Method and system for high-precision detection of profile size of auxiliary washing machine glass observation window

Abstract

The present application relates to the field of washing machine glass observation window profile size detection, disclose a kind of auxiliary washing machine glass observation window's profile size high-precision detection method and system, method includes: obtaining step is using equal spatial interval sampling, obtain the point cloud data set of spatial uniform distribution;In preprocessing step, local surface feature analysis is carried out to point cloud data, edge candidate point is marked and spatial connectivity analysis is carried out, and after isolated point is removed, edge point cloud data is obtained;In fitting step, standard geometric model is iteratively adjusted based on geometric constraint condition, until overall cumulative deviation variation is less than convergence threshold;In size generation step, profile size parameter is extracted.The system includes signal acquisition module, data preprocessing module, geometric fitting module and parameter extraction module.The present application eliminates the influence of motion speed fluctuation on sampling uniformity, effectively eliminates surface defect interference, and realizes high-precision, high-efficiency automatic detection of washing machine glass observation window profile size.

Description

Technical Field

[0001] This invention relates to the field of washing machine glass viewing window contour dimension detection technology, specifically to a high-precision detection method and system for the contour dimension of a washing machine glass viewing window. Background Technology

[0002] The viewing window of a washing machine is a key component, and its dimensional accuracy directly affects the sealing performance and appearance quality of the entire machine. Currently, the inspection of the viewing window's dimensions primarily relies on manual inspection tools or random sampling using a contact coordinate measuring machine (CMM). Manual inspection depends on the operator's experience, resulting in low efficiency and high subjectivity, making it difficult to guarantee consistent and accurate results. While contact CMMs offer high accuracy, their slow speed, potential scratches from probe contact with the glass surface, and high equipment cost limit their suitability for laboratory sampling and cannot meet the demands of full-scale online inspection on production lines. Furthermore, traditional laser scanning methods are prone to measurement noise and edge recognition errors due to the light-transmitting properties and complex surface curvature of glass, making it difficult to achieve the required dimensional accuracy for assembly. Summary of the Invention

[0003] The purpose of this invention is to provide a high-precision detection method and system for the outline dimensions of the glass observation window of an auxiliary washing machine, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a high-precision detection method for the contour dimensions of an auxiliary washing machine glass viewing window, applied to a detection system including a point laser displacement sensor and a drive mechanism, comprising the following steps:

[0005] Acquisition steps: During the process of the drive mechanism driving the point laser displacement sensor to move along the preset scanning path, the trigger sampling method is set to equal spatial interval sampling according to the movement trajectory of the drive mechanism to eliminate the influence of motion speed fluctuation on the uniformity of sampling point distribution. The sensor is triggered to collect data according to the preset spatial sampling interval, and the collected analog signal is converted from analog to digital to obtain a set of point cloud data that is spatially uniformly distributed on the preset scanning path.

[0006] Preprocessing steps: Perform local surface feature analysis on the point cloud dataset to obtain the curvature value at each sampling point. Sampling points with curvature values ​​exceeding a preset curvature threshold are marked as edge candidate points. Perform spatial connectivity analysis on the edge candidate points to remove isolated edge candidate points that cannot form a connected region with the main edge. The remaining edge candidate points with continuous distribution characteristics are determined as edge point cloud data.

[0007] Fitting steps: Select multiple feature points from the edge point cloud data, iteratively adjust the pose of the standard geometric model based on the preset geometric constraints, redetermine the overall cumulative deviation between the standard geometric model and multiple feature points after each adjustment, stop the adjustment when the change in the overall cumulative deviation after two adjacent adjustments is less than the preset convergence threshold, and take the current standard geometric model as the final fitting result.

[0008] Size generation steps: Based on the standard geometric model, extract the contour size parameters of the washing machine glass viewing window. The contour size parameters include at least one of diameter, roundness and radius of curvature.

[0009] As a preferred embodiment of the present invention, the preprocessing step of performing noise filtering on the point cloud data set further includes:

[0010] Statistical analysis is performed on the point cloud dataset to determine the average distance between each sampling point and its neighboring points. Sampling points whose average distance exceeds the overall distribution threshold are identified as outliers and removed to obtain the first point cloud dataset.

[0011] The first point cloud data set is smoothed and filtered to eliminate random fluctuations caused by sensor electrical noise or surface micro-irregularities.

[0012] As a preferred embodiment of the present invention, the preset geometric constraints include the spatial relationship between the central axis of the standard geometric model and the design rotation central axis of the washing machine glass viewing window; the fitting step further includes:

[0013] Determine an initial pose for the standard geometric model such that the central axis of the standard geometric model is located in the vicinity of the design rotation center axis;

[0014] During the iterative adjustment process, the degree of deviation between the central axis of the standard geometric model and the design rotation center axis is continuously monitored;

[0015] When the change in the overall cumulative deviation is less than the preset convergence threshold and the degree of deviation tends to stabilize, the adjustment stops, and the current standard geometric model is taken as the final fitting result.

[0016] A high-precision detection system for the outline dimensions of an auxiliary washing machine glass viewing window, used to implement the method described in any one of the above, comprising:

[0017] The signal acquisition module is used to set the trigger sampling mode to equal spatial interval sampling according to the movement trajectory of the drive mechanism to eliminate the influence of motion speed fluctuation on the uniformity of sampling point distribution during the process of the drive mechanism driving the point laser displacement sensor to move along the preset scanning path. It also triggers the sensor to collect data according to the preset spatial sampling interval, performs analog-to-digital conversion on the collected analog signals, and obtains a set of point cloud data that is spatially uniformly distributed on the preset scanning path.

[0018] The data preprocessing module is used to perform local surface feature analysis on the point cloud dataset, obtain the curvature value at each sampling point, mark the sampling points with curvature values ​​exceeding the preset curvature threshold as edge candidate points, and perform spatial connectivity analysis on the edge candidate points to remove isolated edge candidate points that cannot form a connected region with the main edge, and determine the retained edge candidate points with continuous distribution characteristics as edge point cloud data.

[0019] The geometric fitting module is used to select multiple feature points from edge point cloud data and iteratively adjust the pose of the standard geometric model based on preset geometric constraints. After each adjustment, the overall cumulative deviation between the standard geometric model and multiple feature points is redefined. When the change in the overall cumulative deviation after two adjacent adjustments is less than the preset convergence threshold, the adjustment is stopped and the current standard geometric model is taken as the final fitting result.

[0020] The parameter extraction module is used to extract the contour dimension parameters of the washing machine glass viewing window based on a standard geometric model. The contour dimension parameters include at least one of diameter, roundness, and radius of curvature.

[0021] As a preferred embodiment of the present invention, the data preprocessing module includes a noise filtering unit and an edge recognition unit;

[0022] The noise filtering unit is used to perform statistical analysis on the point cloud data set, determine the average distance between each sampling point and its neighboring points, identify sampling points whose average distance exceeds the overall distribution threshold range as outliers and remove them, and obtain the first point cloud data set. The first point cloud data set is then smoothed and filtered to eliminate random fluctuations.

[0023] The edge recognition unit is used to perform local surface feature analysis on the point cloud data set after noise filtering, analyze and obtain the curvature value at each sampling point, mark the sampling points with curvature values ​​exceeding the preset curvature threshold as edge candidate points, and perform spatial connectivity analysis on the edge candidate points. Isolated edge candidate points that cannot form a connected region with the main edge are removed as interference points, and edge candidate points with continuous distribution characteristics are retained and determined as edge point cloud data.

[0024] As a preferred embodiment of the present invention, the geometric fitting module is further configured to:

[0025] Select multiple feature points from edge point cloud data;

[0026] Determine an initial pose for the standard geometric model so that the model initially corresponds to multiple feature points in space;

[0027] Based on preset geometric constraints, the pose of the standard geometric model is iteratively adjusted, and the overall cumulative deviation between the standard geometric model and multiple feature points is redetermined after each adjustment.

[0028] Compare the overall cumulative deviation obtained after each adjustment with the overall cumulative deviation after the previous adjustment;

[0029] When the change in the overall cumulative deviation obtained after two consecutive adjustments is less than the preset convergence threshold, the iterative process is determined to have converged, the adjustment is stopped, and the current standard geometric model is taken as the final fitting result.

[0030] Compared with the prior art, the beneficial effects of the present invention are:

[0031] 1. By adopting an equal spatial interval sampling method in the acquisition step, this invention eliminates the influence of the speed fluctuation of the driving mechanism on the uniformity of the sampling point distribution, so that the point cloud data set is evenly distributed on the preset scanning path, laying a data foundation for subsequent high-precision edge recognition and model fitting, and effectively avoiding measurement errors caused by uneven distribution of sampling points.

[0032] 2. In the preprocessing step, this invention performs noise filtering and spatial connectivity analysis on the point cloud data set. Outliers are removed through statistical analysis and random fluctuations are eliminated by smoothing filtering. Then, the edge point cloud data belonging to the edge feature area of ​​the washing machine glass observation window is accurately identified through curvature analysis and spatial connectivity analysis. This effectively eliminates isolated interference points caused by glass surface defects and ensures the purity and continuity of the edge point cloud data.

[0033] 3. In the fitting step, this invention uses an iterative adjustment method based on preset geometric constraints to construct a standard geometric model. By continuously monitoring the change in the overall cumulative deviation and the degree of deviation between the model's central axis and the designed rotation center axis, the adjustment stops when the convergence condition is met. This achieves high-precision fitting of the geometric features of the washing machine's glass observation window contour and significantly improves the accuracy of contour dimension parameter extraction.

[0034] 4. This invention, through a systematic modular setup, organically combines the signal acquisition module, data preprocessing module, geometric fitting module, and parameter extraction module, realizing fully automated detection from raw signal acquisition to final dimensional parameter output, effectively improving detection efficiency and meeting the actual needs of online full inspection on production lines.

[0035] 5. This invention is applicable to the contour detection of washing machine glass viewing windows of different sizes, specifications and curvature characteristics. It has strong versatility and adaptability. By optimizing the sampling interval and threshold parameters, the detection accuracy for specific products can be optimized, which has significant economic benefits and application prospects.

[0036] In summary, this method and system are suitable for high-precision detection of the outline dimensions of the glass viewing window of a washing machine. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the overall process of a high-precision detection method for the outline dimensions of an auxiliary washing machine glass observation window according to the present invention. Detailed Implementation

[0038] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0039] Example 1

[0040] Acquisition steps: During the process of the drive mechanism driving the point laser displacement sensor to move along the preset scanning path, the trigger sampling method is set to equal spatial interval sampling according to the movement trajectory of the drive mechanism to eliminate the influence of motion speed fluctuation on the uniformity of sampling point distribution. The sensor is triggered to collect data according to the preset spatial sampling interval, and the collected analog signal is converted from analog to digital to obtain a set of point cloud data that is spatially uniformly distributed on the preset scanning path.

[0041] Preprocessing steps: Perform local surface feature analysis on the point cloud dataset to obtain the curvature value at each sampling point. Sampling points with curvature values ​​exceeding a preset curvature threshold are marked as edge candidate points. Perform spatial connectivity analysis on the edge candidate points to remove isolated edge candidate points that cannot form a connected region with the main edge. The remaining edge candidate points with continuous distribution characteristics are determined as edge point cloud data.

[0042] Fitting steps: Select multiple feature points from the edge point cloud data, iteratively adjust the pose of the standard geometric model based on the preset geometric constraints, redetermine the overall cumulative deviation between the standard geometric model and multiple feature points after each adjustment, stop the adjustment when the change in the overall cumulative deviation after two adjacent adjustments is less than the preset convergence threshold, and take the current standard geometric model as the final fitting result.

[0043] Size generation steps: Based on the standard geometric model, extract the contour size parameters of the washing machine glass viewing window. The contour size parameters include at least one of diameter, roundness and radius of curvature.

[0044] Furthermore, the preprocessing step involves noise filtering of the point cloud dataset, which further includes:

[0045] Statistical analysis is performed on the point cloud dataset to determine the average distance between each sampling point and its neighboring points. Sampling points whose average distance exceeds the overall distribution threshold are identified as outliers and removed to obtain the first point cloud dataset.

[0046] The first point cloud data set is smoothed and filtered to eliminate random fluctuations caused by sensor electrical noise or surface micro-irregularities.

[0047] Furthermore, the preset geometric constraints include the spatial relationship between the central axis of the standard geometric model and the design rotation central axis of the washing machine glass viewing window; the fitting step further includes:

[0048] Determine an initial pose for the standard geometric model such that the central axis of the standard geometric model is located in the vicinity of the design rotation center axis;

[0049] During the iterative adjustment process, the degree of deviation between the central axis of the standard geometric model and the design rotation center axis is continuously monitored;

[0050] When the change in the overall cumulative deviation is less than the preset convergence threshold and the degree of deviation tends to stabilize, the adjustment stops, and the current standard geometric model is taken as the final fitting result.

[0051] A high-precision detection system for the outline dimensions of an auxiliary washing machine glass viewing window, comprising the method for achieving any of the above, including:

[0052] The signal acquisition module is used to set the trigger sampling mode to equal spatial interval sampling according to the movement trajectory of the drive mechanism to eliminate the influence of motion speed fluctuation on the uniformity of sampling point distribution during the process of the drive mechanism driving the point laser displacement sensor to move along the preset scanning path. It also triggers the sensor to collect data according to the preset spatial sampling interval, performs analog-to-digital conversion on the collected analog signals, and obtains a set of point cloud data that is spatially uniformly distributed on the preset scanning path.

[0053] The data preprocessing module is used to perform local surface feature analysis on the point cloud dataset, obtain the curvature value at each sampling point, mark the sampling points with curvature values ​​exceeding the preset curvature threshold as edge candidate points, and perform spatial connectivity analysis on the edge candidate points to remove isolated edge candidate points that cannot form a connected region with the main edge, and determine the retained edge candidate points with continuous distribution characteristics as edge point cloud data.

[0054] The geometric fitting module is used to select multiple feature points from edge point cloud data and iteratively adjust the pose of the standard geometric model based on preset geometric constraints. After each adjustment, the overall cumulative deviation between the standard geometric model and multiple feature points is redefined. When the change in the overall cumulative deviation after two adjacent adjustments is less than the preset convergence threshold, the adjustment is stopped and the current standard geometric model is taken as the final fitting result.

[0055] The parameter extraction module is used to extract the contour dimension parameters of the washing machine glass viewing window based on a standard geometric model. The contour dimension parameters include at least one of diameter, roundness, and radius of curvature.

[0056] Furthermore, the data preprocessing module includes a noise filtering unit and an edge recognition unit;

[0057] The noise filtering unit is used to perform statistical analysis on the point cloud dataset, determine the average distance between each sampling point and its neighboring points, identify sampling points whose average distance exceeds the overall distribution threshold as outliers and remove them, and obtain the first point cloud dataset. The first point cloud dataset is then smoothed and filtered to eliminate random fluctuations.

[0058] The edge recognition unit is used to perform local surface feature analysis on the point cloud data set after noise filtering. It analyzes and obtains the curvature value at each sampling point, marks the sampling points with curvature values ​​exceeding the preset curvature threshold as edge candidate points, and performs spatial connectivity analysis on the edge candidate points. Isolated edge candidate points that cannot form a connected region with the main edge are removed as interference points, and edge candidate points with continuous distribution characteristics are retained and identified as edge point cloud data.

[0059] Furthermore, the geometry fitting module is further used for:

[0060] Select multiple feature points from edge point cloud data;

[0061] Determine an initial pose for the standard geometric model so that the model initially corresponds to multiple feature points in space;

[0062] Based on preset geometric constraints, the pose of the standard geometric model is iteratively adjusted, and the overall cumulative deviation between the standard geometric model and multiple feature points is redetermined after each adjustment.

[0063] Compare the overall cumulative deviation obtained after each adjustment with the overall cumulative deviation after the previous adjustment;

[0064] When the change in the overall cumulative deviation obtained after two consecutive adjustments is less than the preset convergence threshold, the iterative process is determined to have converged, the adjustment is stopped, and the current standard geometric model is taken as the final fitting result.

[0065] Example 2

[0066] This embodiment has the same basic steps as Embodiment 1. The difference lies in the optimization of the specific settings of the preset scanning path in the acquisition step and the parameter selection of noise filtering in the preprocessing step, so as to adapt to the contour detection of the washing machine glass observation window with different curvature characteristics.

[0067] Taking the glass observation window of a 6.5kg capacity drum washing machine produced by a washing machine manufacturer in East China as an example, the observation window is made of high borosilicate glass material. Its edge contour includes an arc-shaped area with a curvature radius of about 120mm and a straight area. The detection system uses Keyence LK-G5000 series point laser displacement sensor, in conjunction with a linear module drive mechanism.

[0068] In the acquisition step, based on the contour features of the observation window, the preset scanning path is a composite path: first, scanning is performed along the arc edge of the observation window at equal spatial intervals of 0.05mm, and then scanning is performed along the straight edge at equal spatial intervals of 0.08mm. The purpose of using equal spatial interval sampling is to eliminate the uneven distribution of sampling points that may occur during the speed change of the drive mechanism. The moving speed of the drive mechanism is set to 10mm / s. The signal acquisition module triggers the sensor to collect data according to the preset spatial sampling interval, and a total of 158,432 sampling points of point cloud data are obtained.

[0069] In the preprocessing step, when performing noise filtering on the point cloud dataset, the statistical analysis parameters are set as follows: the number of neighboring points k is 15, and the overall distribution threshold range is set to ±2.5 times the standard deviation of the average distance. Sampling points whose average distance exceeds this range are identified as outliers and removed. A total of 2347 noise points that significantly deviate from the main distribution are removed, resulting in the first point cloud dataset. Subsequently, a moving average filtering algorithm is used to smooth the first point cloud dataset, with the filtering window size set to 5 points to eliminate random fluctuations caused by sensor electrical noise or microscopic unevenness of the glass surface.

[0070] During the edge point cloud data identification process, the curvature threshold was set to 0.02 during local surface feature analysis, and the radius of the connected region was set to 0.1mm during spatial connectivity analysis. Isolated edge candidate points that could not form a connected region with the main edge were removed as interference points, and the edge point cloud data finally contained 21,568 valid edge points.

[0071] In the fitting step, 500 feature points are selected from the edge point cloud data to determine the initial pose of the circular standard geometric model, so that the model and the feature points initially correspond. The preset geometric constraints include the spatial position constraints between the central axis of the model and the rotation center axis of the washing machine glass observation window design. During the iterative adjustment process, the convergence threshold is set to 0.001mm. When the change in the overall cumulative deviation obtained after two adjacent adjustments is less than 0.001mm, the iterative process is judged to have converged, the adjustment is stopped, and the current standard geometric model is taken as the final fitting result.

[0072] In the dimension generation step, based on the fitted standard geometric model, the radius of curvature of the arc-shaped region of the washing machine glass observation window is extracted as 120.15 mm, and the straightness error of the straight region is 0.03 mm.

[0073] Example 3

[0074] This embodiment has the same basic steps as Embodiment 1. The difference lies in the optimization of the specific parameters for edge point cloud data identification in the preprocessing step and the application of geometric constraints in the fitting step, so as to be suitable for the contour detection of the glass observation window of a washing machine with a large radius of curvature.

[0075] Taking the glass observation window of a 9kg capacity drum washing machine produced by a home appliance company in South China as an example, the observation window is made of heat-shock resistant aluminosilicate glass material, and its edge contour is mainly approximately circular with a design diameter of 320mm. The detection system uses an Omron ZW-5000 series point laser displacement sensor, combined with a high-precision servo motor drive mechanism.

[0076] In the acquisition process, the preset scanning path is a closed path around the edge of the observation window, and scanning is performed at equal spatial intervals of 0.03mm. The driving mechanism's moving speed is set to 8mm / s, and the signal acquisition module triggers the sensor to acquire data according to the preset spatial sampling interval, obtaining a total of 33,493 point cloud data sampling points.

[0077] In the preprocessing step, when performing noise filtering on the point cloud dataset, the statistical analysis parameters were set as follows: the number of neighboring points k was set to 20, and the overall distribution threshold range was set to ±3.0 times the standard deviation of the average distance. Sampling points whose average distance exceeded this range were identified as outliers and removed. A total of 1876 noise points that significantly deviated from the main distribution were removed, resulting in the first point cloud dataset. Subsequently, a Gaussian filtering algorithm was used to smooth the first point cloud dataset, with the filtering parameter σ set to 0.02 mm to eliminate random fluctuations while preserving edge features.

[0078] During the edge point cloud data recognition process, the curvature threshold was set to 0.005 when analyzing local surface features. Since the edge curvature of the large curvature radius contour changes relatively gently, a lower curvature threshold helps to capture complete edge features. During spatial connectivity analysis, the radius of the connected region was set to 0.15 mm. Isolated edge candidate points that could not form a connected region with the main edge were removed as interference points. Finally, the edge point cloud data contained 8247 valid edge points.

[0079] In the fitting step, 300 feature points are selected from the edge point cloud data to determine the initial pose of the circular standard geometric model, so that the central axis of the model is located in the vicinity of the designed rotation center axis. In the preset geometric constraints, the designed rotation center axis is determined according to the mechanical installation reference of the observation window. During the iterative adjustment process, the deviation between the central axis of the standard geometric model and the designed rotation center axis is continuously monitored. The convergence threshold is set to 0.005mm. When the change in the overall cumulative deviation is less than 0.005mm and the deviation tends to stabilize, the adjustment is stopped, and the current standard geometric model is taken as the final fitting result.

[0080] In the size generation step, based on the fitted standard geometric model, the diameter of the washing machine glass observation window is extracted as 320.28 mm, and the roundness error is 0.12 mm.

[0081] Compare with Example 1

[0082] This comparative example uses the same washing machine glass observation window sample and testing equipment as Example 1. The difference is that equal spatial interval sampling is not used in the acquisition step, but equal time interval sampling is used instead.

[0083] Taking the glass observation window of a 6.5kg capacity drum washing machine produced by a washing machine manufacturer in East China as an example, the detection system also uses Keyence LK-G5000 series point laser displacement sensor and linear module drive mechanism. The moving speed of the drive mechanism is set to 10mm / s, but the trigger sampling method is sampling at equal time intervals, and the sampling frequency is set to 200Hz.

[0084] During the acquisition process, the speed fluctuations of the drive mechanism during startup, uniform motion, deceleration, and steering resulted in uneven spatial distribution of sampling points. During the acceleration phase, the interval between adjacent sampling points was small; during the deceleration phase, the interval between adjacent sampling points was large. A total of 15,000 point cloud data points were obtained, but the distribution density of the point cloud along the scanning path varied significantly.

[0085] In the preprocessing step, when performing noise filtering on the point cloud dataset, the same parameter settings as in Example 1 were used: the number of neighboring points k was set to 15, and the overall distribution threshold range was set to ±2.5 times the standard deviation of the average distance. However, due to the uneven distribution of the point cloud and the sparse point cloud density in some areas, the baseline value of the average distance fluctuated greatly during statistical analysis, making it difficult to accurately identify outliers. A total of 1856 suspected noise points were removed.

[0086] During edge point cloud data identification, the curvature threshold was set to 0.02. Due to the uneven distribution of point clouds, the reliability of curvature analysis results decreased in sparse point cloud areas, and some real edge points failed to be accurately identified due to insufficient local point cloud data. In dense point cloud areas, a large number of redundant points were generated. During spatial connectivity analysis, due to the large variation in point cloud intervals, the setting of a connectivity region determination radius of 0.1 mm was difficult to adapt to both sparse and dense areas, resulting in some continuous edges being misjudged as isolated points and removed.

[0087] In the fitting step, 500 feature points are selected from the edge point cloud data. Due to the presence of some missing and redundant features in the edge point cloud data, the selected feature points fail to uniformly cover the entire edge contour. During the iterative adjustment process, the convergence threshold is set to 0.001 mm. However, due to the deterioration of the input data quality, there is a deviation between the fitted standard geometric model and the real contour.

[0088] In the dimension generation step, based on the fitted standard geometric model, the radius of curvature of the arc-shaped region of the washing machine glass observation window is extracted as 119.32 mm, and the straightness error of the straight region is 0.11 mm.

[0089] Compare with Example 2

[0090] This comparative example uses the same washing machine glass observation window sample and testing equipment as Example 1. The difference is that spatial connectivity analysis was not performed in the preprocessing step, and edge point cloud data was identified only by relying on curvature threshold.

[0091] Taking the glass observation window of a 6.5kg capacity drum washing machine produced by a washing machine manufacturer in East China as an example, the detection system uses Keyence LK-G5000 series point laser displacement sensor and linear module drive mechanism. In the acquisition step, the same equal spatial interval sampling method as in Example 1 is adopted, with spatial sampling intervals of 0.05mm and 0.08mm, and a total of 158,432 sampling points are obtained in the point cloud data.

[0092] In the preprocessing step, when performing noise filtering on the point cloud data set, the same parameter settings as in Example 1 are used: the number of neighboring points k is set to 15, the overall distribution threshold range is set to ±2.5 times the standard deviation of the average distance, 2347 outliers are removed to obtain the first point cloud data set, and then the moving average filtering algorithm is used to smooth the first point cloud data set, with the filtering window size set to 5 points.

[0093] During the edge point cloud data identification process, the curvature threshold was set to 0.02 during local surface feature analysis. Sampling points with curvature values ​​exceeding 0.02 were directly marked as edge points without spatial connectivity analysis. Due to the presence of minor scratches, bubbles, and other defects on the surface of the washing machine glass viewing window, these defects also generate high curvature values, resulting in a large number of isolated interference points being incorrectly marked as edge points. A total of 28,936 edge candidate points were marked, including a large number of isolated interference points that did not belong to the main edge of the viewing window.

[0094] In the fitting step, 500 feature points are selected from the edge point cloud data containing a large number of interference points. Due to the presence of interference points, the selected feature points include some outliers that do not belong to the real edges. During the iterative adjustment process, these outliers have a significant negative impact on the fitting results, causing the standard geometric model to be distorted in order to adapt to the outliers.

[0095] In the dimension generation step, based on the fitted standard geometric model, the radius of curvature of the arc-shaped region of the washing machine glass observation window is extracted as 121.78 mm, and the straightness error of the straight region is 0.18 mm.

[0096] Experimental Data and Results Analysis

[0097] To verify the technical effect of the present invention, the experimental settings of Examples 1-3 and Comparative Examples 1-2 were used to detect the contour dimensions of 50 washing machine glass observation window samples produced in the same batch. The detection indicators included the radius of curvature of the arc area, the straightness error of the straight area, the diameter and roundness error of the circular contour. The measurement results of the high-precision coordinate measuring machine were used as the reference standard value. The measurement error (absolute value of the measured value minus the reference value) and detection repeatability (standard deviation of 10 repeated measurements of the same sample) of each detection method were calculated. The experimental results are shown in Table 1.

[0098] Table 1 Comparison of detection accuracy between various embodiments and control examples

[0099]

[0100] As can be seen from the experimental data in Table 1, Examples 1-3 of the present invention exhibit excellent detection accuracy in all detection indicators. Taking Example 1 as an example, the average error of the radius of curvature is only 0.08 mm, the maximum error does not exceed 0.15 mm, and the detection repeatability reaches 0.04 mm, indicating that the method has extremely high measurement accuracy and stability. Example 2 further improves the detection accuracy by optimizing the preset scanning path and noise filtering parameters, reducing the average error of the radius of curvature to 0.07 mm and maintaining the average error of straightness at an excellent level of 0.02 mm. When Example 3 is applied to the detection of circular contours with large radii of curvature, the average diameter error is 0.15 mm and the average roundness error is 0.08 mm, which fully meets the assembly accuracy requirements of the washing machine glass observation window.

[0101] In contrast, Comparative Example 1 used equal-time interval sampling but not equal-spatial-interval sampling. Due to the fluctuation of the driving mechanism's speed, the sampling points were unevenly distributed, resulting in a significant decrease in the quality of the point cloud data. Experimental data showed that the average error of the radius of curvature in Comparative Example 1 reached 0.95 mm, which was 11.8 times that of Example 1; the detection repeatability was 0.58 mm, which was 14.5 times that of Example 1. This result fully demonstrates the necessity and technical effectiveness of using equal-spatial-interval sampling in the acquisition step. Uneven distribution of sampling points directly affects the accuracy of subsequent edge recognition and model fitting, especially in arc-shaped areas with large curvature changes. Insufficient or excessive point cloud density will lead to distortion of curvature analysis results, which will ultimately be reflected in a significant increase in measurement error.

[0102] In contrast, Example 2 did not perform spatial connectivity analysis in the preprocessing step, relying solely on curvature thresholds to identify edge point cloud data. Experimental data showed that Example 2 had the worst detection accuracy, with an average curvature radius error as high as 1.86 mm and a maximum error of 3.52 mm. The detection repeatability was 0.92 mm. Further analysis revealed that the minute defects on the glass observation window surface generated a large number of isolated interference points with high curvature. These interference points were incorrectly marked as edge points, causing distortion of the standard geometric model during the fitting process. The influence of interference points was particularly significant in straight areas, with a straightness error reaching 0.16 mm, far exceeding the 0.02 mm in Example 1. This result fully validates the crucial role of spatial connectivity analysis: by eliminating isolated edge candidate points that cannot form a connected region with the main edge as interference points, the interference of surface defects on edge identification was effectively eliminated, ensuring the purity and continuity of the edge point cloud data and laying a solid foundation for subsequent high-precision fitting.

[0103] Further analysis of the comparative data between Example 2 and Example 1 reveals that using differentiated sampling intervals for contour regions with different curvature characteristics can further improve detection accuracy. In Example 2, a smaller sampling interval of 0.05 mm is used for arc-shaped regions with large curvature changes, while a larger sampling interval of 0.08 mm is used for straight regions. This ensures both the integrity of the key region's features and the detection efficiency. Experimental results show that the maximum error of the curvature radius in Example 2 is reduced from 0.15 mm in Example 1 to 0.12 mm, demonstrating a significant improvement.

[0104] The experimental results of Example 3 verify the applicability of the technical solution of the present invention to glass observation windows of different sizes. For a large observation window with a diameter of 320mm, a low curvature threshold of 0.005 and a relatively loose convergence threshold of 0.005mm were used to successfully achieve high-precision detection. The average diameter error was 0.15mm, and the relative error was only 0.047% relative to the design size of 320mm, achieving extremely high detection accuracy.

[0105] 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.

Claims

1. A high-precision detection method for the outline dimensions of an auxiliary washing machine glass viewing window, applied to a detection system including a point laser displacement sensor and a drive mechanism, characterized in that, Includes the following steps: Acquisition steps: During the process of the drive mechanism driving the point laser displacement sensor to move along the preset scanning path, the trigger sampling method is set to equal spatial interval sampling according to the movement trajectory of the drive mechanism to eliminate the influence of motion speed fluctuation on the uniformity of sampling point distribution. The sensor is triggered to collect data according to the preset spatial sampling interval, and the collected analog signal is converted from analog to digital to obtain a set of point cloud data that is spatially uniformly distributed on the preset scanning path. Preprocessing steps: Perform local surface feature analysis on the point cloud dataset to obtain the curvature value at each sampling point. Sampling points with curvature values ​​exceeding a preset curvature threshold are marked as edge candidate points. Perform spatial connectivity analysis on the edge candidate points to remove isolated edge candidate points that cannot form a connected region with the main edge. The remaining edge candidate points with continuous distribution characteristics are determined as edge point cloud data. Fitting steps: Select multiple feature points from the edge point cloud data, iteratively adjust the pose of the standard geometric model based on the preset geometric constraints, redetermine the overall cumulative deviation between the standard geometric model and multiple feature points after each adjustment, stop the adjustment when the change in the overall cumulative deviation after two adjacent adjustments is less than the preset convergence threshold, and take the current standard geometric model as the final fitting result. Size generation steps: Based on the standard geometric model, extract the contour size parameters of the washing machine glass viewing window. The contour size parameters include at least one of diameter, roundness and radius of curvature.

2. The method for high-precision detection of the outline dimensions of an auxiliary washing machine glass observation window according to claim 1, characterized in that, The preprocessing step, which involves noise filtering of the point cloud dataset, further includes: Statistical analysis is performed on the point cloud dataset to determine the average distance between each sampling point and its neighboring points. Sampling points whose average distance exceeds the overall distribution threshold are identified as outliers and removed to obtain the first point cloud dataset. The first point cloud data set is smoothed and filtered to eliminate random fluctuations caused by sensor electrical noise or surface micro-irregularities.

3. The method for high-precision detection of the outline dimensions of an auxiliary washing machine glass observation window according to claim 1, characterized in that, The preset geometric constraints include the spatial relationship between the central axis of the standard geometric model and the design rotation central axis of the washing machine glass viewing window; the fitting step further includes: Determine an initial pose for the standard geometric model such that the central axis of the standard geometric model is located in the vicinity of the design rotation center axis; During the iterative adjustment process, the degree of deviation between the central axis of the standard geometric model and the design rotation center axis is continuously monitored; When the change in the overall cumulative deviation is less than the preset convergence threshold and the degree of deviation tends to stabilize, the adjustment stops, and the current standard geometric model is taken as the final fitting result.

4. A high-precision detection system for the outline dimensions of an auxiliary washing machine glass viewing window, used to implement the method according to any one of claims 1 to 3, characterized in that, include: The signal acquisition module is used to set the trigger sampling mode to equal spatial interval sampling according to the movement trajectory of the drive mechanism to eliminate the influence of motion speed fluctuation on the uniformity of sampling point distribution during the process of the drive mechanism driving the point laser displacement sensor to move along the preset scanning path. It also triggers the sensor to collect data according to the preset spatial sampling interval, performs analog-to-digital conversion on the collected analog signals, and obtains a set of point cloud data that is spatially uniformly distributed on the preset scanning path. The data preprocessing module is used to perform local surface feature analysis on the point cloud dataset, obtain the curvature value at each sampling point, mark the sampling points with curvature values ​​exceeding the preset curvature threshold as edge candidate points, and perform spatial connectivity analysis on the edge candidate points to remove isolated edge candidate points that cannot form a connected region with the main edge, and determine the retained edge candidate points with continuous distribution characteristics as edge point cloud data. The geometric fitting module is used to select multiple feature points from edge point cloud data and iteratively adjust the pose of the standard geometric model based on preset geometric constraints. After each adjustment, the overall cumulative deviation between the standard geometric model and multiple feature points is redefined. When the change in the overall cumulative deviation after two adjacent adjustments is less than the preset convergence threshold, the adjustment is stopped and the current standard geometric model is taken as the final fitting result. The parameter extraction module is used to extract the contour dimension parameters of the washing machine glass viewing window based on a standard geometric model. The contour dimension parameters include at least one of diameter, roundness, and radius of curvature.

5. The high-precision detection system for the outline dimensions of the auxiliary washing machine glass observation window according to claim 4, characterized in that, The data preprocessing module includes a noise filtering unit and an edge recognition unit; The noise filtering unit is used to perform statistical analysis on the point cloud data set, determine the average distance between each sampling point and its neighboring points, identify sampling points whose average distance exceeds the overall distribution threshold range as outliers and remove them, and obtain the first point cloud data set. The first point cloud data set is then smoothed and filtered to eliminate random fluctuations. The edge recognition unit is used to perform local surface feature analysis on the point cloud data set after noise filtering, analyze and obtain the curvature value at each sampling point, mark the sampling points with curvature values ​​exceeding the preset curvature threshold as edge candidate points, and perform spatial connectivity analysis on the edge candidate points. Isolated edge candidate points that cannot form a connected region with the main edge are removed as interference points, and edge candidate points with continuous distribution characteristics are retained and determined as edge point cloud data.

6. The high-precision detection system for the outline dimensions of the auxiliary washing machine glass observation window according to claim 4, characterized in that, The geometric fitting module is further used for: Select multiple feature points from edge point cloud data; Determine an initial pose for the standard geometric model so that the model initially corresponds to multiple feature points in space; Based on preset geometric constraints, the pose of the standard geometric model is iteratively adjusted, and the overall cumulative deviation between the standard geometric model and multiple feature points is redetermined after each adjustment. Compare the overall cumulative deviation obtained after each adjustment with the overall cumulative deviation after the previous adjustment; When the change in the overall cumulative deviation obtained after two consecutive adjustments is less than the preset convergence threshold, the iterative process is determined to have converged, the adjustment is stopped, and the current standard geometric model is taken as the final fitting result.

Images