Method, device and system for measuring linear size of foamed plastic product

Abstract

The invention relates to the technical field of model measurement, in particular to a method, a device and a system for measuring the linear size of a foamed plastic product. According to the method, registration analysis is carried out based on point cloud data obtained for multiple times from different scanning equipment angles of a foamed plastic product, and a target moment is screened through stable integrity analysis of the point cloud data after registration; point cloud data interfered by light transmission are screened through comparison and analysis of overall and local voxels of movement and reflection at continuous target moments; the key part in the product structure is considered, and the important area in the point cloud data is analyzed for precision adjustment, so that the accurate three-dimensional model of the foamed plastic product is determined, and the accuracy of the linear size measurement of the foamed plastic product is improved. According to the method, the light transmission interference is eliminated, the importance degrees of different areas are analyzed, and the precision of the three-dimensional models of the different areas is adjusted, so that the more accurate three-dimensional model is obtained, and the linear size measurement precision based on the three-dimensional model is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of model measurement, and in particular to a linear dimension measurement method, device and system for foam plastic products. Background Art

[0002] Traditional methods for linear dimensional measurement of foam plastic products typically utilize contact measurement methods, such as calipers. However, because foam materials are generally soft and elastic, contact measurement can cause sample compression or deformation, affecting measurement accuracy. Furthermore, foam products can contain complex curved surfaces and internal structures, making them inaccurately measured using traditional methods. Therefore, existing methods for linear dimensional measurement of foam plastic products typically utilize 3D scanning to capture the product's three-dimensional features and accurately determine its structure. This allows for linear measurement based on the 3D model structure.

[0003] However, due to interference in the production process of foam plastic products, such as uneven foaming, pressure fluctuations that cause bubble nuclei to expand and rupture rapidly to form a low-density open-pore structure, and inappropriate cooling rates, local density differences may exist in the foam plastic products, resulting in local light transmission interference during scanning, such as the formation of speckle noise, making the obtained three-dimensional model of the foam plastic product inaccurate, and unable to accurately adjust the accuracy of the three-dimensional model according to the structural complexity and importance of the foam plastic product, resulting in redundant detailed data of simple areas or missing data of complex structures, resulting in inaccurate three-dimensional models. Summary of the Invention

[0004] In order to solve the technical problem of inaccurate three-dimensional models of foam plastic products obtained in the prior art, the present invention aims to provide a method, device and system for measuring the linear dimensions of foam plastic products. The technical solutions adopted are as follows: The present invention provides a method for measuring the linear dimensions of a foam plastic product, the method comprising: On a foam plastic product conveyor belt, point cloud data is collected at each sampling moment using different devices; the structural features of each voxel in the point cloud data are obtained; the structural features include the maximum principal curvature, normal vector, and reflectivity; At each sampling moment, voxel registration is performed based on the proximity between the maximum principal curvature and reflectivity of voxels in point cloud data from different devices, as well as the similarity between normal vectors, to obtain valid point cloud data at each sampling moment and determine stable voxels. Target moments are selected based on the approximation between the volume of the valid point cloud data at the sampling moment and the standard volume, as well as the distribution of stable voxels and the point cloud density in the entire system. The valid point cloud data at consecutive target moments are converted to the same coordinate system and registered to obtain the preliminary screening point cloud data. Based on the reflectivity fluctuation differences and temporal movement differences between the unstable voxels and the whole point cloud data, as well as the instability, the light-transmitting interference points are screened out. Based on the local structural complexity and position distribution of each voxel in the initial screening point cloud data, the important indicators of each voxel are analyzed and the voxels are iteratively merged to obtain the important areas; A three-dimensional model is obtained by removing noise from light-transmitting interference points and adjusting them according to the distribution of important areas and important indicators; linear dimension measurement is achieved based on the three-dimensional model.

[0005] Furthermore, performing voxel registration based on the approximation of voxels between various structural features in point cloud data from different devices to obtain valid point cloud data at each sampling moment and determine stable voxels includes: At each sampling moment, the maximum principal curvature difference and reflectivity difference between any two voxels in the point cloud data of any two devices are calculated, and the product of the maximum principal curvature difference and the reflectivity difference is negatively correlated to obtain the structural similarity of the two voxels. The cosine similarity of the normal vectors between any two voxels is multiplied by the structural similarity to obtain the matching degree of the two voxels. When the matching degree between two voxels is the largest, the corresponding two voxels will be registered; the registered voxels will not be registered with other voxels between the two devices, and the point cloud data after all devices are registered will be used as the valid point cloud data at each sampling moment; In the valid point cloud data, the mean of all matching degrees of each voxel during registration is normalized and used as the stability index of each voxel; the voxels whose stability index is greater than the preset stability threshold are regarded as stable voxels.

[0006] Furthermore, the method for obtaining the target time includes: For any sampling moment, the mean value of the number of points contained in all voxels in the valid point cloud data at that sampling moment is taken as the point cloud density value at that sampling moment; the difference between the volume of the valid point cloud data at that sampling moment and the volume of the standard CAD model is negatively correlated to obtain the volume approximation at that sampling moment; Combine the total number of stable voxel midpoints, point cloud density value and volume approximation of the valid point cloud data at the sampling moment to obtain the complete index at the sampling moment; The moment when the completeness indicator is greater than the preset completeness threshold is taken as the target moment.

[0007] Furthermore, the method for obtaining the light-transmitting interference point includes: In the preliminary screening point cloud data, the voxels that are stable at all target moments are regarded as stable voxels of the preliminary screening point cloud data, and the remaining voxels are regarded as unstable voxels; For any unstable voxel in the initial screening point cloud data, the movement deviation index of the unstable voxel is obtained according to the degree of deviation between the displacement of the unstable voxel at the target time in the time sequence and the overall displacement of the valid point cloud data at the target time; Obtaining a reflectivity trend deviation index of the unstable voxel based on the difference between the reflectivity change fluctuation of the overall stable voxel and the reflectivity change fluctuation of the unstable voxel at consecutive target moments; Performing negative correlation mapping on the number of times the unstable voxel is used as a stable voxel at all target moments to obtain a stability deviation index of the unstable voxel; combining the stability deviation index, reflectivity trend deviation index, and movement deviation index of the unstable voxel to obtain an interference index of the unstable voxel; When the interference index is greater than the preset interference threshold, the point in the corresponding unstable voxel is regarded as a light-transmitting interference point.

[0008] Furthermore, the method for obtaining the movement deviation index includes: Obtain the center of gravity of the valid point cloud data at each target moment; between each two target moments, point the vector from the center of gravity of the previous target moment to the center of gravity of the next target moment as the overall movement vector between each two target moments; Between each two target moments, a vector pointing from the position of the unstable voxel at the previous target moment to the position at the next target moment is used as the displacement vector of the unstable voxel between each two target moments; The cosine similarity between the displacement vector of the unstable voxel and the overall motion vector between each two target moments is used as the approximation of the movement trend of the unstable voxel between each two target moments; The mean of all movement trend approximations of the unstable voxel between consecutive target moments is negatively correlated to obtain the movement deviation index of the unstable voxel.

[0009] Furthermore, the method for obtaining the reflectivity trend deviation index includes: Between every two target moments, the difference between the reflectivity of each stable voxel in the initial screening point cloud data at the next target moment and the reflectivity at the previous target moment is calculated as the fluctuation of each stable voxel in the initial screening point cloud data; the average of the fluctuations of all stable voxels in the initial screening point cloud data is used as the overall fluctuation index between every two target moments; the average of all overall fluctuation indexes in consecutive target moments is used as the overall fluctuation trend index; Between each two target moments, the difference between the reflectivity of the unstable voxel at the latter target moment and the reflectivity at the previous target moment is used as the local fluctuation index of the unstable voxel; the average of all local fluctuation indexes of the unstable voxel at consecutive target moments is used as the local fluctuation trend index of the unstable voxel; The difference between the local fluctuation trend index and the overall fluctuation trend index of the unstable voxel is used as the reflectivity trend deviation index of the unstable voxel.

[0010] Furthermore, based on the local structural complexity and position distribution of each voxel in the initial screening point cloud data, the important indicators of each voxel are analyzed and the voxels are iteratively merged to obtain the important areas, including: For any voxel in the initial screening point cloud data, excluding the light-transmitting interference points, the cosine similarity of the normal vector between the voxel and each voxel in the preset local neighborhood is calculated. Then, the mean of the cosine similarities is calculated and negative correlation mapping is performed to obtain the local structural deviation of the voxel. The mean of the normal vectors of the voxel and all the points in all voxels in the preset local neighborhood is used as the local density distribution of the voxel. The number of times the voxel is a stable voxel at all target moments is used as the stability of the voxel; and a stress estimation value at the position of the voxel is obtained; Based on the fact that the voxel belongs to the assembly surface, as well as the stress estimation value, local structure deviation, local density distribution and stability of the voxel, the important index of the voxel is obtained; The voxel with the largest importance index is used as the region to be merged. Within the preset local neighborhood of the region to be merged, when the difference between the importance index of the existing voxel and the importance index of the region to be merged is less than the preset merging threshold, the corresponding voxel is merged with the region to be merged to obtain a new region to be merged, and the average of the importance indexes of all voxels in the region to be merged is used as the importance index of the region to be merged. This process continues until the regions to be merged cannot be merged and a merged region is obtained. The remaining unmerged voxels are iteratively determined to determine the area to be merged and obtain the merged area; when the important index of the merged area is greater than the preset important threshold, the corresponding merged area is regarded as the important area.

[0011] Furthermore, the method for obtaining the three-dimensional model includes: De-noising the light-transmitting interference points in the initial screening point cloud data to obtain de-noised point cloud data; The average of the important indicators of all voxels in each non-important region is negatively correlated and normalized to the value used as the adjustment indicator of the region; the product of the adjustment indicator and the initial downsampled voxel size is used as the adjustment value; the sum of the initial downsampled voxel size and the adjustment value is used as the adjusted downsampled voxel size of each non-important region; The product of the normalized value of the important index mean of all voxels in each important region and the initial downsampled voxel size is used as the adjustment value; the difference between the initial downsampled voxel size and the adjustment value is used as the adjusted downsampled voxel size of each important region; The denoised point cloud data is adaptively downsampled based on the region-based adjusted downsampled voxel size, and a 3D model is generated through surface reconstruction.

[0012] The present invention also provides a linear dimension measurement system for foam plastic products, the system comprising: The data acquisition module is used to collect point cloud data at each sampling moment on the foam plastic product conveyor belt through different devices; obtain the structural characteristics of each voxel in the point cloud data; the structural characteristics include the maximum principal curvature, normal vector and reflectivity; The target moment analysis module is used to perform voxel registration at each sampling moment based on the approximation of voxels between various structural features in the point cloud data of different devices, obtain valid point cloud data at each sampling moment, and determine stable voxels. The target moment is selected based on the volume standard approximation of the valid point cloud data at the sampling moment, as well as the distribution of stable voxels and the point cloud density in the whole. The interference point analysis module is used to convert the valid point cloud data at consecutive target moments into the same coordinate system and align them to obtain preliminary screening point cloud data. Based on the reflectivity fluctuation differences and temporal movement differences between unstable voxels and the entire point cloud data, as well as the instability, light-transmitting interference points are screened out. The key area analysis module is used to analyze the important indicators of each voxel and iteratively merge the voxels to obtain the important areas based on the local structural complexity and position distribution of each voxel in the initial screening point cloud data; The measurement module is used to remove noise from light-transmitting interference points and make adjustments based on the distribution of important areas and important indicators to obtain a three-dimensional model; linear dimension measurement is achieved based on the three-dimensional model.

[0013] The present invention also provides a linear dimension measuring device for foam plastic products, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned methods for measuring the linear dimension of foam plastic products are implemented.

[0014] The present invention has the following beneficial effects: The present invention is based on the registration analysis of point cloud data obtained multiple times at different scanning device angles for a foam plastic product. The stability and integrity analysis of the registered point cloud data is then used to screen target moments to determine the scanned point cloud data of the complete foam plastic product structure for subsequent analysis, thereby ensuring the credibility of the continuous structural analysis. By performing a global and local voxel comparison analysis of the movement and reflection of consecutive target moments, point cloud data affected by light transmission interference is screened to reduce the possibility of noise being mismodeled as surface features. Furthermore, the point cloud data is analyzed for important areas within the product structure and precision adjustments are made based on key components. The adjusted detection results retain high-precision point clouds in assembly areas and structurally complex areas, while reducing the point cloud density in other areas. Only points with a high degree of contribution to the foam product structure are retained, thereby determining an accurate three-dimensional model of the foam plastic product and improving the accuracy of linear dimension measurements of the foam plastic product. The present invention eliminates light transmission interference, analyzes the importance of different areas, and adjusts the precision of the three-dimensional models of different areas to obtain a more accurate three-dimensional model, thereby increasing the accuracy of linear dimension measurements based on the three-dimensional model. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A flow chart of a method for measuring the linear dimensions of a foam plastic product provided by one embodiment of the present invention; Figure 2 A schematic diagram of a scanning device acquisition method provided by one embodiment of the present invention; Figure 3 A flow chart of a method for obtaining light-transmitting interference points provided by one embodiment of the present invention; Figure 4 A structural diagram of a linear dimension measurement system for foam plastic products provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0017] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effectiveness of a method, device, and system for measuring the linear dimensions of a foam plastic product according to the present invention. In the following description, references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The following describes in detail a method, device and system for measuring the linear dimensions of a foam plastic product provided by the present invention with reference to the accompanying drawings.

[0020] Example 1: See also Figure 1 , which shows a flow chart of a method for measuring the linear dimensions of a foam plastic product provided by one embodiment of the present invention, the method comprising the following steps: S1: On a foam plastic product conveyor belt, point cloud data is collected at each sampling moment using different devices; the structural features of each voxel in the point cloud data are obtained; the structural features include the maximum principal curvature, normal vector, and reflectivity.

[0021] In the embodiment of the present invention, foam plastic products are usually inspected in large quantities, so a conveyor belt is used for 3D scanning. Scanning equipment is installed directly above the conveyor belt and at a 60-degree angle on both sides. The scanning equipment is a structured light 3D scanner. Figure 2 , which shows a schematic diagram of a scanning device acquisition provided by an embodiment of the present invention.

[0022] Among them, the conveyor belt runs at a constant speed, the scanning probe scans once per second and the three probes scan simultaneously. The scanning moment is used as the sampling moment. Therefore, each scanning device can collect a type of point cloud data at the same sampling moment. The implementer of the specific multi-device scanning and acquisition process can adjust it according to the implementation scenario, and there is no restriction here.

[0023] The point cloud data can be matched and analyzed through the local structural features of the point cloud data in the voxels, because the curvature features and normal vector features can reflect the local morphological features of the surface of the foam plastic product. For example, the curvature of the edge of the foam gap is much higher than that of the plane area, and it is less affected by the angle change. The reflectivity feature can reflect the roughness feature of the surface of the foam plastic product. Therefore, the structural features in each voxel are obtained, and the structural features include the maximum principal curvature, normal vector and reflectivity. In an embodiment of the present invention, the side length of the voxel is 0.03mm, which can be adjusted by the implementer. The maximum principal curvature of each voxel is estimated by the fixed-point neighborhood, and the normal vector of each voxel is estimated by the local plane fitting method. It should be noted that the structural feature acquisition method is a public technical means well known to those skilled in the art and will not be elaborated here.

[0024] S2: At each sampling moment, voxel registration is performed based on the approximation of voxels between various structural features in the point cloud data of different devices to obtain valid point cloud data at each sampling moment and determine stable voxels; the target moment is screened out through the volume standard approximation of the valid point cloud data at the sampling moment, as well as the stable voxels and the point cloud density distribution in the whole.

[0025] The foam plastic product moves forward at a constant speed on the conveyor belt, gradually entering and then leaving the scanning range of the three-dimensional scanning device. Therefore, when the scanning device scans the foam plastic product, the scanning result obtained at a certain moment may fail to obtain the complete structure of the foam plastic product because the foam plastic product has not completely entered the scanning range or its own structure partially blocks the scanning light.

[0026] Therefore, the scanning results obtained by all scanning devices at the same time are matched. If the original point cloud density obtained at a certain moment is high, the obtained volume is less different from the standard CAD model and the local structural features are more similar to the CAD model, then it is more likely to cover a more complete structure of the foam plastic product, and the moment when the complete structure is covered is selected as the target moment.

[0027] First, a preliminary registration analysis is performed on the point cloud data of different devices. The registration process is mainly based on feature registration, that is, an approximate analysis is performed on each structural feature obtained to determine the matching result. The approximate situation is reflected by the closeness between the maximum principal curvature and the reflectivity, as well as the similarity between the normal vectors. The complete characterization of the point cloud data at each moment is obtained for analysis, and the stable voxels with better registration and less impact are determined to facilitate the subsequent screening of interference points. Therefore, in the embodiment of the present invention, the method for obtaining valid point cloud data and stable voxels includes: At each sampling moment, the maximum principal curvature difference and reflectivity difference between every two voxels in the point cloud data of any two devices are calculated, and the product of the maximum principal curvature difference and the reflectivity difference is negatively correlated to obtain the structural approximation of the two voxels, where the two voxels are located in the point cloud data of the two devices respectively. The smaller the difference, the closer the maximum principal curvature and reflectivity, the more similar the local rough morphological features, and the higher the possibility that they correspond to the same structural voxel.

[0028] The cosine similarity of the normal vectors between each pair of voxels is then multiplied by the structural similarity to obtain the matching degree of the two voxels. A higher cosine similarity indicates a more similar normal vector. Combined with the structural similarity, the matching degree of the two voxels is characterized. It should be noted that cosine similarity and negative correlation mapping are well known to those skilled in the art. Negative correlation mapping can take the form of, for example, an inverse proportional value or a negative exponential power, and is not further elaborated or limited here.

[0029] When the matching degree between two voxels is the largest, the corresponding two voxels will be registered. The registered voxels will not be registered with other voxels between the two devices. That is, when the two voxels have the maximum matching degree with each other, the two voxels will be matched. The voxels that are successfully matched will no longer participate in the calculation of the matching index of other voxels.

[0030] The point cloud data of the scanning results of each device at the same time are aligned voxel by voxel, and the point cloud data after all devices are aligned is used as the effective point cloud data at each sampling moment. The effective point cloud data is the data that characterizes a single foam plastic product at each sampling moment.

[0031] Furthermore, the mean of all matching degrees during registration for each voxel in the valid point cloud data is normalized to serve as a stability indicator for each voxel. In this embodiment of the present invention, point cloud data from three scanning devices is registered, and there are three matching degrees for each voxel. The higher the average matching degree, the more consistent the scanning results, and the voxel is less affected by the fiber angle. It should be noted that normalization is a technical means well known to those skilled in the art, and the normalization method can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0032] Therefore, voxels with stability indexes greater than a preset stability threshold are regarded as stable voxels. In the embodiment of the present invention, the preset stability threshold is set to 0.8. The specific value can be adjusted by the implementer and is not limited here.

[0033] Because foam plastic products are continuously scanned during transport, not every valid point cloud scanned represents the complete product structure. Therefore, further analysis reveals that if the volume of the valid point cloud at a given sampling moment is similar to that of the standard CAD model and contains a large number of stable voxels, indicating a relatively stable structure less affected by scanning angles, and the corresponding point cloud density is high, then the valid point cloud is more likely to retain relatively complete foam plastic structural features.

[0034] Preferably, in an embodiment of the present invention, the method for obtaining the target time includes: First, for any sampling moment, an analysis is performed for each sampling moment. The mean value of the number of points contained in each voxel in the valid point cloud data at that sampling moment is used as the point cloud density value at that sampling moment, reflecting the degree of point distribution in the single voxel. A higher distribution is likely to retain more complete information.

[0035] The difference between the volume of the valid point cloud data at the sampling moment and the volume of the standard CAD model is then negatively correlated to obtain the volume approximation at that sampling moment. The closer the volume is to the standard, the more likely the valid point cloud is to represent the complete foam plastic product structure. In this embodiment of the present invention, the volume of the valid point cloud data is the product of the number of voxels and the volume of each voxel.

[0036] Finally, the integrity index for that sampling moment is derived by combining the total number of stable voxel midpoints, the point cloud density, and the volume approximation of the valid point cloud data at that sampling moment. A greater total number of stable voxel midpoints indicates greater stability of the scanned structure, less impact, and more reliable data for subsequent analysis. In this embodiment of the present invention, the product of the total number of stable voxel midpoints, the point cloud density, and the volume approximation is used as the integrity index for that sampling moment. A higher integrity index indicates a greater likelihood of the presence of complete and reliable valid point cloud data at that sampling moment, useful for analyzing the complete structure of the product.

[0037] Therefore, the moment when the completeness index is greater than the preset completeness threshold is used as the target moment. In the embodiment of the present invention, the preset completeness threshold is set to 0.8. The specific value can be adjusted by the implementer and is not limited here.

[0038] S3: Convert the valid point cloud data at consecutive target moments to the same coordinate system and align them to obtain the preliminary screening point cloud data; based on the reflectivity fluctuation differences and temporal movement differences between the unstable voxels and the whole in the preliminary screening point cloud data, as well as the unstable conditions, screen out the translucent interference points.

[0039] During the scanning process, the porous structure and production process of foam plastic products may cause light transmittance interference in the scanning results. For example, the pores in the foam plastic make it easy for the scanning light source to penetrate the product surface, and may cause the laser to be reflected multiple times within the pores, resulting in point cloud coordinate offset. If the temperature, pressure, etc. are not properly controlled during the production process, the pores may increase, the light transmittance is stronger, and thus more noise points are generated.

[0040] Therefore, considering the differences and instabilities of the valid point cloud data at consecutive target moments, we screen out noise points that interfere with light transmission, thereby improving the accuracy of 3D structure determination in subsequent denoising. To facilitate analysis of consecutive target moments, the valid point cloud data at consecutive target moments are converted to the same coordinate system and aligned to obtain preliminary screening point cloud data.

[0041] In this embodiment of the present invention, the corresponding valid point cloud data is matched voxel by voxel according to the order of the target moments in the temporal sequence. Based on the voxel matching relationship, a rigid body transformation is performed using the least squares method to align all target point clouds to the same coordinate system. The matched point cloud data is then marked as the preliminary screening point cloud. It should be noted that the conversion method to the same coordinate system is a well-known technical means well known to those skilled in the art and will not be elaborated here.

[0042] Under normal circumstances, foam plastic products move at a constant speed along the conveyor belt, and the movement paths of their local areas also match this. This means that the movement trend of the normal, stable voxels in the valid point cloud data matches the overall movement trend of the product. Furthermore, due to the varying angle of incidence of the scanning device, the intensity of the reflected light varies with the angle of incidence, even when the material is uniform and the surface is flat. This means that the reflectivity of all normal, stable voxels in the valid point cloud data exhibits a certain regularity. However, when light transmission interference is present, the laser's reflection path may change, causing the reflectivity and path of the noise points of light transmission interference to fluctuate erratically.

[0043] Therefore, by analyzing the deviation of the unstable voxel in movement and reflectivity fluctuation, combined with the instability, preferably, in the embodiment of the present invention, the method for obtaining the light-transmitting interference point can be found in Figure 3 , which shows a flow chart of a method for obtaining light-transmitting interference points provided by an embodiment of the present invention, the method comprising the following steps: S301: In the preliminary screening point cloud data, the voxels that are stable at all target moments are regarded as stable voxels of the preliminary screening point cloud data, and the remaining voxels are regarded as unstable voxels; for any unstable voxel in the preliminary screening point cloud data, the movement deviation index of the unstable voxel is obtained according to the degree of deviation between the displacement of the unstable voxel at consecutive target moments in time sequence and the overall displacement of the valid point cloud data at the target moment.

[0044] The stable voxels and unstable voxels in the preliminary screening point cloud data are determined by the stable voxel conditions at consecutive target moments. The unstable voxels in the preliminary screening point cloud data are unstable voxels at at least one target moment, and there is a certain possibility that they are affected by the light angle.

[0045] Then, any unstable voxel is analyzed. When the movement of the unstable voxel at consecutive target moments deviates from the movement of the overall point cloud data, the less credible the corresponding unstable voxel is as a structural point cloud, and the higher the interference may be.

[0046] Preferably, in an embodiment of the present invention, the method for obtaining the movement deviation index includes: Obtain the center of gravity of the valid point cloud data at each target moment to analyze the overall movement. Between each target moment, use the vector from the center of gravity of the previous target moment to the center of gravity of the next target moment as the overall movement vector between the two target moments. The vector of the center of gravity coordinates represents the direction and degree of movement of the entire foam plastic product between the target moments.

[0047] Furthermore, between every two target moments, the vector pointing from the position of the unstable voxel at the previous target moment to the position at the next target moment is used as the displacement vector of the unstable voxel between every two target moments, and the degree of movement of the voxel between the target moments is represented by the vector of the voxel position coordinates.

[0048] Then, the cosine similarity between the displacement vector of the unstable voxel and the overall motion vector between each two target moments is used as the approximation of the movement trend of the unstable voxel between each two target moments. The larger the cosine similarity, the more consistent the movement situation.

[0049] Therefore, the approximation of all consecutive moments is comprehensively considered, and the mean of all movement trend approximations of the unstable voxel between consecutive target moments is negatively correlated to obtain the movement deviation index of the unstable voxel. The smaller the overall approximation, the higher the movement trend deviation, and the more likely the unstable voxel is an interference situation.

[0050] S302: Obtaining a reflectivity trend deviation index of the unstable voxel based on the difference between the reflectivity change fluctuation of the overall stable voxel and the reflectivity change fluctuation of the unstable voxel at consecutive target moments.

[0051] The degree of fluctuation deviation of angle change is further reflected by the continuous reflectivity. The higher the deviation between the fluctuations of reflectivity change, the less consistent the transmittance fluctuation of the unstable element is with the fluctuation trend of the overall reflectivity of the foam plastic product.

[0052] Preferably, in an embodiment of the present invention, the method for obtaining the reflectivity trend deviation index includes: Between every two target moments, the difference between the reflectivity of each stable voxel in the preliminary screening point cloud data at the subsequent target moment and the reflectivity at the previous target moment is calculated as the fluctuation degree of each stable voxel in the preliminary screening point cloud data. The average fluctuation degree of all stable voxels in the preliminary screening point cloud data is taken as the overall fluctuation index between every two target moments. The change in the reflectivity of the overall stable voxels reflects the degree of fluctuation of the overall reflectivity at adjacent moments.

[0053] Then, the average of all overall fluctuation indicators in consecutive target moments is used as the overall fluctuation trend indicator, and the fluctuation conditions of consecutive adjacent target moments are integrated to reflect the trend of fluctuation changes.

[0054] Then, the reflectivity change trend of each voxel is analyzed. Between every two target moments, the difference between the reflectivity of the unstable voxel in the latter target moment and the reflectivity in the previous target moment is used as the local fluctuation index of the unstable voxel. The average of all local fluctuation indexes of the unstable voxel in consecutive target moments is used as the local fluctuation trend index of the unstable voxel, reflecting the reflectivity change trend of the unstable voxel affected by the angle light in time series.

[0055] Finally, the difference between the local fluctuation trend index and the overall fluctuation trend index of the unstable voxel is used as the reflectivity trend deviation index of the unstable voxel. The greater the difference in the change trend, the worse the impact of the unstable voxel and the overall consistency, and the more likely the unstable voxel is an interference.

[0056] S303: Perform negative correlation mapping on the number of times the unstable voxel is used as a stable voxel at all target moments to obtain a stability deviation index of the unstable voxel; and combine the stability deviation index, reflectivity trend deviation index, and movement deviation index of the unstable voxel to obtain an interference index of the unstable voxel.

[0057] Finally, the unstable voxels may be stable in the continuous target moments. The stability deviation index of the unstable voxel is obtained by negative correlation mapping the number of stable voxels at the target moment. The larger the stability deviation index, the lower the probability of the stable voxel being analyzed as affected, and the greater the degree of instability.

[0058] The interference index of the unstable voxel is obtained by combining the stability deviation index, reflectivity trend deviation index and movement deviation index of the unstable voxel. In an embodiment of the present invention, the product of the stability deviation index, reflectivity trend deviation index and movement deviation index of the unstable voxel is used as the interference index of the unstable voxel. The larger the deviation, the higher the degree of instability, and the more likely the unstable voxel is an interference situation.

[0059] S304: Filtering light-transmitting interference points based on the interference index.

[0060] Therefore, through threshold judgment, when the interference index exceeds the preset interference threshold, the point in the corresponding unstable voxel is regarded as a light-transmitting interference point. In this embodiment of the present invention, the preset interference threshold can be set to 0.6. The specific value can be adjusted by the implementer and is not limited here. The light-transmitting interference point can be directly eliminated and denoised in the subsequent process.

[0061] S4: Based on the local structural complexity and position distribution of each voxel in the initial screening point cloud data, analyze the important indicators of each voxel and iteratively merge the voxels to obtain the important areas.

[0062] Generally speaking, due to the different application scenarios of foam plastic products themselves, the importance of different areas of them also varies. For example, if the foam plastic product currently being tested is used in scenarios such as automobile assembly, the importance of the foam plastic assembly surface is higher than that of the other surfaces. Moreover, if the structure of the assembly surface is more complex and the stress at the corresponding position is greater, more detailed information needs to be retained for this position.

[0063] Therefore, the importance of the region is further analyzed based on the location of each voxel and the complexity of the local structure. If the voxel is marked as a stable voxel more times in the target point cloud corresponding to different times, the geometric structure of the location is relatively stable, and it is more likely to be the core point of the local structure. If the stress at the voxel is large, the structure at that location is more prone to fatigue damage and the more high-precision detection is needed. If the normal vector of the voxel is significantly different from the normal vectors of other voxels in the adjacent range, the voxel is more likely to reflect local key features. Therefore, when the voxel is located on the assembly surface and the location is more complex, the importance index of the voxel should be larger.

[0064] In an embodiment of the present invention, a method for obtaining an importance index includes: For any voxel in the initial screening point cloud data except for the light-transmitting interference points, the light-transmitting interference points are the parts that need to be denoised and do not need to be analyzed in detail. After calculating the cosine similarity of the normal vector between the voxel and each voxel in the preset local neighborhood range, the mean of the cosine similarity is calculated and a negative correlation mapping is performed to obtain the local structural deviation of the voxel. When the cosine similarity of the normal vector is lower, the local deviation of the normal vector is larger, the local feature information may be more, and the voxel position may be more important. In an embodiment of the present invention, the preset local neighborhood range is set to a 6-neighborhood voxel range with the voxel as the center. The 6 neighborhoods include the front, back, left, right, up and down directions of the central voxel. The specific range setting can be adjusted by the implementer.

[0065] Furthermore, the mean of all points in each voxel and all voxels in the preset local neighborhood range is used as the local density distribution of the voxel. When the point distribution density of the voxels in the local range is higher, it may represent more information and the voxel position is more critical and important.

[0066] Furthermore, the number of times the voxel is used as a stable voxel in all target moments is used as the stability of the voxel. The more times it is used as a stable voxel, the higher the structural stability, the greater the possibility that the voxel position is the core of the structure, and the higher the importance.

[0067] Furthermore, a stress estimate is obtained for the voxel location. A larger stress estimate indicates that the voxel location is more susceptible to fatigue damage and requires greater attention. In this embodiment of the present invention, stress is the internal force generated per unit area within an object due to external forces. FEA is used to calculate and output the stress estimate at the voxel. It should be noted that stress estimation methods are well known to those skilled in the art and are not detailed or limited herein.

[0068] Finally, the important index of the voxel is obtained by combining the situation that the voxel belongs to the assembly surface, the stress estimation value, local structural deviation, local density distribution and stability of the voxel. In an embodiment of the present invention, when the voxel is on the assembly surface, the assembly index of the voxel is marked as a value of 1, otherwise the assembly index is marked as a value of 0. The product of the stress estimation value, local structural deviation, local density distribution and stability of the voxel is normalized as the structural important index of the voxel, and the sum of the structural important index and the assembly index is used as the important index of the voxel.

[0069] After evaluating the important situations based on all important key estimates of the voxels, voxels with similar important indices and close spatial positions are merged into the same region through local neighborhood analysis. In an embodiment of the present invention, the method for obtaining the important region includes: The voxel with the largest importance index is the region to be merged, and merging begins from the voxel with the largest importance index. Within the preset local neighborhood of the region to be merged, when the difference between the importance index of a voxel and the importance index of the region to be merged is less than a preset merging threshold, the corresponding voxel is merged with the region to be merged to obtain a new region to be merged. In this embodiment of the present invention, the preset merging threshold is set to 0.2, and the specific value can be adjusted by the implementer. That is, when the difference between the importance index of the region to be merged and the importance index of the voxel is less than 0.2, the region can be merged.

[0070] The average of the important indicators of all voxels in the area to be merged is used as the important indicator of the area to be merged, and the important indicators of the area are updated until the area to be merged cannot be merged, and a merged area is obtained. When there are no voxels that meet the conditions in the neighborhood, the current merging can be stopped.

[0071] The remaining unmerged voxels are iteratively determined to determine the region to be merged and obtain the merged region. For the remaining unmerged voxels, the voxels with the largest importance index are reselected as the region to be merged and merged. The iteration stops when no more merging is found. At this time, if the importance index of the merged region is greater than a preset importance threshold, the corresponding merged region is regarded as the important region. In this embodiment of the present invention, the preset importance threshold can be set to 1.6. The specific value can be adjusted voluntarily and is not limited here.

[0072] S5: Obtain a 3D model by removing noise from light-transmitting interference points and adjusting them based on the distribution of important areas and important indicators; perform linear dimension measurement based on the 3D model.

[0073] The accuracy of the three-dimensional model of the foam plastic product is adjusted based on the importance of the region where each voxel is located. That is, when a translucent noise point does not belong to an important region, the point is directly screened out. When it belongs to an important region, the voxel size is adjusted according to the importance of each region.

[0074] In this embodiment of the present invention, the light-transmitting interference points in the initial screening point cloud data are first denoised to obtain denoised point cloud data. The denoised initial screening point cloud data is then subjected to precision adjustment. Specifically, when a region is of high importance, its detail information should be retained, i.e., the region is downsampled to a smaller voxel size; otherwise, the downsampled voxel size should be increased.

[0075] Therefore, the mean of the important indicators of all voxels in each non-important region is negatively correlated and normalized to the value used as the region's adjustment indicator. The product of the adjustment indicator and the initial downsampled voxel size is used as the adjustment value. For non-important regions, the larger the important indicator, the smaller the increase. Therefore, the mean of the important indicators is negatively correlated to obtain the adjustment value. The sum of the initial downsampled voxel size and the adjustment value is used as the adjusted downsampled voxel size for each non-important region.

[0076] Furthermore, the product of the normalized mean of the importance index of all voxels in each important region and the initial downsampled voxel size is used as the adjustment value. For important regions, the larger the importance index, the greater the degree of reduction required to retain more information. Therefore, the difference between the initial downsampled voxel size and the adjustment value is used as the adjusted downsampled voxel size for each important region.

[0077] Finally, the denoised point cloud data is adaptively downsampled based on the region-based downsampled voxel size, and a 3D model is generated through surface reconstruction. The adjusted 3D model can be matched with the standard CAD model of the foam plastic product and linear dimension measurements can be performed. It should be noted that modeling of point cloud data is a well-known technique for those skilled in the art and will not be detailed here.

[0078] In an embodiment of the present invention, since the definition of linear dimension refers to the shortest distance between two specific points of a foam material sample, two parallel lines or two parallel planes determined by angles, edges or faces, a reference plane in a standard CAD model is taken as an example to represent the linear dimension acquisition process (CAD is converted to an STL model), and the corresponding surface in the point cloud three-dimensional model is projected onto the CAD reference plane, wherein the corresponding surface of the point cloud data needs to be plane fitted to obtain a projection point set, and the distance from the coordinates of the center of gravity of the projection point set to the reference plane is calculated as the distance between the two planes. The obtained distance is output to the staff to realize the linear dimension measurement of the foam plastic, wherein, for Class A injection molded parts, the tolerance range of the linear dimension is ±0.1% to ±0.2%, and for Class B injection molded parts, the tolerance range of the linear dimension is ±0.2% to ±0.3%. If the distance between the planes is greater than the design tolerance, a warning message can be output.

[0079] In summary, the present invention is based on the registration analysis of point cloud data obtained multiple times at different scanning device angles of foam plastic products, and through the stability and integrity analysis of the registered point cloud data, the target moment is screened to determine the point cloud data of the complete foam plastic product structure obtained by scanning for subsequent analysis, thereby ensuring the credibility of the continuous structure analysis. Through the overall and local voxel comparison analysis of the movement and reflection of consecutive target moments, the point cloud data affected by light transmittance interference is screened to reduce the situation where noise is mistakenly modeled as surface features, and considering the key parts of the product structure, the important areas in the point cloud data are analyzed for accuracy adjustment. The adjusted detection results retain high-precision point clouds in assembly areas and complex structural areas, reduce the point cloud density of other areas, and only retain points that contribute more to the structure of the foam product, thereby determining the accurate three-dimensional model of the foam plastic product and improving the accuracy of the linear dimension measurement of the foam plastic product. The present invention eliminates light transmittance interference, analyzes the importance of different areas, and adjusts the accuracy of the three-dimensional models of different areas, thereby obtaining a more accurate three-dimensional model and making the linear dimension measurement based on the three-dimensional model more accurate.

[0080] Example 2: The present invention also provides a linear dimension measurement system for foam plastic products, see Figure 4 , which shows a structural diagram of a linear dimension measurement system for foam plastic products provided by an embodiment of the present invention, the system includes: a data acquisition module 401, a target moment analysis module 402, an interference point analysis module 403, a key area analysis module 404 and a measurement module 405.

[0081] The data acquisition module 401 is used to collect point cloud data at each sampling moment on the foam plastic product conveyor belt using different devices; obtain the structural features of each voxel in the point cloud data; the structural features include the maximum principal curvature, normal vector and reflectivity; The target moment analysis module 402 is configured to perform voxel registration at each sampling moment based on the approximation of voxels between various structural features in point cloud data from different devices, thereby obtaining valid point cloud data at each sampling moment and determining stable voxels. The target moment is selected based on the volume standard approximation of the valid point cloud data at the sampling moment, as well as the distribution of stable voxels and the overall point cloud density. Interference point analysis module 403 is used to convert the valid point cloud data at consecutive target time points into the same coordinate system and align them to obtain preliminary screening point cloud data; based on the reflectivity fluctuation difference and time sequence movement difference between the unstable voxels and the whole point cloud data, as well as the instability, light-transmitting interference points are screened out; The key area analysis module 404 is used to analyze the important indicators of each voxel and iteratively merge the voxels based on the local structural complexity and position distribution of each voxel in the initial screening point cloud data to obtain the key areas; The measurement module 405 is used to remove noise from the light-transmitting interference points and make adjustments based on the distribution of important areas and important indicators to obtain a three-dimensional model; and to achieve linear dimension measurement based on the three-dimensional model.

[0082] It should be noted that the system provided in the above embodiment is merely illustrated by the division of the aforementioned functional modules. In actual applications, the aforementioned functions can be distributed among different functional modules as needed, i.e., the system's internal structure can be divided into different functional modules to perform all or part of the functions described above. Since the specific implementation process of the linear dimension measurement system for foam plastic products in this embodiment is the same as the specific implementation process of the linear dimension measurement method for foam plastic products described above, it will not be further elaborated here.

[0083] Example 3: The present invention also provides a linear dimension measuring device for foam plastic products, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned methods for measuring the linear dimension of foam plastic products are implemented.

[0084] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0085] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A method for measuring the linear dimensions of a foam plastic product, characterized in that: The method comprises: On a foam plastic product conveyor belt, point cloud data is collected at each sampling moment using different devices; the structural features of each voxel in the point cloud data are obtained; the structural features include the maximum principal curvature, normal vector, and reflectivity; At each sampling moment, voxel registration is performed based on the proximity between the maximum principal curvature and reflectivity of voxels in point cloud data from different devices, as well as the similarity between normal vectors, to obtain valid point cloud data at each sampling moment and determine stable voxels. Target moments are selected based on the approximation between the volume of the valid point cloud data at the sampling moment and the standard volume, as well as the distribution of stable voxels and the point cloud density in the entire system. The valid point cloud data at consecutive target moments are converted to the same coordinate system and registered to obtain the preliminary screening point cloud data. Based on the reflectivity fluctuation differences and temporal movement differences between the unstable voxels and the whole point cloud data, as well as the instability, the light-transmitting interference points are screened out. Based on the local structural complexity and position distribution of each voxel in the initial screening point cloud data, the important indicators of each voxel are analyzed and the voxels are iteratively merged to obtain the important areas; A three-dimensional model is obtained by removing noise from light-transmitting interference points and adjusting them according to the distribution of important areas and important indicators; linear dimension measurement is achieved based on the three-dimensional model.

2. A method for measuring the linear dimensions of a foam plastic product according to claim 1, characterized in that: The voxel registration is performed based on the approximation between the voxels of the point cloud data of different devices and the structural features, to obtain the valid point cloud data at each sampling moment and determine the stable voxels, including: At each sampling moment, the maximum principal curvature difference and reflectivity difference between any two voxels in the point cloud data of any two devices are calculated, and the product of the maximum principal curvature difference and the reflectivity difference is negatively correlated to obtain the structural similarity of the two voxels. The cosine similarity of the normal vectors between any two voxels is multiplied by the structural similarity to obtain the matching degree of the two voxels. When the matching degree between two voxels is the largest, the corresponding two voxels will be registered; the registered voxels will not be registered with other voxels between the two devices, and the point cloud data after all devices are registered will be used as the valid point cloud data at each sampling moment; In the valid point cloud data, the mean of all matching degrees of each voxel during registration is normalized and used as the stability index of each voxel; the voxels whose stability index is greater than the preset stability threshold are regarded as stable voxels.

3. The method for measuring the linear dimensions of a foam plastic product according to claim 1, wherein: The method for obtaining the target time includes: For any sampling moment, the mean value of the number of points contained in all voxels in the valid point cloud data at that sampling moment is taken as the point cloud density value at that sampling moment; the difference between the volume of the valid point cloud data at that sampling moment and the volume of the standard CAD model is negatively correlated to obtain the volume approximation at that sampling moment; Combine the total number of stable voxel midpoints, point cloud density value and volume approximation of the valid point cloud data at the sampling moment to obtain the complete index at the sampling moment; The moment when the completeness indicator is greater than the preset completeness threshold is taken as the target moment.

4. The method for measuring the linear dimensions of a foam plastic product according to claim 1, wherein: The method for obtaining the light transmittance interference point includes: In the preliminary screening point cloud data, the voxels that are stable at all target moments are regarded as stable voxels of the preliminary screening point cloud data, and the remaining voxels are regarded as unstable voxels; For any unstable voxel in the initial screening point cloud data, the movement deviation index of the unstable voxel is obtained according to the degree of deviation between the displacement of the unstable voxel at the target time in the time sequence and the overall displacement of the valid point cloud data at the target time; Obtaining a reflectivity trend deviation index of the unstable voxel based on the difference between the reflectivity change fluctuation of the overall stable voxel and the reflectivity change fluctuation of the unstable voxel at consecutive target moments; Performing negative correlation mapping on the number of times the unstable voxel is used as a stable voxel at all target moments to obtain a stability deviation index of the unstable voxel; combining the stability deviation index, reflectivity trend deviation index, and movement deviation index of the unstable voxel to obtain an interference index of the unstable voxel; When the interference index is greater than the preset interference threshold, the point in the corresponding unstable voxel is regarded as a light-transmitting interference point.

5. A method for measuring the linear dimensions of a foam plastic product according to claim 4, characterized in that: The method for obtaining the movement deviation index includes: Obtain the center of gravity of the valid point cloud data at each target moment; between each two target moments, point the vector from the center of gravity of the previous target moment to the center of gravity of the next target moment as the overall movement vector between each two target moments; Between each two target moments, a vector pointing from the position of the unstable voxel at the previous target moment to the position at the next target moment is used as the displacement vector of the unstable voxel between each two target moments; The cosine similarity between the displacement vector of the unstable voxel and the overall motion vector between each two target moments is used as the approximation of the movement trend of the unstable voxel between each two target moments; The mean of all movement trend approximations of the unstable voxel between consecutive target moments is negatively correlated to obtain the movement deviation index of the unstable voxel.

6. A method for measuring the linear dimensions of a foam plastic product according to claim 4, characterized in that: The method for obtaining the reflectivity trend deviation index includes: Between every two target moments, the difference between the reflectivity of each stable voxel in the initial screening point cloud data at the next target moment and the reflectivity at the previous target moment is calculated as the fluctuation of each stable voxel in the initial screening point cloud data; the average of the fluctuations of all stable voxels in the initial screening point cloud data is used as the overall fluctuation index between every two target moments; the average of all overall fluctuation indexes in consecutive target moments is used as the overall fluctuation trend index; Between each two target moments, the difference between the reflectivity of the unstable voxel at the latter target moment and the reflectivity at the previous target moment is used as the local fluctuation index of the unstable voxel; the average of all local fluctuation indexes of the unstable voxel at consecutive target moments is used as the local fluctuation trend index of the unstable voxel; The difference between the local fluctuation trend index and the overall fluctuation trend index of the unstable voxel is used as the reflectivity trend deviation index of the unstable voxel.

7. A method for measuring the linear dimensions of a foam plastic product according to claim 1, characterized in that: Based on the local structural complexity and position distribution of each voxel in the initial screening point cloud data, the important indicators of each voxel are analyzed and the voxels are iteratively merged to obtain important areas, including: For any voxel in the initial screening point cloud data, excluding the light-transmitting interference points, the cosine similarity of the normal vector between the voxel and each voxel in the preset local neighborhood is calculated. Then, the mean of the cosine similarities is calculated and negative correlation mapping is performed to obtain the local structural deviation of the voxel. The mean of the normal vectors of the voxel and all the points in all voxels in the preset local neighborhood is used as the local density distribution of the voxel. The number of times the voxel is a stable voxel at all target moments is used as the stability of the voxel; and a stress estimation value at the position of the voxel is obtained; Based on the fact that the voxel belongs to the assembly surface, as well as the stress estimation value, local structure deviation, local density distribution and stability of the voxel, the important index of the voxel is obtained; The voxel with the largest importance index is used as the region to be merged. Within the preset local neighborhood of the region to be merged, when the difference between the importance index of the existing voxel and the importance index of the region to be merged is less than the preset merging threshold, the corresponding voxel is merged with the region to be merged to obtain a new region to be merged, and the average of the importance indexes of all voxels in the region to be merged is used as the importance index of the region to be merged. This process continues until the regions to be merged cannot be merged and a merged region is obtained. The remaining unmerged voxels are iteratively determined to determine the area to be merged and obtain the merged area; when the important index of the merged area is greater than the preset important threshold, the corresponding merged area is regarded as the important area.

8. The method for measuring the linear dimensions of a foam plastic product according to claim 1, wherein: The method for obtaining the three-dimensional model includes: De-noising the light-transmitting interference points in the initial screening point cloud data to obtain de-noised point cloud data; The average of the important indicators of all voxels in each non-important region is negatively correlated and normalized to the value used as the adjustment indicator of the region; the product of the adjustment indicator and the initial downsampled voxel size is used as the adjustment value; the sum of the initial downsampled voxel size and the adjustment value is used as the adjusted downsampled voxel size of each non-important region; The product of the normalized value of the important index mean of all voxels in each important region and the initial downsampled voxel size is used as the adjustment value; the difference between the initial downsampled voxel size and the adjustment value is used as the adjusted downsampled voxel size of each important region; The denoised point cloud data is adaptively downsampled based on the region-based adjusted downsampled voxel size, and a 3D model is generated through surface reconstruction.

9. A linear dimension measurement system for foam plastic products, characterized in that: The system comprises: The data acquisition module is used to collect point cloud data at each sampling moment on the foam plastic product conveyor belt through different devices; obtain the structural characteristics of each voxel in the point cloud data; the structural characteristics include the maximum principal curvature, normal vector and reflectivity; The target moment analysis module is used to perform voxel registration at each sampling moment based on the approximation of voxels between various structural features in the point cloud data of different devices, obtain valid point cloud data at each sampling moment, and determine stable voxels. The target moment is selected based on the volume standard approximation of the valid point cloud data at the sampling moment, as well as the distribution of stable voxels and the point cloud density in the whole. The interference point analysis module is used to convert the valid point cloud data at consecutive target moments into the same coordinate system and align them to obtain preliminary screening point cloud data. Based on the reflectivity fluctuation differences and temporal movement differences between unstable voxels and the entire point cloud data, as well as the instability, light-transmitting interference points are screened out. The key area analysis module is used to analyze the important indicators of each voxel and iteratively merge the voxels to obtain the important areas based on the local structural complexity and position distribution of each voxel in the initial screening point cloud data; The measurement module is used to remove noise from light-transmitting interference points and make adjustments based on the distribution of important areas and important indicators to obtain a three-dimensional model; linear dimension measurement is achieved based on the three-dimensional model.

10. A linear dimension measuring device for a foam plastic product, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method for measuring the linear dimensions of a foam plastic product as claimed in any one of claims 1 to 8 are implemented.

Images