A welding piece production remote monitoring platform and method based on an internet of things

Abstract

This invention discloses an IoT-based remote monitoring platform and method for welded component production, relating to the field of industrial automation monitoring technology. The IoT-based remote monitoring platform and method for welded component production includes the following steps: S1: Acquiring and preprocessing image frame sequences, extracting target positions and dwell frequencies, and generating structured spatial trajectory data; S2: Constructing a standard task template and locating key areas, determining the area hit response status based on coverage; S3: Performing coordinate alignment analysis between the identified target spatial position and the standard task template, identifying coverage status and offset features; S4: Integrating coverage range, distribution balance, and key area status to comprehensively determine task completion status. This solves the problem that existing platforms cannot determine whether welding tasks are completed correctly based on the distribution of the operating area.

Description

Technical Field

[0001] This invention relates to the field of industrial automation monitoring technology, specifically to a remote monitoring platform and method for welding production based on the Internet of Things. Background Technology

[0002] With the continued application of monitoring, control, and data acquisition systems in the manufacturing industry, welding operations are gradually incorporating IoT-based data acquisition and control architectures. These architectures leverage edge computing nodes and instrument data acquisition devices to achieve real-time perception and remote monitoring of the work process. Control systems centered on parameter acquisition and task execution status monitoring have become a crucial component of welding production line information management, providing a technological foundation for the processing and feedback of multi-source heterogeneous data.

[0003] For example, the invention patent with announcement number CN110895874B discloses an RS485 interface sensor acquisition module and its acquisition method, belonging to the field of industrial communication engineering and applied in the structural safety and health monitoring industry. It includes a communication module, a multi-serial port microcontroller module, an isolated RS485 driver module, an interface protection module, a power management module, an isolated power supply module, a current sampling and monitoring module, a data storage module, a power distribution module, and an RS485 bus driver. It solves the problems of current RS485 network devices in the industry, such as the inability to remotely monitor operating status, the inability to upgrade online, and the inability to self-diagnose. It realizes the intelligence of RS485 network devices, improves the reliability and convenience of RS485 field system integration, and enhances the ability to achieve low maintenance costs and rapid fault location.

[0004] For example, invention patent CN114937349B discloses a toll lane status monitoring device, method, and medium. The device includes: an instruction receiving module for receiving a first control instruction from a control platform; an instruction analysis module for analyzing the first control instruction to obtain the issuing device and instruction content corresponding to the first control instruction; a heartbeat acquisition module for obtaining the device heartbeat of the issuing device corresponding to the first control instruction; a heartbeat analysis module for determining the device status of the issuing device based on the device heartbeat; the device status includes at least a normal status and an abnormal status; a status feedback module for uploading the device number of the issuing device in an abnormal status to the control platform based on the device status of the issuing device, so that the control platform can send a second control instruction to the lane indicator and information board corresponding to the issuing device; and an instruction issuing module for transmitting the first control instruction and the second control instruction corresponding to the issuing device to the issuing device in a normal status.

[0005] Current welding monitoring methods still largely rely on changes in process parameters such as current and voltage for judgment, lacking a spatial-level identification mechanism to determine whether critical work areas are actually covered. In welding tasks with complex structures and multiple workstations, omissions, offsets, and skipped areas are prone to occur. Traditional acquisition methods are insufficient to reflect the distribution of welding behavior in the image space, making it difficult to meet the needs for refined supervision of work standardization.

[0006] To address the above issues, there is an urgent need for a remote monitoring platform and method for welding production based on the Internet of Things. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a remote monitoring platform and method for welding production based on the Internet of Things, which solves the problem that existing technologies cannot determine whether welding tasks have been completed in a standardized manner based on the distribution of the operating area.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a remote monitoring method for welding production based on the Internet of Things, comprising the following steps: S1: acquiring and preprocessing image frame sequences, extracting target positions and dwell frequencies, and generating structured spatial trajectory data; S2: using the spatial trajectory data as a matching basis, constructing a standard task template and locating key areas, determining the area hit response status based on coverage, and triggering frequency adjustment, trajectory completion, area comparison, and status update; S3: using the area status update result as input, performing coordinate alignment analysis between the identified target spatial position and the standard task template, identifying coverage status and offset characteristics, and linking the offset monitoring, boundary re-collection, and early warning response mechanisms; S4: combining coverage status and offset characteristics as normative judgment criteria, integrating coverage range, distribution balance, and key area status, completing the comprehensive judgment of task completion status, and executing qualified marking, re-inspection assignment, abnormal re-collection, and coverage correction instructions.

[0011] Further, the specific steps for acquiring and preprocessing the image frame sequence, extracting the target position and dwell frequency, and generating structured spatial trajectory data are as follows: Acquire spatial trajectory data during the welding process. The spatial trajectory data includes: image frames, coordinates of the upper left corner of the recognition box, height of the recognition box, and width of the recognition box; capture image frames from the on-site video at set time intervals to form an image frame sequence; obtain the position, category, and confidence level of the recognition box of the operation target in each frame using a target recognition algorithm; extract the coordinates of the upper left corner of the recognition box and calculate the coordinates of the center point of the recognition target based on its height and width; construct the trajectory sequence of the operation target in the image space; perform blur, repetition, and illumination anomaly filtering on the image frames; unify the image scale and coordinate units; filter low-confidence recognition results and correct inter-frame positional abrupt changes; archive the processed target center point sequence and dwell frequency to form structured spatial trajectory data that can be used for spatial coverage analysis; and perform standardization and normalization processing on the spatial trajectory data.

[0012] Furthermore, the specific steps for constructing a standard task template and locating key areas based on spatial trajectory data are as follows: According to the welding task's operational specifications and structural drawings, mark all key welding areas that need to be covered and clarify their positional boundaries in the image space; set high-weight labels for key points such as connection points, structural corners, and edge junctions to establish regional priority levels; construct a standard task template from the marked areas according to category, coordinates, and weight information, unify the image coordinate scale to adapt to the spatial trajectory data, and assign a unique code to each marked area; support calling multiple versions of templates based on the weldment model and task type to meet the matching needs of different operational scenarios.

[0013] Further, the specific steps for determining the region hit response status based on coverage are as follows: Obtain the image frame sequence and the coordinates of the target center point; combine the constructed standard task template, read the region numbers marked as needing coverage in the template, and count the number of key task regions; read the category label of each region in the template, and automatically assign region weight coefficients according to the structure type mapping rules; perform spatial point statistics on the target center point in the image frame, and match it with the positional relationship of each region in the template, obtaining the number of pixels where the center point falls into the region as the region hit area; analyze the coordinate contour range of each marked region in the template, and calculate its enclosed pixel area as the region template area; count the number of times the target center point appears in each region in the image frame sequence to obtain the region operation frequency; take the number of key task regions as the total number of regions, and process each region separately: take the region weight coefficient, multiply it by the ratio between the region hit area and the region template area, and then multiply the product by the natural logarithm of the region operation frequency to obtain the weighted coverage result of the region; sum the weighted coverage results of all regions, and finally divide this accumulated value by the number of key task regions to obtain the region hit response value.

[0014] Further, the specific steps for trigger frequency adjustment, trajectory completion, region comparison, and status update are as follows: Real-time comparison of region hit response values ​​and response thresholds, where the response thresholds include a first-level response threshold and a second-level response threshold; when the region hit response value is greater than or equal to the first-level response threshold, increase the image acquisition frequency to enhance spatial data density; compare the center point of the identified target with the key area of ​​the template to identify uncovered solder joints; send an alarm to the main control interface to de-stabilize the target; when the region hit response value is greater than or equal to the second-level response threshold but less than the first-level response threshold, extend the image recording time to complete the coverage trajectory; add the target to the pending confirmation queue and periodically compare the key area reach rate; activate the spatial coverage recognition module to monitor the contact between the coverage boundary and high-weight solder joints; when the region hit response value is less than or equal to the second-level response threshold, save the current operation data as a stable sample; reduce the recognition frequency while maintaining periodic trajectory acquisition; pause spatial coverage recognition and only record the hit rate and time distribution information.

[0015] Furthermore, the specific steps for analyzing the coordinate alignment between the spatial location of the identified target and the standard task template, using the region state update result as input, and identifying the coverage state and offset features, are as follows: Obtain the image frame sequence and the center point of the identified target; count the number of consecutively captured image frames during task execution as the total number of image frames; combine the positional annotations of each key coverage area in the standard task template, and calculate the Euclidean distance between the center point of the identified target and the centroid of the template structure in each frame to obtain the operation offset distance; based on the spatial distribution of all center points in the image frame sequence, perform principal direction vector clustering analysis to extract the hit counts in the concentrated distribution directions. The number of operations is calculated as the frequency of the main direction; the cumulative number of hits of the target center point within the task area in all image frames is counted as the total number of operations; the total number of image frames is used as the total number of region traversals, and the following operations are performed on each image frame: the operation offset distance is squared and summed, the sum of the squared offset distances of all frames is divided by the total number of image frames, and the square root of the quotient is taken to obtain the average offset distance; the average offset distance is multiplied by the direction correction term in parentheses, which consists of a constant plus the direction enhancement coefficient multiplied by the ratio of the frequency of the main direction to the total number of operations; the final product is the operation offset intensity value.

[0016] Furthermore, the specific steps of the linkage execution offset monitoring, boundary re-import, and early warning response mechanism are as follows: real-time comparison of the operation offset intensity value and the offset threshold; when the operation offset intensity value is greater than or equal to the offset threshold, increase the image acquisition frequency, re-import the current area boundary image, compare the operation range with the maximum deviation range set by the template, if it exceeds the limit, trigger an offset early warning and record the abnormal information; when the operation offset intensity value is less than the offset threshold, write the current operation position into the stable coverage record for use as a spatial comparison reference, pause the offset direction identification and boundary expansion analysis, and if it remains stable continuously, mark it as a high-quality coverage period and archive it.

[0017] Furthermore, the specific steps for comprehensively judging the task completion status by combining coverage status and offset features as normative judgment criteria, integrating coverage range, distribution balance, and key area status are as follows: Acquire image frames and target center points; combine the obtained region hit response values ​​to extract the spatial distribution characteristics of the target in each key area during task execution; statistically analyze the hit frequency of the target center points in high-weight regions of all image frames and calculate their arithmetic mean as the key area density mean; based on the number of target hits in each region, calculate the standard deviation of the hit frequency of all task regions to obtain the spatial distribution standard deviation; by retrospectively analyzing the standard deviation of the hit frequency of each region in historical normative task samples and taking their average, obtain the standard distribution benchmark value; multiply the region hit response value by the key area density mean as the numerator of the formula; calculate the difference between the spatial distribution standard deviation and the standard distribution benchmark value, multiply by the distribution sensitivity coefficient, take the opposite of the product as the exponent, substitute it into the natural exponential function, add one to the exponent value, and use it as the denominator of the formula; divide the numerator by the denominator to finally obtain the spatial coverage comprehensive value.

[0018] Further, the specific steps for executing the qualified marking, re-inspection assignment, abnormal supplementary sampling, and coverage correction instructions are as follows: Real-time comparison of the spatial coverage comprehensive value with the confirmation threshold, which includes a primary confirmation threshold and a secondary confirmation threshold; when the spatial coverage comprehensive value is greater than or equal to the primary confirmation threshold, the task is marked as qualified, subsequent verification is terminated, and the evaluation report is generated directly; when the spatial coverage comprehensive value is greater than the secondary confirmation threshold but less than the primary confirmation threshold, the spatial coverage identification module is invoked for regional verification, and a supplementary sampling process is assigned through the regulatory interface to supplement key area images; when the spatial coverage comprehensive value is less than or equal to the secondary confirmation threshold, the task is marked as unqualified, the distribution of the target center point is compared with the template area, missing locations are identified, a redo prompt is issued, and coverage reinforcement suggestions are generated.

[0019] The second aspect of this invention provides an IoT-based remote monitoring platform for weldment production, comprising: an operation data acquisition module, a spatial offset calculation module, a spatial coverage identification module, and a structural specification determination module. The operation data acquisition module is used to acquire and preprocess image frame sequences, extract target positions and dwell frequencies, and generate structured spatial trajectory data. The spatial offset calculation module is used to construct a standard task template and locate key areas based on the spatial trajectory data, determine the area hit response status based on coverage, and trigger frequency adjustment, trajectory completion, area comparison, and status update. The spatial coverage identification module is used to perform coordinate alignment analysis between the identified target spatial position and the standard task template, identify coverage status and offset characteristics, and link offset monitoring, boundary re-sampling, and early warning response mechanisms, using the area status update results as input. The structural specification determination module is used to combine coverage status and offset characteristics as the basis for specification judgment, integrate coverage range, distribution balance, and key area status to complete a comprehensive judgment of task completion, and execute qualified marking, re-inspection assignment, abnormal re-sampling, and coverage correction instructions.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) This invention extracts the center point coordinates and dwell frequency of the target by acquiring and preprocessing the image frame sequence, constructs spatial trajectory data, and unifies the image scale, coordinate unit and data format to realize the computable expression of the operation process, which solves the problem that the existing monitoring data structure is loose and difficult to support subsequent accurate judgment.

[0023] (2) This invention analyzes the key areas marked on the structural drawings, sets weight labels and unique codes, constructs a standard task template, compares it with the spatial trajectory collected in real time, and performs identification and update according to the area hit response value, providing a stable foundation for spatial offset analysis and task standardization judgment, and improving the accuracy of area-level coverage identification.

[0024] (3) The present invention uses two indicators, offset intensity value and hit response value, to evaluate the operational stability, sets up a multi-level threshold trigger frequency adjustment, boundary supplementation and jump area early warning mechanism, dynamically controls the image acquisition and behavior monitoring mode, realizes the immediate response to abnormal behavior, and enhances the software’s ability in real-time feedback and intelligent intervention.

[0025] (4) This invention calculates the spatial coverage comprehensive value, integrates the hit density of operation behavior in key areas, the standard deviation of spatial distribution and historical standard benchmarks, realizes multi-dimensional data-driven qualification judgment and graded processing, supports task marking, review assignment and supplementary sampling suggestion generation, and improves the software's comprehensive decision support capability in welding quality management scenarios.

[0026] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0027] Figure 1 This is a flowchart of a remote monitoring method for welded parts production based on the Internet of Things according to the present invention;

[0028] Figure 2 This is a structural diagram of a remote monitoring platform for welding production based on the Internet of Things according to the present invention;

[0029] Figure 3 This is a line graph showing the operational offset intensity values ​​of the present invention.

[0030] Figure 4 This is a standard task template diagram for the present invention;

[0031] Figure 5 This is a thermal diagram showing the actual operation of the present invention. Detailed Implementation

[0032] 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.

[0033] Please see Figures 1-5 This invention provides a technical solution: a remote monitoring method for welding production based on the Internet of Things, comprising the following steps: S1: acquiring and preprocessing image frame sequences, extracting target positions and dwell frequencies, and generating structured spatial trajectory data; S2: using spatial trajectory data as a matching basis, constructing a standard task template and locating key areas, determining the area hit response status based on coverage, and triggering frequency adjustment, trajectory completion, area comparison, and status update; S3: using the area status update result as input, performing coordinate alignment analysis between the identified target spatial position and the standard task template, identifying coverage status and offset characteristics, and linking the offset monitoring, boundary re-collection, and early warning response mechanisms; S4: combining coverage status and offset characteristics as normative judgment criteria, integrating coverage range, distribution balance, and key area status to complete the comprehensive judgment of task completion, and executing qualified marking, re-inspection assignment, abnormal re-collection, and coverage correction instructions.

[0034] Specifically, the process involves acquiring and preprocessing image frame sequences, extracting target positions and dwell frequencies, and generating structured spatial trajectory data. The specific steps are as follows: First, spatial trajectory data during the welding process is acquired. This data includes image frames, the coordinates of the top-left corner of the recognition box, the height of the recognition box, and the width of the recognition box, ensuring that the spatial position information of the object being operated on can be accurately located in each frame. Second, image frames are extracted from the live video at set time intervals to form an image frame sequence. Each frame is then identified using a target recognition algorithm, extracting the recognition box position, category, confidence level, and other attributes. Third, the geometric parameters of the recognition box are combined to extract the coordinates of the top-left corner and calculate its center point coordinates, constructing the motion trajectory of the target in the image space frame by frame. To improve data reliability, image frames are filtered for blur, repetition, and lighting anomalies, eliminating unstable image segments. Simultaneously, image scale and coordinate units are standardized to filter low-confidence recognition results, and abrupt position changes between frames are corrected to eliminate errors caused by abnormal jumps. Finally, the processed center point sequence and corresponding dwell frequency are archived to generate spatial trajectory data with clear structure and time sequence. Further standardization and normalization processing is then performed to meet the algorithm input requirements for subsequent region matching and coverage analysis.

[0035] In this implementation scheme, the precise extraction and standardized representation of the spatial behavior of the welding process are achieved through the acquisition and preprocessing of image frame sequences. The center point position of the target is calculated using bounding box parameters, and the dwell frequency is statistically analyzed based on inter-frame time information, constructing complete and clearly structured spatial trajectory data. Through image quality screening, coordinate unification, confidence filtering, and position correction, the stability and usability of the data are effectively improved. The resulting structured spatial trajectory data not only retains the core positional features of the operation but also possesses a unified scale and standard format, providing a reliable data foundation for subsequent area coverage matching, offset analysis, and standardization judgment.

[0036] Specifically, based on spatial trajectory data, a standard task template is constructed and key areas are located. The specific steps are as follows: According to the welding task's operational specifications and structural drawings, all key welding areas that need to be covered are marked, clearly defining the specific location boundaries of each area in the image space to ensure clear spatial guidance for the task objective. High-weight labels are assigned to connection points, structural corners and edge junctions, and other operation-sensitive points to reflect their importance in the overall structure, establishing a region priority level based on structural characteristics. The category attributes, spatial coordinates, and weight information of the marked areas are integrated to construct a standard task template, and the image coordinate scale is unified to ensure coordinate consistency with the subsequently extracted spatial trajectory data. A unique code is assigned to each marked area to facilitate region identification and result indexing during execution. This template supports switching and calling multiple versions of structural configurations according to weldment type and task type, meeting diverse operational needs and achieving dynamic adaptation between the task template and the actual scene.

[0037] In this implementation plan, a foundation for matching spatial trajectory data with operational requirements was established by constructing a standard task template and accurately locating key welding areas. Through structural drawing annotations and weight settings, the position and importance of each area in the image space were clarified, ensuring the distinguishability and priority of key components in subsequent analysis. The template construction process achieved uniformity in area attributes, coordinate information, and image scale, while introducing a unique coding mechanism to enhance the stability and traceability of area identification. The final standard task template not only possesses structural integrity and standardized format but also supports flexible switching between multiple versions, providing a clear and adjustable spatial reference for subsequent coverage judgment and offset analysis.

[0038] Specifically, the region hit response status is determined based on coverage, and the specific steps are as follows: First, acquire the image frame sequence and the coordinates of the target center point, and extract the spatial distribution information during the operation. Second, using the established standard task template, read the region numbers marked as needing coverage and count the number of key regions that should be fully covered in the task. Third, further read the category label corresponding to each region and automatically assign weight coefficients to different regions according to the structure type mapping rules, reflecting their relative importance in the task structure. Fourth, in the image frame, perform spatial landing point statistics on the target center point and match it with the boundary positions of each region in the template to obtain the number of pixels where the center point falls into each region, which is used as the region hit area. Fifth, simultaneously analyze the coordinate contours of each region in the template, calculate the number of pixels within its enclosed area, and obtain the region template area. Sixth, count the cumulative occurrences of the target center point in each region in the image frame sequence to obtain the region operation frequency. The number of critical task regions is used as the total number of regions. Each region is processed separately: its weight coefficient is taken, multiplied by the ratio between the region's hit area and the template area, and then the product is multiplied by the natural logarithm of the operation frequency to obtain the weighted coverage result. Finally, the weighted coverage results of all regions are summed and divided by the number of critical task regions to calculate the region hit response value, which is used to measure the overall coverage quality.

[0039] The specific calculation method for the area hit response value is as follows:

[0040]

[0041] In the formula, Indicates the region hit response value. Indicates the number of critical areas in the mission. Indicates the regional weighting coefficient. Indicates the area hit. Indicates the area of ​​the template region. Indicates the frequency of operations within a region.

[0042] In this implementation plan, the spatial distribution of targets within key areas is statistically identified. Combined with regional attributes, area proportions, and operation frequency, the actual coverage of the standard task template by welding operations is quantified. Introducing regional weighting coefficients differentiates the impact of different structural locations on coverage assessment, avoiding the problem of treating key and auxiliary areas equally. By constructing a weighted coverage result and summarizing it into a comprehensive spatial coverage value, a unified expression of the overall coverage quality of the welding task is achieved, providing a clear and quantifiable basis for subsequent anomaly identification, standardization assessment, and control response.

[0043] Specifically, the triggering frequency adjustment, trajectory completion, area comparison, and status update involve the following steps: Real-time comparison of the area hit response value with response thresholds, including a primary response threshold and a secondary response threshold, used to dynamically classify coverage status levels. When the area hit response value is greater than or equal to the primary response threshold, the current coverage anomaly is determined to be high, and the image acquisition frequency is immediately increased to obtain higher-density spatial behavior data. The target center point is compared one by one with the key areas in the template to identify solder joints with coverage gaps, and an alarm message is sent to the main control interface to de-stabilize the target for subsequent intervention. When the area hit response value is between the secondary and primary response thresholds, it indicates a potential coverage deviation that is not yet severe. The image recording time is extended to complete the continuity of the coverage trajectory, and the target is added to the confirmation queue. Automatic comparison of key area reach rates is performed periodically, and the spatial coverage recognition module is invoked to monitor boundary coverage and the reach status of high-weight solder joints. When the area hit response value is less than or equal to the secondary response threshold, the current coverage status is determined to be stable. The current operation data is archived as a stable sample, the image recognition frequency is reduced, the periodic acquisition of the trajectory is maintained, and spatial coverage recognition is paused. Only the hit rate and time distribution records are retained to reduce redundant calculations.

[0044] In this implementation plan, by setting multi-level response thresholds, the acquisition frequency, trajectory completion, and recognition strategies are dynamically adjusted based on the regional hit response value, enabling the classification and differentiated processing of coverage status. According to changes in coverage, the image acquisition density, trajectory recording duration, and comparison behavior are intelligently switched to promptly identify potential missed solder joints and insufficient boundary coverage. Operational behaviors under unstable conditions are incorporated into the alarm and supplementary acquisition mechanisms. For stable coverage conditions, data simplification and resource optimization are achieved through frequency reduction and sample archiving, thereby improving overall monitoring efficiency and response sensitivity.

[0045] Specifically, using the region state update results as input, coordinate alignment analysis is performed on the spatial location of the identified target and the standard task template to identify coverage status and offset features. The specific steps are as follows: Image frame sequences and the center point of the identified target are acquired, and spatial location data throughout the task is extracted. The number of consecutively captured image frames during task execution is counted as the total number of image frames, used for normalization processing in subsequent offset calculations. Combining the location annotations of key coverage areas in the standard task template, the Euclidean distance between the center point of the identified target and the centroid of the template structure in each frame is calculated to obtain the operation offset distance, reflecting the spatial deviation degree of the behavior in each frame. Based on the overall distribution of all center points in the image frame sequence, principal direction vector clustering analysis is performed to identify the clustering of hit points in concentrated directions and extract the principal direction frequency. Finally, the cumulative number of hits of the identified target within the task region is counted as the total operation frequency, used to characterize the overall behavior density. The total number of image frames is used as the total number of region traversals. For each image frame, the following operations are performed: the operation offset distance is squared and summed. After the sum of the squared offset distances of all frames is completed, it is divided by the total number of image frames, and the square root is taken to obtain the average offset distance. Then, the average offset distance is multiplied by the direction correction term, which consists of a constant plus the ratio of the direction enhancement coefficient to the frequency of the main direction and the total frequency of operations. The direction enhancement coefficient is obtained by analyzing the relationship between the distribution density of the operation target in the main direction of the task area and the spatial balance curve, and the value range is 0-2. The final product is the operation offset intensity value, which is used to comprehensively reflect the concentration and spatial stability of the welding behavior.

[0046] The specific calculation method for the operational offset intensity value is as follows:

[0047]

[0048] In the formula, Indicates the operational offset intensity value. This indicates the total number of image frames. Indicates the operation offset distance. Indicates the frequency of the main direction. This indicates the total number of operations. This represents the directional enhancement coefficient.

[0049] Table 1 shows the data table of operational offset intensity values ​​provided in the embodiments of this application. In this embodiment, the total number of image frames for measurement point 1 is set to 10, the operation offset distance is set to [4.93, 3.93, 4.36, 5.28, 5.00, 2.93, 4.34, 3.53, 3.57, 3.94], the main direction frequency is set to 35, and the total operation frequency is set to 90; the total number of image frames for measurement point 2 is set to 12, the operation offset distance is set to [4.83, 6.05, 5.40, 4.80, 5.11, 5.00, 6.09, 4.50, 4.98, 3.89, 2.30, 5.30], the main direction frequency is set to 45, and the total operation frequency is set to 100; the total number of image frames for measurement point 3 is set to 11, and the operation offset distance is set to [5.72, 4.15, 7.09, 3.46, 4.92, ... [4.69, 6.37, 6.31, 5.03, 5.25, 4.01], the main direction frequency is set to 55, and the total number of operations is set to 110; the total number of image frames for measurement point 4 is set to 9, the operation offset distance is set to [2.99, 4.60, 5.10, 6.17, 6.14, 4.57, 4.65, 3.91, 3.54], the main direction frequency is set to 58, and the total number of operations is set to 100; the total number of image frames for measurement point 5 is set to 13, the operation offset distance is set to [3.85, 8.13, 5.25, 5.34, 4.38, 6.76, 3.96, 5.60, 4.80, 6.30, 5.25, 4.47, 5.82], the main direction frequency is set to 62, and the total number of operations is set to 130.

[0050] Table 1 Data table of operational offset intensity values

[0051]

[0052] like Figure 3The figure shows a line graph of the operation offset intensity values ​​provided in this embodiment of the application. According to the data in the image and table, the set offset threshold is 1.95. The operation offset intensity values ​​of the five measuring points fluctuate between 1.65 and 2.13, showing an overall trend of first increasing and then decreasing. The operation offset intensity value of measuring point 4 is 2.13, the highest among the five groups, exceeding the threshold, indicating that there is a significant concentrated offset in this area, possibly deviating from the key area set in the standard task template. The offset intensity value of measuring point 3 is 2.07, also higher than the threshold, suggesting a certain degree of uneven coverage and behavioral divergence risk. The offset intensity value of measuring point 5 is 1.96, slightly higher than the threshold boundary, requiring further confirmation of its offset direction and boundary contact. In contrast, the offset intensity values ​​of measuring points 2 and 1 are 1.81 and 1.65 respectively, both below the threshold, indicating relatively stable operation behavior, concentrated spatial distribution, and relatively standardized coverage. This figure is used to assist in judging whether there are abnormal offsets in the operation, providing a reference for subsequent boundary re-sampling, offset monitoring, and task correction.

[0053] In this implementation plan, spatial offset characteristics during operation are quantified by analyzing the coordinate alignment between the identified target location and the standard task template. By calculating the offset distance between the identification center point and the template's centroid, and combining this with directional clustering and operation frequency, an offset intensity value is constructed. This effectively reflects whether there are issues such as concentrated offsets, deviations from key areas, and behavioral divergence. This indicator introduces a directional correction term while considering the average spatial offset, making the analysis results more directionally sensitive and adaptable to distribution, providing clear and calculable quantitative behavioral basis for subsequent offset monitoring and anomaly detection.

[0054] Specifically, the mechanism for coordinated execution of offset monitoring, boundary re-import, and early warning response involves the following steps: Real-time comparison of the operation offset intensity value with the offset threshold is used to determine if there is significant spatial deviation in the current operation. When the operation offset intensity value is greater than or equal to the offset threshold, the current behavior is identified as having a concentrated offset trend. The image acquisition frequency is immediately increased, and boundary images of the current area are re-imported to obtain more complete spatial trajectory data. Simultaneously, the operation range is compared with the maximum deviation range set in the standard task template. If it exceeds the limit, the offset early warning mechanism is triggered, and relevant abnormal information is recorded for subsequent analysis. When the operation offset intensity value is less than the offset threshold, the operation behavior is deemed spatially stable. The current operation position is written into the stable coverage record as a reference for future spatial comparisons, and offset direction identification and boundary expansion analysis are paused to reduce the load. If this state remains unchanged over multiple consecutive time periods, it is marked as a high-quality coverage period and archived to provide reliable sample support for compliance assessment.

[0055] In this implementation scheme, dynamic identification and classification of welding behavior offset states are achieved through real-time comparison of the operation offset intensity value and the offset threshold. Image acquisition density is adjusted according to the offset degree, boundary area images are automatically supplemented, and offset warnings are triggered, effectively addressing coverage divergence and location boundary violations. For operations in a stable state, archiving and optimized identification tasks are performed to form high-quality coverage samples, improving the accuracy and efficiency of the monitoring process and providing a continuous and reliable data foundation for subsequent standardized judgments and behavior assessments.

[0056] Specifically, combining coverage status and offset features as normative judgment criteria, and integrating coverage range, distribution balance, and key area status, a comprehensive judgment of task completion is achieved. The specific steps are as follows: First, acquire image frames and target center points. Second, combine the obtained region hit response values ​​to extract the spatial distribution characteristics of the identified targets in each key area during task execution, constructing a multi-dimensional data foundation describing coverage quality. Third, statistically analyze the hit frequency of target center points in high-weight regions across all image frames and calculate their arithmetic mean as the key area density mean, used to measure the operational activity level of key areas. Fourth, based on the number of target hits in each region, calculate the standard deviation of the hit frequency across all task regions to obtain the spatial distribution standard deviation, reflecting the balance of operational distribution. Fifth, by retrospectively analyzing the standard deviation of the hit frequency of each region in historical normative task samples, take their average as the standard distribution benchmark value, providing a reference for judging the degree of deviation. The numerator of the formula is formed by multiplying the regional hit response value by the mean density of the key regions. A distribution sensitivity coefficient, ranging from 0 to 1.5, is obtained by fitting the response relationship between spatial distribution fluctuations and task completion levels in historical tasks. The difference between the current standard deviation and the baseline value is then calculated and multiplied by the distribution sensitivity coefficient. The negative of the product is used as the exponent, substituted into the natural exponential function, and incremented by one as the denominator. Finally, the numerator is divided by the denominator to obtain the comprehensive spatial coverage value, which comprehensively reflects the degree of standardization in task completion.

[0057] The specific calculation method for the comprehensive value of spatial coverage is as follows:

[0058]

[0059] In the formula, Indicates the overall value of spatial coverage. Indicates the region hit response value. This represents the mean density of the key area. Indicates the standard deviation of the spatial distribution. Represents the baseline value of the standard distribution. This represents the distribution sensitivity coefficient.

[0060] This implementation plan establishes a comprehensive assessment mechanism for welding task completion by integrating regional coverage status and spatial offset characteristics. Based on an examination of hit intensity, it introduces the mean density of key areas, the standard deviation of spatial distribution, and historical distribution benchmark values ​​to achieve a coordinated evaluation of coverage, operational balance, and structural response. The calculation of the comprehensive spatial coverage value not only quantifies the overall compliance level but also effectively distinguishes between stable coverage and distribution anomalies, supporting various application scenarios such as qualification determination, review screening, and subsequent feedback control.

[0061] Specifically, the process involves executing commands for qualification marking, re-inspection assignment, anomaly re-collection, and coverage correction. The specific steps are as follows: Real-time comparison of the spatial coverage composite value with preset confirmation thresholds, including primary and secondary confirmation thresholds to differentiate between different coverage quality levels. When the spatial coverage composite value is greater than or equal to the primary confirmation threshold, the task coverage is deemed compliant, directly marked as qualified, all subsequent verification operations are terminated, and the evaluation report generation process begins, achieving closed-loop task management. When the spatial coverage composite value is between the secondary and primary confirmation thresholds, a slight coverage deviation is considered, requiring further confirmation. The spatial coverage identification module is invoked to verify key areas one by one, and a re-collection process is assigned through the regulatory interface to collect images of missing locations for supplementary comparison. When the spatial coverage composite value is less than or equal to the secondary confirmation threshold, coverage is deemed severely insufficient, and the task is marked as unqualified. Simultaneously, the correspondence between the target center point and the template area is compared and identified to identify unreached locations, a redo prompt is issued, and structured coverage reinforcement suggestions are generated to guide subsequent corrective operations.

[0062] In this implementation plan, by comparing the comprehensive spatial coverage value with the confirmation threshold, the completion status of welding tasks is automatically graded and responded to. Tasks can be categorized into three types based on coverage level: qualified, pending review, and unqualified, with linked marking, re-image acquisition, and correction operations. Qualified tasks can be directly archived and an evaluation report generated; tasks pending review are further confirmed through area verification and image re-acquisition; and unqualified tasks output missing locations and re-welding suggestions, achieving closed-loop control from judgment to repair, thus improving the automation and accuracy of quality management.

[0063] like Figure 2The diagram shown is a structural schematic of a remote monitoring platform for welded parts production based on the Internet of Things (IoT) provided in this application embodiment. This IoT-based remote monitoring platform for welded parts production applies an IoT-based remote monitoring method, including: an operation data acquisition module, a spatial offset calculation module, a spatial coverage identification module, and a structural specification determination module. The operation data acquisition module is used to acquire and preprocess image frame sequences, extract the position coordinates and dwell frequency information of the identified target, clean abnormal frames and unify coordinate scales, and generate structured spatial trajectory data to provide basic input for subsequent spatial analysis. The spatial offset calculation module is used to construct a standard task template corresponding to the task requirements based on the spatial trajectory data, mark key welding areas, and set... The weighting level determines the region hit response status based on the trajectory landing point, triggering image acquisition frequency adjustment, trajectory data completion, region matching comparison, and status result update; the spatial coverage recognition module takes the region status update result as input, performs coordinate alignment analysis on the distribution of the recognition target in the image space and the template structure, identifies the overall coverage status and local offset features, and links to execute offset behavior monitoring, boundary area re-imaging, and anomaly warning response; the structural specification judgment module comprehensively analyzes the coverage status and offset features, using this as the basis for judging welding specifications, integrates the overall coverage range, spatial distribution balance, and key area reach, outputs a comprehensive conclusion on task completion, and executes qualified marking, review task assignment, anomaly re-imaging call, and coverage correction instructions accordingly.

[0064] In this implementation plan, a modular design enables spatial behavior monitoring and compliance assessment throughout the entire welding process. The operational data acquisition module preprocesses image frames and extracts trajectories, providing foundational data for spatial analysis. The spatial offset calculation module locates key areas and determines hit status based on the trajectory and task template, driving dynamic data updates. The spatial coverage recognition module focuses on coverage integrity and offset feature identification, supporting real-time monitoring and response. The structural compliance assessment module integrates multi-dimensional indicators, outputs task completion results, and triggers feedback commands. This closed-loop data collaboration among modules ensures the accuracy of welding quality analysis, the timeliness of response, and the targeted nature of control.

[0065] like Figure 4The diagram shown is a standard task template provided in this application embodiment, used to mark key areas that need to be covered in a welding task and their spatial locations. Three marked areas are defined in the diagram: Area A (connection point, weight 3), Area B (corner position, weight 2), and Area C (boundary area, weight 1), with their boundaries clearly marked in the image space using rectangles. Each area is distinguished by a different color, and the weight values ​​reflect their importance in the task structure. Area A is a high-weight core area located in the left-middle of the image; Area B has medium weight and is located in the upper-middle part; and Area C has low weight and is located in the lower right corner. This template unifies the image coordinate scale and structural category encoding, providing a clear reference for subsequent spatial trajectory data matching, coverage judgment, and offset identification.

[0066] like Figure 5 The image shown is a heatmap of actual operations provided in this application embodiment, used to display the landing positions and dwell frequency distribution of the identified target in the image space. Combined with the standard task template image, it can be seen that regions A and B show consecutive hits, with the highest landing frequency reaching 2.00, indicating that the operation is densely covered in high-weight regions and exhibits repetitive behavior. Region C also shows obvious operation traces, indicating that low-weight regions are also reached. The remaining image space shows no obvious hotspots, and the operation distribution is highly concentrated. This image, combined with template comparison results, indicates that the current operation basically covers all key areas, but the distribution between regions is still uneven. It is recommended to strengthen trajectory stability control in boundary regions to prevent coverage shifts and structural skipping risks.

[0067] 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.

[0068] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A remote monitoring method for welding piece production based on Internet of Things, characterized in that, Includes the following steps: S1: Acquire and preprocess image frame sequences, extract target location and dwell frequency, and generate structured spatial trajectory data; S2: Based on spatial trajectory data, construct a standard task template and locate key areas. Determine the area hit response status based on coverage, and trigger frequency adjustment, trajectory completion, area comparison and status update. The specific steps for determining the area hit response status based on coverage are as follows: The process involves: acquiring image frame sequences and identifying the coordinates of the target center point; combining a pre-constructed standard task template, reading the region numbers marked as needing coverage, and counting the number of key task regions; reading the category label of each region in the template and automatically assigning region weight coefficients according to the structure type mapping rules; counting the spatial landing points of the target center point in the image frames and matching them with the positional relationships of each region in the template, obtaining the number of pixels in which the center point falls as the region hit area; analyzing the coordinate contour range of each marked region in the template and calculating its enclosed pixel area as the region template area; and counting the number of times the target center point appears in each region in the image frame sequence to obtain the region operation frequency. The number of critical task regions is taken as the total number of regions. Each region is processed separately: the region weight coefficient is taken, multiplied by the ratio between the region hit area and the region template area, and then the product is multiplied by the natural logarithm of the region operation frequency to obtain the weighted coverage result of the region; the weighted coverage results of all regions are added together, and finally this cumulative value is divided by the number of critical task regions to obtain the region hit response value. S3: Using the regional status update results as input, perform coordinate alignment analysis between the spatial location of the identified target and the standard task template, identify the coverage status and offset characteristics, and link the offset monitoring, boundary re-sampling and early warning response mechanisms. S4: Combining coverage status and offset characteristics as the basis for normative judgment, integrating coverage range, distribution balance and key area status, to complete the comprehensive judgment of task completion, and execute qualified marking, re-examination and assignment, abnormal supplementary collection and coverage correction instructions.

2. The welding piece production remote monitoring method based on the Internet of Things according to claim 1, characterized in that: The specific steps for acquiring and preprocessing the image frame sequence, extracting the target location and dwell frequency, and generating structured spatial trajectory data are as follows: Spatial trajectory data during the welding process is collected, including image frames, coordinates of the top left corner of the recognition box, and the height and width of the recognition box. Image frames are captured from the on-site video at set time intervals to form an image frame sequence. A target recognition algorithm is used to obtain the position, category, and confidence level of the recognition box for the target in each frame. The coordinates of the top left corner of the recognition box are extracted and combined with its height and width to calculate the coordinates of the target's center point, constructing a trajectory sequence of the target in the image space. Blur, repetition, and lighting anomaly filtering are performed on the image frames. Image scale and coordinate units are standardized, low-confidence recognition results are filtered, and abrupt changes in position between frames are corrected. The processed target center point sequence and dwell frequency are archived to form structured spatial trajectory data that can be used for spatial coverage analysis. The spatial trajectory data is then standardized and normalized.

3. The welding piece production remote monitoring method based on the Internet of Things according to claim 1, characterized in that: The specific steps for constructing a standard task template and locating key areas based on spatial trajectory data are as follows: Based on the welding task's work specifications and structural drawings, mark all critical welding areas that need to be covered and clarify their positional boundaries in the image space; set high-weight labels for key points such as connection parts, structural corners, and edge junctions to establish regional priority levels; construct standard task templates for the marked areas according to category, coordinates, and weight information, unify the image coordinate scale to adapt to spatial trajectory data, and assign a unique code to each marked area; It supports calling multiple versions of templates based on the welding part model and task type to meet the matching needs of different work scenarios.

4. The welding piece production remote monitoring method based on the Internet of Things according to claim 1, characterized in that: The specific steps for trigger frequency adjustment, trajectory completion, region comparison, and status update are as follows: Real-time comparison of the region hit response value with the response threshold, wherein the response threshold includes a first-level response threshold and a second-level response threshold; When the area hit response value is greater than or equal to the first-level response threshold, the image acquisition frequency is increased to enhance the spatial data density; the center point of the identified target is compared with the key area of ​​the template to identify the uncovered solder joints; an alarm is sent to the main control interface to release the target from the stable state. When the area hit response value is greater than or equal to the secondary response threshold but less than the primary response threshold, extend the image recording time and supplement the coverage trajectory; add the target to the queue to be confirmed, and compare the reach rate of key areas periodically; Activate the spatial coverage identification module to monitor the contact between the coverage boundary and high-weight solder joints; When the region hit response value is less than or equal to the secondary response threshold, save the current operation data as a stable sample; reduce the recognition frequency and maintain trajectory periodic acquisition. Spatial coverage identification is paused; only hit rate and temporal distribution information are recorded.

5. The welding piece production remote monitoring method based on the Internet of Things according to claim 1, characterized in that: The specific steps for using the region state update results as input to perform coordinate alignment analysis between the spatial location of the identified target and the standard task template, and to identify the coverage status and offset features, are as follows: The process involves acquiring image frame sequences and identifying target center points; counting the number of consecutively captured image frames during task execution as the total number of image frames; combining the location annotations of each key coverage area in the standard task template, calculating the Euclidean distance between the target center point in each frame and the centroid of the template structure to obtain the operation offset distance; based on the spatial distribution of all center points in the image frame sequence, performing principal direction vector clustering analysis to extract the number of hits in the concentrated distribution direction as the principal direction frequency; and counting the cumulative number of hits of the target center point within the task area in all image frames as the total operation frequency. Using the total number of image frames as the total number of region traversals, perform the following operations on each image frame: Square the operation offset distance and sum them up. After summing the squared offset distances of all frames, divide the sum by the total number of image frames, and then take the square root of the quotient to obtain the average offset distance. Multiply the average offset distance by the direction correction term in parentheses. The direction correction term consists of a constant plus the direction enhancement coefficient multiplied by the ratio of the main direction frequency to the total operation frequency. The final product is the operation offset intensity value.

6. The welding piece production remote monitoring method based on the Internet of Things according to claim 1, characterized in that: The specific steps of the linkage execution offset monitoring, boundary supplementary sampling, and early warning response mechanism are as follows: The system compares the operation offset intensity value with the offset threshold in real time. When the operation offset intensity value is greater than or equal to the offset threshold, the image acquisition frequency is increased, and the current area boundary image is acquired. The operation range is compared with the maximum deviation range set by the template. If the deviation is exceeded, an offset warning is triggered and abnormal information is recorded. When the operation offset intensity value is less than the offset threshold, the current operation position is written into the stable coverage record for use as a spatial comparison reference. The offset direction identification and boundary expansion analysis are paused. If the stable state is maintained continuously, it is marked as a high-quality coverage period and archived.

7. The welding piece production remote monitoring method based on the Internet of Things according to claim 1, characterized in that: The specific steps for comprehensively determining the task completion status by combining coverage status and offset features as normative judgment criteria, and integrating coverage range, distribution balance, and key area status, are as follows: Acquire image frames and identify target center points, and combine them with the obtained region hit response values ​​to extract the spatial distribution features of the identified targets in each key region during the task execution process; The hit frequency of the target center point in the high-weight region in all image frames is counted and its arithmetic mean is calculated as the density mean of the key region. Based on the number of target hits in each region, the standard deviation of the hit frequency of all task regions is calculated to obtain the spatial distribution standard deviation. By retrospectively analyzing the standard deviation of the hit frequency of each region in the historical standard task samples, the average value is taken to obtain the standard distribution benchmark value. Multiply the regional hit response value by the mean density of the critical region as the numerator of the formula; calculate the difference between the standard deviation of the spatial distribution and the baseline value of the standard distribution, multiply it by the distribution sensitivity coefficient, take the opposite of the product as the exponent, substitute it into the natural exponential function, add one to the exponent value as the denominator of the formula; divide the numerator by the denominator to finally obtain the comprehensive spatial coverage value.

8. The welding piece production remote monitoring method based on the Internet of Things according to claim 1, characterized in that: The specific steps for executing the qualified marking, re-checking and assignment, abnormal re-sampling and coverage correction instructions are as follows: The system compares the overall spatial coverage value with the confirmation threshold in real time. The confirmation threshold includes a primary confirmation threshold and a secondary confirmation threshold. When the overall spatial coverage value is greater than or equal to the primary confirmation threshold, the task is marked as qualified, subsequent verification is terminated, and the system directly proceeds to the evaluation report generation stage. When the overall spatial coverage value is greater than the secondary confirmation threshold but less than the primary confirmation threshold, the spatial coverage identification module is invoked to perform regional verification, and a supplementary acquisition process is allocated through the regulatory interface to supplement images of key areas. When the overall spatial coverage value is less than or equal to the secondary confirmation threshold, the task is marked as unqualified, the distribution of the target center point is compared with the template area, missing locations are identified, a redo prompt is issued, and coverage reinforcement suggestions are generated.

9. An IoT-based remote monitoring platform for welded component production, employing the IoT-based remote monitoring method for welded component production as described in any one of claims 1-8, comprising: The system comprises an operational data acquisition module, a spatial offset calculation module, a spatial coverage identification module, and a structural specification determination module, characterized in that: The operation data acquisition module is used to acquire and preprocess image frame sequences, extract target positions and dwell frequencies, and generate structured spatial trajectory data. The spatial offset calculation module is used to construct a standard task template and locate key areas based on spatial trajectory data, determine the area hit response status based on coverage, and trigger frequency adjustment, trajectory completion, area comparison and status update. The spatial coverage identification module is used to take the regional status update result as input, perform coordinate alignment analysis between the spatial location of the identification target and the standard task template, identify the coverage status and offset characteristics, and link and execute offset monitoring, boundary re-sampling and early warning response mechanisms. The structural specification determination module is used to combine coverage status and offset characteristics as the basis for standardization judgment, integrate coverage range, distribution balance and key area status to complete the comprehensive judgment of task completion, and execute qualified marking, re-examination assignment, abnormal supplementary sampling and coverage correction instructions.

Images