all position points within a certain time window before and after the time is extracted. Based on the effective measurement point set, a local trajectory complexity index is calculated. The total time consumption from the beginning of the measurement to the end of the measurement is obtained, and the aforementioned sequence deviation factor and the local trajectory complexity index are combined to obtain an original operation efficiency value through a preset efficiency calculation model. A pre-stored reference dimension value for the measured object is introduced, and the original operation efficiency value and the reference dimension value are normalized and fused, and finally a comprehensive measurement score is output.

[0075] Embodiment 1: refer to Figure 2 A minimum duration threshold and a maximum displacement tolerance threshold are set. In the continuous spatial position stream, a fixed-length analysis window is slid in time sequence. The standard deviation of the spatial coordinates of all position points in each analysis window is calculated. When the standard deviation of the spatial coordinates is less than the maximum displacement tolerance threshold, and the duration of this low standard deviation state exceeds the minimum duration threshold, it is determined that a pause event occurs in the time period covered by the analysis window. The average spatial coordinates of all position points in this time period are extracted as the representative position points of the pause event.

[0076] In specific implementation, a minimum duration threshold and a maximum displacement tolerance threshold are set, the minimum duration threshold is used to define the minimum time length that an effective pause should last, and the maximum displacement tolerance threshold is used to define the maximum position fluctuation range within which a point set is considered to be sufficiently stationary. In some embodiments, in the continuous spatial position stream, a fixed-length analysis window is slid in time sequence, and the length of the analysis window is predefined. For each analysis window, the standard deviation of the spatial coordinates of all position points in the window is calculated. The calculation of the standard deviation of the spatial coordinates represents the dispersion degree of all position points in the analysis window relative to their average position. The standard deviation of the spatial coordinates σ can be calculated by the following formula:

[0077]

[0078] wherein: denotes the spatial coordinate standard deviation, denotes the number of position points within the analysis window, denotes the spatial coordinate vector of the th position point, denotes the mean vector of the spatial coordinates of all position points within the analysis window, denotes the Euclidean norm of a vector. The calculated spatial coordinate standard deviation is used for comparison with the maximum displacement tolerance threshold.

[0079] In some embodiments, when the calculated spatial coordinate standard deviation is less than the set maximum displacement tolerance threshold, and the state of the spatial coordinate standard deviation being below the maximum displacement tolerance threshold is continuously maintained for a time duration exceeding a minimum duration threshold, it is determined that a pause event has occurred within the time period covered by the current analysis window. It is understood that the maximum displacement tolerance threshold defines an upper limit of the spatial tightness of the position point cluster, and the minimum duration threshold defines a lower limit of the time duration for which the tightness state needs to be maintained, and together they constrain the physical definition of a pause event. A plurality of consecutive analysis windows all satisfy the above conditions, and these windows collectively cover a continuous time period, which is identified as one complete pause event.

[0080] Optionally, after identifying the time period corresponding to the pause event, the average spatial coordinates of all position points within the time period are extracted as the representative position point of the pause event. The calculation method of the representative position point is to take the arithmetic mean of the spatial coordinate components of each position point captured within the time period, generating a new coordinate point.

[0081] Example 2: Obtain the theoretical spatial range of each standard measurement stage in the multi-step measurement framework. Calculate the Euclidean distance between the representative location of the pause event and the center point of each theoretical spatial range. Select the standard measurement stage corresponding to the theoretical spatial range with the smallest Euclidean distance as a candidate associated stage. Verify whether the representative location of the pause event is within the theoretical spatial range of the candidate associated stage. If it is, successfully bind the current pause event to the candidate associated stage, generating a step binding record. Gather all step binding records in chronological order to form an associated step sequence. Extract the stage identifier of the standard measurement stage bound to each step binding record from the associated step sequence, and arrange the extracted stage identifiers in order to generate an actual execution sequence. Compare the actual execution sequence with the preset logical order in the multi-step measurement framework position by position. When a stage identifier in the actual execution sequence is found to be in a position before the position corresponding to the previous stage identifier in the preset logical order, it is determined that a sequence misalignment has occurred. The total number of sequence misalignments that occur throughout the entire actual execution sequence is counted, and this total number is converted into a sequence deviation factor based on a preset mapping relationship.

[0082] In practice, the theoretical spatial region of each standard measurement stage in the multi-step measurement framework is obtained. This theoretical spatial region is a pre-defined boundary in space for each standard measurement stage. The Euclidean distance between the representative location point of the pause event and the center point of each theoretical spatial region is calculated. The calculation of the Euclidean distance quantifies the spatial proximity of the representative location point to the center of each theoretical spatial region. It can be calculated using the following formula:

[0083]

[0084] in: The three-dimensional spatial coordinates representing the location of the pause event. In a multi-step measurement framework, the first step is... The three-dimensional spatial coordinates of the center point of the theoretical spatial region within each standard measurement stage. That is, representing the position point and the first The Euclidean distance between the center points is compared. The Euclidean distances calculated for all standard measurement phases are also compared. The numerical value is used to select the standard measurement stage corresponding to the theoretical spatial region with the smallest Euclidean distance as the candidate correlation stage.

[0085] In some embodiments, it is verified whether the representative location of the pause event is within the theoretical spatial region of the candidate association stage. The verification process is completed by comparing whether each coordinate component of the representative location falls within the corresponding coordinate axis interval defined by the theoretical spatial region of the candidate association stage. If the representative location is within the theoretical spatial region of the candidate association stage, the current pause event is successfully bound to the candidate association stage, generating a step binding record containing the pause event identifier, representative location information, and the bound standard measurement stage identifier. All step binding records are compiled in chronological order to form an associated step sequence. From the associated step sequence, the stage identifier of the standard measurement stage bound to each step binding record is extracted sequentially. The stage identifier is a symbol that uniquely identifies a standard measurement stage. The extracted stage identifiers are arranged in the extraction order to generate an actual execution sequence, which reflects the order in which the standard measurement stages associated with the pause event are triggered during the operation.

[0086] Optionally, the actual execution sequence is compared bit-by-bit with the preset logical order in the multi-step measurement framework. The preset logical order in the multi-step measurement framework is an ordered list of standard measurement stage identifiers, defining the correct order of operations. During the bit-by-bit comparison, when a stage identifier is found in the actual execution sequence whose position number in the preset logical order appears before the corresponding position number of the preceding stage identifier in the preset logical order, a sequence misalignment is determined to have occurred. It can be understood that the determination of a sequence misalignment means that a standard measurement stage that should theoretically be executed later has been executed prematurely in the operation process. The total number of sequence misalignments occurring throughout the entire actual execution sequence is counted, and the total number is converted into a sequence deviation factor according to a preset mapping relationship. The sequence deviation factor is a numerical value used to quantify the degree of sequence error.

[0087] See Figure 3 This is a line graph showing the average Euclidean distance during the standard measurement phase. The vertical axis, "Average Euclidean Distance," represents the average three-dimensional spatial distance between all points at the current stage of the measurement operation and the center point of the corresponding theoretical reference area. A higher value indicates a greater deviation of the operation position from the reference area, resulting in lower accuracy. The horizontal axis covers the entire standard process of dimensional measurement, corresponding to key operational steps in machine vision measurement. "Extracting dimensional features" is clearly identified as the stage with the highest risk of deviation, requiring focused optimization of operational procedures at this stage. This graph serves as a process quality monitoring tool for machine vision dimensional measurement—by tracking the degree of spatial deviation at each stage, it achieves closed-loop management of "visualizing the operation process → locating weak points → targeted accuracy optimization," ultimately improving the overall accuracy of dimensional measurement.

[0088] In one embodiment, video frames under the measurement view are acquired in real time by an image sensor on the measurement device. A binaryzation segmentation operation is applied to each frame of the video frames, a global threshold is selected on a specific color space channel of the video frame, the component value of each pixel in the video frame on the specific color space channel is compared with the global threshold, the pixels with component value greater than the global threshold are set as foreground pixels, and the pixels with component value less than or equal to the global threshold are set as background pixels. A morphological opening operation is performed on the set image to remove small noise points in the foreground, and a morphological closing operation is performed on the image after the morphological opening operation to fill small holes in the foreground, to obtain a final binaryzation image. All connected pixel regions are found in the obtained binaryzation image by a region growing algorithm. The geometric feature descriptor of each connected pixel region is calculated, and the geometric feature descriptor includes area, perimeter and rectangularity. The geometric feature descriptor of each connected pixel region is similarity matched with the feature descriptors in the pre-stored feature template library, and when there is a connected pixel region with a matching similarity exceeding a preset similarity threshold, it is determined that an image region conforming to a preset morphological feature is detected.

[0089] In a specific implementation, video frames under the measurement view are acquired in real time by an image sensor on the measurement device, and the image sensor continuously captures images during the measurement process. A binaryzation segmentation operation is applied to each frame of the video frames acquired in real time, and the purpose of the binaryzation segmentation operation is to separate the foreground target from the background in the image. In a specific implementation, a global threshold is selected on a specific color space channel of the video frame, and the color space can be grayscale, a component channel of HSV or other. The component value of each pixel in the video frame on the specific color space channel is compared with the global threshold, and the comparison operation is performed pixel by pixel. The pixels with component value greater than the global threshold are set as foreground pixels, and the pixels with component value less than or equal to the global threshold are set as background pixels, to generate an initial binary image through the comparison and setting operations. A morphological opening operation is performed on the set image to remove small noise points in the foreground, and the morphological opening operation is an operation of erosion followed by dilation. A morphological closing operation is performed on the image after the morphological opening operation to fill small holes in the foreground, and the morphological closing operation is an operation of dilation followed by erosion.

[0090] In some embodiments, a region growing algorithm is used to find all connected pixel regions in the final binarized image. The region growing algorithm aggregates neighboring pixels with similar attributes, starting from a seed pixel. A geometric feature descriptor is calculated for each connected pixel region, including area, perimeter, and rectangularity. The area describes the total number of pixels contained in the connected pixel region, the perimeter describes the length of the outer boundary of the connected pixel region, and the rectangularity describes the ratio of the area of ​​the connected pixel region to its smallest bounding rectangle. The geometric feature descriptor of each connected pixel region is then matched for similarity with feature descriptors in a pre-stored feature template library. The feature template library stores standard geometric feature data of the preset morphological features to be detected. The results of the similarity matching are then presented. It can be calculated using the following formula:

[0091]

[0092] in: This represents the area of ​​the currently connected pixel region. This represents the perimeter of the currently connected pixel region. This represents the rectangularity of the currently connected pixel region. The first one in the feature template library The area of ​​each feature template, The first one in the feature template library The perimeter of each feature template, The first one in the feature template library The rectangularity of each feature template. These are preset weighting coefficients corresponding to the three feature components of area, perimeter, and rectangularity, respectively. , Indicates the currently connected pixel region and the first... The matching similarity of each feature template. When there is a connected pixel region, the matching similarity... When the similarity exceeds a preset threshold, it is determined that an image region matching the preset morphological characteristics has been detected in the current video frame.

[0093] See Figure 4This is a multi-dimensional polyline chart of the geometric features of the connected pixel area. The area fluctuation amplitude is large, reaching a peak (about 800) in areas 2, 5, and 8, and a trough (about 500) in areas 4, 7, and 10; the perimeter is stable at 200-300 overall, with only areas 4 and 7 showing a small decrease; the rectangularity reflects the closeness of the area shape to a rectangle, with the rectangularity of areas 2, 5, and 8 approaching 0.95, and that of area 9 dropping to 0.6. This chart is used for morphological feature analysis in machine vision size measurement. The fluctuations in area and perimeter reflect the spatial proportion and boundary complexity of different areas; changes in rectangularity can assist in identifying the shape type of the target area; this multi-dimensional comparison can provide geometric feature basis for subsequent "feature template matching" and help determine whether the area meets the preset morphology.

[0094] Embodiment 4: see Figure 5 The spatial coordinates of each position point are extracted from the effective measurement point set. A non-uniform rational B-spline curve fitting method is used to fit the spatial coordinates of all position points in the effective measurement point set as control points. A weight value is assigned to each control point and a node vector is determined. The parameter equation of the smooth curve is generated based on the control points, weight values, and node vectors. The coordinates of a series of densely sampled points on the curve are calculated through the parameter equation of the smooth curve to represent the fitted smooth curve. The variance value of the first derivative sequence of the smooth curve is calculated based on the smooth curve, and the variance value is recorded as the direction change intensity. The average value of the curvature sequence of the smooth curve is calculated, and the average value is recorded as the path bending degree. The direction change intensity and the path bending degree are weighted and summed, and the result is the local trajectory complexity index.

[0095] In a specific implementation, the spatial coordinates of each position point are extracted from the effective measurement point set, which is a collection of all position points captured within a specific time window before and after the feature capture moment. A non-uniform rational B-spline curve fitting method is used to fit the spatial coordinates of all position points in the effective measurement point set as control points. A weight value is assigned to each control point, which is used to adjust the influence of the corresponding control point on the curve shape. At the same time, a node vector is determined, which is a non-decreasing real number sequence used to define the parameterization range of the B-spline basis function. The parameter equation of the smooth curve is generated based on the control points, the weight values assigned to each control point, and the node vector. The coordinates of a series of densely sampled points on the curve are calculated through the parameter equation of the smooth curve, which are used to represent the fitted smooth curve.

[0096] In some embodiments, a variance value of a sequence of first derivatives of the fitted smooth curve is calculated, the first derivatives describe the tangent direction or the rate of change of the curve at each parameter point, the calculated variance value is denoted as the direction change intensity. An average value of a sequence of curvatures of the smooth curve is calculated, the curvatures describe the bending degree of the curve at a point, the calculated average value is denoted as the path bending degree. In order to synthesize a local trajectory complexity index with consistent dimension, the direction change intensity and the path bending degree are respectively normalized. The calculation formula of the local trajectory complexity index is:

[0097]

[0098] wherein: represents the calculated direction change intensity, represents the calculated path bending degree, is a preset reference direction change intensity constant value, is a preset reference path bending degree constant value, and are preset weight coefficients of the normalized direction change intensity component and the path bending degree component, respectively. Optionally, the sum of the weight coefficients and is 1, refer to Table 1.

[0099] Table 1: Calculation parameter table of local trajectory complexity index

[0100]

[0101] In some embodiments, a first timestamp corresponding to a starting moment of measurement and a second timestamp corresponding to an ending moment of measurement are obtained, a difference between the second timestamp and the first timestamp is calculated to obtain a total time consumption. An efficiency calculation model including a sequence deviation factor, a local trajectory complexity index, and a total time consumption is established. The sequence deviation factor is input into a first calculation channel of the efficiency calculation model to obtain a sequence penalty component. The local trajectory complexity index is input into a second calculation channel of the efficiency calculation model to obtain a trajectory regularity component. The total time consumption is input into a third calculation channel of the efficiency calculation model to obtain a time efficiency component. The sequence penalty component, the trajectory regularity component, and the time efficiency component are synthesized to output an original operation efficiency value.

[0102] ​In specific implementation, a first timestamp corresponding to the measurement start time and a second timestamp corresponding to the measurement end time are obtained, the first timestamp and the second timestamp are values with unified time reference obtained from the time service of the measurement system. The numerical difference between the second timestamp and the first timestamp is calculated, and the difference is the total time consumption from the start of the measurement operation to the end of the measurement operation. An efficiency calculation model including the sequence deviation factor, the local trajectory complexity index and the total time consumption is established, and the efficiency calculation model defines the mathematical relationship from the input parameters to the output original operation efficiency value. In specific implementation, the efficiency calculation model defines a mathematical framework, which takes the sequence deviation factor, the local trajectory complexity index and the total time consumption as input parameters, and converts these parameters into intermediate components through a preset processing channel, and finally synthesizes the original operation efficiency value. The efficiency calculation model includes three independent calculation channels: the first calculation channel is used to process the sequence deviation factor, and generates a sequence penalty component by multiplying the sequence deviation factor by a preset penalty coefficient; the second calculation channel is used to process the local trajectory complexity index, and generates a trajectory regularization component by multiplying the local trajectory complexity index by a preset regularization coefficient; the third calculation channel is used to process the total time consumption, and generates a time efficiency component by performing mathematical transformation such as inverse operation on the total time consumption and multiplying by a scaling factor. The design of these calculation channels ensures that each input parameter can be independently quantified and adjusted to reflect its influence on operation efficiency in different dimensions. Then, the model outputs the original operation efficiency value by performing weighted summation operation on the sequence penalty component, the trajectory regularization component and the time efficiency component, wherein the weight coefficients are preset based on the relative importance of each component, and the dimension consistency and reasonableness of the synthesis process are ensured.

[0103] In some embodiments, the sequence deviation factor is input into the first calculation channel of the efficiency calculation model, and the first calculation channel processes the sequence deviation factor to obtain a sequence penalty component. It can be understood that the processing of the first calculation channel can be multiplying the sequence deviation factor by a preset penalty coefficient to obtain the sequence penalty component. The local trajectory complexity index is input into the second calculation channel of the efficiency calculation model, and the second calculation channel processes the local trajectory complexity index to obtain a trajectory regularization component. It can be understood that the processing of the second calculation channel can be multiplying the local trajectory complexity index by a preset regularization coefficient to obtain the trajectory regularization component. The total time consumption is input into the third calculation channel of the efficiency calculation model, and the third calculation channel processes the total time consumption to obtain a time efficiency component, which can be obtained by performing mathematical transformation on the total time consumption. In order to ensure the dimension consistency of the synthesis operation, the sequence penalty component, the trajectory regularization component and the time efficiency component are standardized. The calculation formula of the original operation efficiency value is:

[0104] ​

[0105] in: This represents the sequential penalty component output from the first computation channel. This represents the trajectory normalization component output from the second calculation channel. Indicated by total time The time efficiency component is the input and the output from the third computation channel. It is a preset reference value for the sequential penalty component. It is a preset reference value for trajectory normalization components. It is a preset reference value for the time efficiency component. , , These are the preset composite weight coefficients for the standardized sequential penalty component, trajectory regularization component, and time efficiency component.

[0106] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0107] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A machine vision-based size measurement and scoring method, characterized in that, include: A multi-step measurement framework for the object under test is established, and a continuous spatial position stream and a corresponding timestamp stream are captured synchronously during the measurement process. The continuous spatial position stream is analyzed to identify pause events, and each pause event is associated with a standard measurement stage in the multi-step measurement framework to generate an associated step sequence. Based on the associated sequence of steps, the logical order of the standard measurement stage is checked to see if it is followed. When a logical order misalignment is found, a sequence deviation event is marked and a corresponding sequence deviation factor is generated. In the continuous spatial position flow, it is dynamically detected whether an image region that conforms to the preset morphological characteristics appears. When the image region is detected, the moment when the image region that conforms to the preset morphological characteristics is detected is recorded as the feature capture moment. The set of position points within the time window before and after the feature capture moment is extracted and the set of position points is recorded as the effective measurement point set. Calculate the local trajectory complexity index based on the effective set of measurement points; The total time taken from the start to the end of the measurement is obtained, and the original operational efficiency value is calculated by combining the order deviation factor and the local trajectory complexity index. The system introduces a pre-stored reference dimension value for the object being measured, normalizes and fuses the original operational performance value with the reference dimension value, and finally outputs a comprehensive measurement score.

2. The machine vision-based size measurement and scoring method according to claim 1, characterized in that, The step of identifying pause events by parsing the continuous spatial location stream includes: Set a minimum duration threshold and a maximum displacement tolerance threshold; In the continuous spatial location stream, an analysis window of a fixed duration is slid in chronological order; Calculate the standard deviation of the spatial coordinates of all points within each analysis window; When the standard deviation of the spatial coordinates is less than the maximum displacement tolerance threshold, and the duration of this state exceeds the minimum duration threshold, it is determined that a pause event has occurred within the time period covered by the analysis window. Extract the average spatial coordinates of all locations within the time period, and use them as the representative location of the pause event.

3. The machine vision-based size measurement and scoring method according to claim 1, characterized in that, The step of associating each pause event with a standard measurement phase in the multi-step measurement framework includes: Obtain the theoretical spatial range of each standard measurement stage in the multi-step measurement framework; Calculate the Euclidean distance between the representative location of the pause event and the center point of each theoretical spatial region. The standard measurement stage corresponding to the theoretical spatial region with the smallest Euclidean distance is selected as the candidate correlation stage; Verify whether the representative location of the pause event is within the theoretical spatial region of the candidate association stage; If it falls within the range, the current pause event is successfully bound to the candidate associated stage, generating a step binding record; All step binding records are compiled in chronological order to form the associated step sequence.

4. The machine vision-based size measurement and scoring method according to claim 3, characterized in that, The step of checking whether the logical order of the standard measurement phases is followed includes: From the associated sequence of steps, extract the stage identifier of the standard measurement stage to which each step is bound; The extracted stage identifiers are arranged in order to generate an actual execution sequence; The actual execution sequence is compared bit by bit with the preset logical order in the multi-step measurement framework; When a stage identifier is found in the actual execution sequence and its position in the preset logical order appears before the position corresponding to the previous stage identifier, it is determined that a sequence misalignment has occurred. The total number of sequence misalignments that occur throughout the entire actual execution sequence is counted, and the total number of misalignments is converted into the sequence deviation factor according to a preset mapping relationship.

5. The machine vision-based size measurement and scoring method according to claim 1, characterized in that, The step of dynamically detecting whether an image region conforming to preset morphological characteristics appears includes: Real-time acquisition of video frames from the measurement perspective captured by the image sensor on the measurement equipment; Binarization segmentation is applied to each video frame to obtain the corresponding binary image; In the binarized image, all connected pixel regions are found using a region growing algorithm; Calculate the geometric feature descriptor for each connected pixel region, the geometric feature descriptor including area, perimeter, and rectangularity; The geometric feature descriptor of each connected pixel region is matched with the feature descriptor in the pre-stored feature template library based on similarity. When the matching similarity of a connected pixel region exceeds a preset similarity threshold, an image region that conforms to the preset morphological features is detected.

6. The machine vision-based size measurement and scoring method according to claim 1, characterized in that, The step of calculating the local trajectory complexity index based on the effective set of measurement points includes: Extract the spatial coordinates of each location point from the set of valid measurement points; Fit the spatial coordinates of all locations into a smooth curve; Calculate the variance of the first derivative sequence of the smooth curve, and denote the variance as the intensity of the direction change; Calculate the average value of the curvature sequence of the smooth curve, and record the average value as the path curvature. The local trajectory complexity index is obtained by weighted summing of the intensity of the directional change and the curvature of the path.

7. The machine vision-based size measurement and scoring method according to claim 1, characterized in that, The step of calculating the original operating performance value includes: Obtain the first timestamp corresponding to the start time of the measurement and the second timestamp corresponding to the end time of the measurement; Calculate the difference between the second timestamp and the first timestamp to obtain the total time elapsed; Establish an efficiency calculation model that includes sequence deviation factor, local trajectory complexity index and total energy consumption; The order deviation factor is input into the first calculation channel of the performance calculation model to obtain an order penalty component; The local trajectory complexity index is input into the second calculation channel of the performance calculation model to obtain a trajectory regularization component. The total time consumption is input into the third calculation channel of the performance calculation model to obtain a time efficiency component; The sequential penalty component, trajectory normalization component, and time efficiency component are combined and processed to output the original operation efficiency value.

8. The machine vision-based size measurement and scoring method according to claim 5, characterized in that, The step of applying binarization segmentation to each video frame includes: Select a global threshold for a specific color space channel in the video frame; Compare the component value of each pixel in the video frame in that color space channel with the global threshold; Pixels with component values ​​greater than the global threshold are set as foreground pixels, and pixels with component values ​​less than or equal to the global threshold are set as background pixels. Perform a morphological opening operation on the image after setting to remove small noise points in the foreground; The image after morphological opening is then subjected to morphological closing to fill the small holes in the foreground, resulting in the final binarized image.

9. The machine vision-based size measurement and scoring method according to claim 6, characterized in that, The step of fitting the spatial coordinates of all location points into a smooth curve includes: The non-uniform rational B-spline curve fitting method is adopted, and the spatial coordinates of all position points in the effective measurement point set are used as control points. Assign a weight value to each control point and determine a node vector; The parametric equation of the smooth curve is generated based on the control points, the weight values, and the node vectors. The coordinates of a series of densely sampled points on the curve are calculated using the parametric equations to characterize the fitted smooth curve.

10. A machine vision-based size measurement and scoring system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the machine vision-based size measurement and scoring method according to any one of claims 1 to 9.

Images