3D point cloud data intelligent processing method and processing device

Abstract

The invention relates to the field of data processing, in particular to a 3D point cloud data intelligent processing method and processing device. The method comprises the following steps: acquiring to-be-processed initial image data; the initial image data comprises a to-be-analyzed target object in the industrial monitoring scene; performing three-dimensional splicing on the initial image data to obtain a spliced image corresponding to the target object; performing virtual modeling based on the spliced image to obtain an initial virtual model of the target object; performing spatial form optimization on the initial virtual model through an adaptive optimization model to obtain an optimized virtual model; and extracting three-dimensional point cloud data of the target object from the optimized virtual model to complete intelligent extraction of the three-dimensional point cloud data. According to the method, automatic three-dimensional image analysis and extraction of the target object in the industrial monitoring scene can be realized, the image processing accuracy and processing efficiency of industrial monitoring are improved, and more comprehensive and timely automatic image processing services are provided for users.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a 3D point cloud data intelligent processing method and processing device. Background Art

[0002] In today's digital age, 3D point cloud data processing technology plays a key role in many fields. However, the amount of 3D point cloud data is too large, and it is difficult for enterprises to access and use it after collection, especially when the point cloud data of a workshop is retrieved online and then manually evaluated.

[0003] Moreover, in the manual evaluation process, due to differences in the experience, knowledge level and subjective judgment of the evaluators, the feature extraction is inaccurate, so different evaluators may get different results when evaluating the workshop data of the same enterprise. Moreover, even if the same evaluator evaluates the workshop data of the same enterprise at different times, the evaluation results may be inconsistent due to factors such as personal status and emotions. This makes the evaluation process lack repeatability and cannot guarantee the reliability and stability of the evaluation results. For example, in the analysis of the main welding point cloud data of the automobile enterprise workshop, it is difficult for manual evaluation to determine the quality of the main welding point in a timely and accurate manner through the data of the main welding point.

[0004] In addition, manual assessments often focus on extracting and assessing the company's current data, but lack an effective early warning mechanism for the risks that the company may face in the future. Assessors mainly analyze and judge based on historical data and current information, making it difficult to predict the potential risks of the company's workshop. Moreover, the manual assessment process lacks real-time monitoring and dynamic analysis of risk factors, making it impossible to detect and respond to changes in risks in a timely manner.

[0005] Therefore, there is an urgent need for a new solution to improve the quality of credit assessment, assist in improving credit assessment efficiency, and reduce credit assessment risks. Summary of the invention

[0006] In view of the technical problems existing in the prior art, the present invention provides a 3D point cloud data intelligent processing method and processing device, which are used to realize automatic three-dimensional image analysis and extraction of target objects in industrial monitoring scenes, improve the image processing accuracy and processing efficiency of industrial monitoring, and provide users with more comprehensive and timely automatic three-dimensional image processing services.

[0007] In a first aspect, an embodiment of the present application provides a method for intelligently processing 3D point cloud data, comprising: Acquire initial image data to be processed; the initial image data includes a target object to be analyzed in an industrial monitoring scene; Performing three-dimensional stitching on the initial image data to obtain a stitched image corresponding to the target object; Perform virtual modeling based on the stitched images to obtain an initial virtual model of the target object; the initial virtual model is a virtual three-dimensional model of the target object in the stitched image space; The initial virtual model is optimized in spatial form by an adaptive optimization model to obtain an optimized virtual model; the optimized virtual model is a virtual three-dimensional model of the target object in the image optimization space; The three-dimensional point cloud data of the target object is extracted from the optimized virtual model to complete the intelligent extraction of the three-dimensional point cloud data.

[0008] In a second aspect, an embodiment of the present application provides a 3D point cloud data intelligent processing system, the system comprising the following units: An acquisition unit, configured to acquire initial image data to be processed; the initial image data includes a target object to be analyzed in an industrial monitoring scene; A stitching unit, used for performing three-dimensional stitching on the initial image data to obtain a stitched image corresponding to the target object; A modeling unit, configured to perform virtual modeling based on the stitched image to obtain an initial virtual model of the target object; the initial virtual model is a virtual three-dimensional model of the target object in the stitched image space; An optimization unit, configured to optimize the spatial form of the initial virtual model through an adaptive optimization model to obtain an optimized virtual model; the optimized virtual model is a virtual three-dimensional model of the target object in the image optimization space; The extraction unit is used to extract the three-dimensional point cloud data of the target object from the optimized virtual model to complete the intelligent extraction of the three-dimensional point cloud data.

[0009] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device comprising: at least one processor, memory, and input-output unit; The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the 3D point cloud data intelligent processing method of the first aspect.

[0010] In a fourth aspect, a computer-readable storage medium is provided, which includes instructions, and when the instructions are executed on a computer, the computer executes the 3D point cloud data intelligent processing method of the first aspect.

[0011] The beneficial effects of the present invention are: providing a 3D point cloud data intelligent processing method and processing device. In this technical solution, the initial image data to be processed is obtained; the initial image data contains the target object to be analyzed in the industrial monitoring scene; the initial image data is three-dimensionally spliced ​​to obtain a spliced ​​image corresponding to the target object; virtual modeling is performed based on the spliced ​​image to obtain an initial virtual model of the target object; the initial virtual model is a virtual three-dimensional model of the target object in the spliced ​​image space; the initial virtual model is spatially optimized by an adaptive optimization model to obtain an optimized virtual model; the optimized virtual model is a virtual three-dimensional model of the target object in the image optimization space; the three-dimensional point cloud data of the target object is extracted from the optimized virtual model to complete the intelligent extraction of the three-dimensional point cloud data. This technical solution can realize the automated three-dimensional image analysis and extraction of target objects in industrial monitoring scenes, improve the image processing accuracy and processing efficiency of industrial monitoring, and provide users with more comprehensive and timely automated image processing services. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a flow chart of a 3D point cloud data intelligent processing method according to an embodiment of the present application; Figure 2 It is a structural schematic diagram of a 3D point cloud data intelligent processing system according to an embodiment of the present application; Figure 3 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application; Figure 4 It is a structural schematic diagram of a medium device according to an embodiment of the present application. DETAILED DESCRIPTION

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0014] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise clearly and specifically defined.

[0015] In the description of the present application, the term "for example" is used to mean "used as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid unnecessary details to obscure the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.

[0016] The embodiments of the present application provide a 3D point cloud data intelligent processing method and processing device.

[0017] In an embodiment of the present application, initial image data to be processed is obtained; the initial image data includes a target object to be analyzed in an industrial monitoring scene; the initial image data is three-dimensionally spliced ​​to obtain a spliced ​​image corresponding to the target object; virtual modeling is performed based on the spliced ​​image to obtain an initial virtual model of the target object; the initial virtual model is a virtual three-dimensional model of the target object in the spliced ​​image space; the initial virtual model is spatially optimized by an adaptive optimization model to obtain an optimized virtual model; the optimized virtual model is a virtual three-dimensional model of the target object in the image optimization space; and the three-dimensional point cloud data of the target object is extracted from the optimized virtual model to complete the intelligent extraction of the three-dimensional point cloud data.

[0018] The embodiments of the present application can realize automated three-dimensional image analysis and extraction of target objects in industrial monitoring scenarios, improve the image processing accuracy and efficiency of industrial monitoring, and provide users with more comprehensive and timely automated image processing services.

[0019] The 3D point cloud data intelligent processing solution provided in the embodiment of the present application can also be executed by an electronic device, which can be a server, a server cluster, or a cloud server. The electronic device can also be a terminal device such as a mobile phone, a computer, a tablet computer, a wearable device, or a dedicated device (such as a dedicated terminal device with a 3D point cloud data intelligent processing system). These electronic devices can also be equipped with the chips introduced in the above embodiments. Alternatively, these electronic devices can also be installed with a service program for executing the 3D point cloud data intelligent processing solution.

[0020] Figure 1 A schematic diagram of a 3D point cloud data intelligent processing method provided in an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps: 101. Obtaining initial image data to be processed; 102. Perform three-dimensional stitching on the initial image data to obtain a stitched image corresponding to the target object; 103. Perform virtual modeling based on the spliced ​​images to obtain an initial virtual model of the target object; 104. Optimizing the spatial form of the initial virtual model through an adaptive optimization model to obtain an optimized virtual model; 105. Extract three-dimensional point cloud data of the target object from the optimized virtual model to complete intelligent extraction of the three-dimensional point cloud data.

[0021] In an embodiment of the present application, the initial image data includes a target object to be analyzed in an industrial monitoring scene.

[0022] Taking mechanical equipment as an example, in industrial production, various types of mechanical equipment are common monitoring targets. For example, the spindle, tool holder, guide rail and other components of machine tool equipment. Power equipment such as engines, steam turbines, compressors, etc., their complex internal structures and moving parts need to be monitored to ensure the normal operation and performance optimization of the equipment. The initial image data of these devices can be used to detect defects such as wear, deformation, cracks, etc. of components, as well as monitor the operating status of the equipment and diagnose faults.

[0023] The target objects can be industrial parts, including various standard parts and customized parts, such as bolts, nuts, gears, bearings, molds, etc. By processing and analyzing the initial image data of these parts, their dimensional accuracy, surface quality, internal defects, etc. can be detected to ensure that the quality of parts meets production requirements and avoid equipment failures and production accidents caused by parts problems.

[0024] The target objects can be pipes and containers. For example, pipes and containers widely used in chemical, petroleum, power and other industries are also important monitoring objects. Pipeline corrosion, blockage, leakage, container deformation, cracks, weld quality and other problems can be discovered by analyzing the initial image data. For example, by processing the 3D point cloud data inside the pipeline, the corrosion of the inner wall of the pipeline and the distribution of sediments can be detected.

[0025] The target object can be an industrial building structure, such as the frame structure, wall, roof, etc. of a factory building. Monitoring of industrial building structures can ensure their safety and stability, and timely detect problems such as cracks and deformation of the structure. The initial image data can be used to perform 3D modeling of the building structure, so as to more intuitively observe and analyze the status of the structure.

[0026] Furthermore, the use of visual sensors such as industrial cameras and video cameras is a common way to collect initial image data. These sensors can be installed in fixed locations or on mobile devices to shoot the target object from multiple angles and directions. On an automated production line, multiple industrial cameras can be installed to monitor the product in real time; in the monitoring of large equipment, drones can be used to carry out all-round photography of the equipment.

[0027] Laser scanning equipment such as LiDAR can quickly obtain the three-dimensional point cloud data of the target object and can also generate corresponding image data. Laser scanning has the advantages of high precision, high speed, and non-contact, and is suitable for monitoring complex shapes and large target objects. In the monitoring of industrial pipelines, laser scanning can accurately obtain the three-dimensional shape and size information of the pipeline.

[0028] The structured light scanner projects a specific structured light pattern onto the target object and obtains the three-dimensional information of the target object based on the changes in the reflected light. This method is suitable for monitoring parts with high precision requirements and can obtain detailed texture and shape information on the surface of the target object.

[0029] Due to the use of multiple acquisition methods, the initial image data may come from different types of sensors with different formats and characteristics. The image data acquired by the visual sensor is usually a two-dimensional color image, while the data acquired by laser scanning and structured light scanning contains three-dimensional point cloud information and intensity information. This multi-source data needs to be effectively fused and processed to give full play to the advantages of various data.

[0030] In industrial monitoring scenarios, in order to fully and accurately reflect the status of the target object, it is usually necessary to collect a large amount of image data. For large industrial facilities or long-term monitoring tasks, the amount of data may be very large. This requires efficient data processing and storage technology to ensure that data management and analysis can proceed smoothly.

[0031] In some industrial monitoring applications, such as real-time monitoring of production processes, it is necessary to obtain and process initial image data in a timely manner in order to find problems and take measures in a timely manner. Therefore, the acquisition and processing of initial image data needs to have a certain degree of real-time performance to meet the needs of industrial production.

[0032] Specifically, for industrial monitoring scenarios, the initial image data containing the target object is obtained and three-dimensional stitching is performed, and a complete stitched image can be constructed from images collected from multiple angles and different positions. Compared with data collection from a single perspective, this can avoid the loss of some information about the monitored object, comprehensively cover all parts of the target object, provide a rich and complete data basis for subsequent analysis, and ensure that key components, complex structures and other details of industrial equipment can be fully captured. In the three-dimensional stitching process, the image information of each part can be accurately fused through image matching, alignment and other technologies. When monitoring industrial pipelines, the images of interfaces, welds and other parts of the pipeline under different shooting angles can be accurately connected, reducing information errors such as position and size caused by perspective deviation, improving the accuracy of the data, and providing a reliable basis for subsequent virtual modeling and analysis.

[0033] In the embodiment of the present application, the initial virtual model is a virtual three-dimensional model of the target object in the spliced ​​image space. The virtual three-dimensional model is constructed based on the spliced ​​image obtained by three-dimensionally splicing the initial image data containing the target object in the industrial monitoring scene. It is equivalent to using the spliced ​​image as the basic material to initially present the target object in a three-dimensional form in the virtual space, providing an initial morphological framework for subsequent processing and analysis.

[0034] It provides a preliminary virtual three-dimensional form for the target object, so that the target object originally in the two-dimensional image can be observed and understood in three-dimensional form, which is convenient for operators to view the overall structure and general outline of the target object from different angles, and helps to have a comprehensive preliminary understanding of the target object.

[0035] As the basic data model for further processing and optimization, it provides input for the adaptive optimization model. Subsequent operations such as spatial morphology optimization are all performed on the basis of this model. Its quality and accuracy will affect the effect and quality of the final 3D point cloud data extraction.

[0036] In industrial monitoring scenarios, the actual situation may be very complex, with various interference factors and complex backgrounds. The initial virtual model can separate the target object from the complex scene through preliminary modeling of the target object, and present it in the form of a relatively concise and clear three-dimensional model, which is convenient for subsequent targeted analysis and processing.

[0037] In the embodiment of the present application, the optimized virtual model is a virtual three-dimensional model of the target object in the image optimization space. The optimized virtual model is obtained by optimizing the spatial form of the initial virtual model through the adaptive optimization model. During the construction process, the adaptive optimization model will adjust the position of each position point in the initial virtual model according to the geometric features of the target object, the pre-set geometric constraints, etc., and then use the surface fitting algorithm to perform surface smoothing. It will also be further adjusted according to the proportional changes and size relationships of the target object in each dimension, and finally obtain the optimized virtual model.

[0038] Thus, through a series of optimization operations, the model can more accurately reflect the real shape and characteristics of the target object. For example, the adjustment of the position points and the surface smoothing can make the surface of the model smoother and more natural, more in line with the actual physical shape of the target object, and improve the visual and geometric accuracy of the model. Factors such as the proportional changes and size relationships of the target object in various dimensions are taken into account, so that the optimized model can better adapt to different observation angles and practical application scenarios. In industrial monitoring, different monitoring tasks may require observing the target object from different angles and scales. The optimized virtual model can maintain good performance in various situations and provide support for accurate analysis and decision-making. The optimized model provides a better basis for extracting 3D point cloud data from it. Due to its more accurate and stable shape, the extracted 3D point cloud data is of higher quality, with less noise and error, which is conducive to various subsequent analyses and applications based on 3D point cloud data, such as target recognition, defect detection, and dimensional measurement.

[0039] Specifically, the initial virtual model is constructed based on the stitched images, which can reflect the three-dimensional form of the target object in the stitched image space. After the adaptive optimization model is optimized in terms of geometric features, surface smoothness, proportional dimensions, etc., the optimized virtual model is closer to the real physical form of the industrial target object. When modeling industrial parts, the optimized model can accurately present the subtle textures on the surface of the parts and the transition of complex surfaces, improve the realism of the model, and help to more realistically simulate the actual state of the object in the industrial production process.

[0040] Spatial morphology optimization makes the model more in line with the needs of industrial monitoring data analysis. When analyzing the operating status of industrial equipment, optimizing the reasonable shape of the virtual model allows the analysis algorithm to better extract the deformation, displacement and other characteristics of key parts of the equipment, providing a more suitable model for fault diagnosis and performance evaluation, and improving the effectiveness and reliability of the analysis results.

[0041] Then, 3D point cloud data is extracted from the optimized virtual model to complete the intelligent extraction process without a lot of manual intervention, which greatly saves time and cost. In large-scale industrial facility monitoring projects, point cloud data can be quickly extracted from many equipment models to improve data processing efficiency, so that monitoring work can keep up with the pace of industrial production and detect potential problems in a timely manner.

[0042] Finally, the 3D point cloud data extracted after the previous multi-step processing has better data integrity, accuracy and consistency because it is based on a high-quality optimized virtual model. In the quality inspection of industrial equipment, high-quality point cloud data can accurately reflect problems such as equipment surface defects and dimensional deviations, reduce the probability of misjudgment and missed judgment, improve the quality and accuracy of industrial monitoring, and ensure the safe and stable operation of industrial production.

[0043] Exemplarily, in 101, the initial image data to be processed is obtained. According to the characteristics of the industrial monitoring scene and the target object, select the appropriate image acquisition device. For close-range, high-precision monitoring of small parts, a high-resolution industrial camera can be selected; for large-scale scenes or objects of complex shapes, lidar can obtain more comprehensive three-dimensional information; and for the monitoring of dynamic targets, a high-speed camera is more suitable. Plan the location, angle and quantity of acquisition. In order to fully obtain the information of the target object, it is necessary to shoot from multiple different positions and angles to ensure that no part is missed. For example, when monitoring a large industrial equipment, it is necessary to set multiple acquisition points around the equipment and shoot from different angles such as the top, side, and bottom. Consider the influence of the acquisition environment, such as lighting conditions, background interference, etc. During the acquisition process, try to avoid factors that affect image quality such as direct strong light and shadows, and use fill light equipment to improve lighting conditions. Operate the acquisition device to acquire images according to a predetermined acquisition plan. During the acquisition process, ensure the stability and accuracy of the equipment to avoid image blur or position deviation due to jitter or displacement.

[0044] Exemplarily, in 102, the initial image data is three-dimensionally stitched to obtain a stitched image corresponding to the target object. Feature extraction is performed on each initial image, and commonly used features include corners, edges, textures, etc. These features can help the subsequent image matching and stitching process. For example, SIFT (Scale Invariant Feature Transform) or SURF (Speeded Up Robust Features) algorithm is used to extract feature points in the image. Matching feature points are found between different images. By comparing the descriptors of the feature points, it is determined which feature points represent the same physical point in different images. The nearest neighbor matching algorithm or RANSAC (Random Sampling Consensus) algorithm can be used to improve the accuracy and reliability of matching.

[0045] Then, the relative position and posture relationship between different images are calculated based on the matched feature points. Then, these images are spatially transformed to align them in three-dimensional space. Finally, the aligned images are fused to obtain a spliced ​​image of the target object. During the fusion process, the overlapping areas between the images should be handled properly to avoid splicing marks and brightness differences.

[0046] Exemplarily, in 103, virtual modeling is performed based on the stitched image to obtain an initial virtual model of the target object. The three-dimensional point cloud data of the target object is extracted from the stitched image. The depth information of each pixel in the image can be calculated by stereoscopic vision, structured light and other methods to obtain the corresponding three-dimensional point coordinates. According to the generated point cloud data, surface reconstruction is performed to construct a three-dimensional surface model of the target object. Commonly used surface reconstruction algorithms include Poisson reconstruction, moving least squares method, etc. These algorithms can fit a smooth surface according to the distribution of the point cloud, so that the model is closer to the real shape of the target object.

[0047] The reconstructed model is optimized to remove noise points, fill holes, smooth the surface, etc. Filtering algorithms can be used to remove noise, interpolation algorithms can be used to fill holes, and surface fitting algorithms can be used to smooth the surface.

[0048] In the above or following embodiments, in step 104, the initial virtual model is optimized in spatial form by using an adaptive optimization model to obtain an optimized virtual model, which can be implemented as follows: Through the adaptive optimization model, the positions of various position points in the initial virtual model are adjusted according to the geometric features of the target object and the pre-set geometric constraints to obtain a first optimization model; the first optimization model is smoothed by using a surface fitting algorithm to obtain a second optimization model; the second optimization model is adjusted according to the proportional changes and size relationships of the target object in various dimensions to obtain the optimized virtual model.

[0049] Step 1: Adjust the position points according to the geometric features and constraints to obtain the first optimization model.

[0050] Specifically, the initial virtual model of the target object is analyzed to extract its key geometric features, such as point density, curvature, normal vector, etc. in the point cloud data. These features can reflect the surface shape, structure and other information of the target object. According to the needs of the industrial monitoring scene and the actual situation of the target object, a series of geometric constraints are set in advance. For example, for a mechanical part, it may be stipulated that certain planes should maintain parallel or vertical relationships, and the diameters of certain holes should be within a specific range. Based on the extracted geometric features and the preset geometric constraints, the positions of various position points in the initial virtual model are adjusted. This can be achieved by solving an optimization problem, such as minimizing a certain objective function, which usually takes into account factors such as the distance between points and the distance from points to the constraint plane. Through iterative calculation, the position of the points is continuously adjusted so that the model meets the geometric constraints.

[0051] Therefore, by adjusting the position points, the position deviation that may exist in the initial virtual model can be corrected to make the model more consistent with the actual geometry of the target object. For example, in the modeling of industrial parts, ensuring the relative position and size relationship of each part is accurate improves the model's restoration of the real object. Models that meet geometric constraints are more stable in subsequent processing and analysis. For example, when performing mechanical analysis or motion simulation, an accurate geometric model can provide a more reliable foundation and reduce the deviation of analysis results caused by model errors.

[0052] Step 2: Use a surface fitting algorithm to smooth the surface and obtain a second optimization model.

[0053] Specifically, according to the characteristics and application requirements of the first optimization model, a suitable surface fitting algorithm is selected, such as least squares surface fitting, spline surface fitting, etc. These algorithms can fit a smooth surface to approximate the surface of the target object based on the point cloud data in the model. The point cloud data in the first optimization model is input into the selected surface fitting algorithm, and the parameters of the fitting surface are obtained by calculation. In this process, some parameter adjustments and optimizations may be required to ensure the quality of the fitting surface. According to the parameters of the fitting surface, the points in the first optimization model are adjusted to make the surface of the model smoother. This can be achieved by projecting the points onto the fitting surface or interpolating it.

[0054] In this way, surface smoothing can eliminate jagged and discontinuous phenomena on the model surface, making the model look smoother and more natural. This is of great significance for some industrial products that require high appearance quality, such as automotive parts and aircraft engine blades. Smooth surface models can reduce the impact of noise and errors and improve data quality during subsequent analysis and processing. For example, when performing 3D measurement or reverse engineering, smooth surface models can more accurately reflect the actual size and shape of the target object.

[0055] Step 3: Adjust the position according to the proportion change and size relationship to obtain the optimized virtual model.

[0056] Specifically, the proportional changes and dimensional relationships of the target object in various dimensions are analyzed. This can be done by comparing with known standard models or design drawings, or by determining based on actual measurement data. Based on the proportional changes and dimensional relationships obtained from the analysis, the contour change trend of the target object in the second optimization model is predicted. This prediction can be achieved by establishing a mathematical model or using a machine learning algorithm. Based on the predicted contour change trend, the spatial position points in various dimensions in the second optimization model are adjusted. This can be achieved by scaling, translating, and other operations on the coordinates of each point, so that the proportions and sizes of the model conform to the actual conditions of the target object.

[0057] Thus, by adjusting the position, the proportion and size of the optimized virtual model are ensured to be consistent with the actual situation of the target object. This is very important for applications such as dimensional measurement and quality control in industrial monitoring, and can accurately determine whether the target object meets the design requirements. Models that conform to the actual proportions and size relationships are more practical in practical applications. For example, when performing virtual assembly or simulating production processes, accurate models can better reflect the actual situation and provide a more reliable basis for decision-making.

[0058] It is understandable that in the above steps, we first need to have a deep understanding and analysis of the geometric structure of the target object. For a three-dimensional target object, we need to consider its , , Observe how the proportional relationship between the length, width, height and other dimensions of the target object changes under different conditions (such as different working conditions, different environmental factors, etc.). For example, in some industrial production processes, due to thermal expansion and contraction, the target object The length of the dimension may change according to a certain proportional factor, and , The proportion of changes in dimensions may be different. Clarify the relationship between the sizes of various parts of the target object in different dimensions. For example, a mechanical part may have specific holes and protrusion structures. The sizes of these structures in different dimensions are mutually constrained. It is necessary to analyze the specific numerical relationship between them and the changing rules of this relationship under different circumstances. Then, for all position points in the second optimization model, according to the adjustment strategies and methods determined based on the proportion changes and size relationships, adjust their positions in each dimension in turn. After adjusting all position points, the original second optimization model is transformed into an optimized virtual model. This optimized virtual model can more accurately reflect the spatial form of the target object in the current actual situation, and better conform to the actual characteristics of the target object such as the proportion changes and size relationships in various dimensions, thereby providing a more accurate model basis for subsequent analysis, monitoring and other work.

[0059] Further optionally, in the above steps, the position of the second optimized model is adjusted according to the proportion change and size relationship of the target object in various dimensions to obtain the optimized virtual model, which can be implemented as follows: According to the proportional changes and size relationships of the target object in various dimensions, the outer contour change trend coefficient of the target object in the second optimization model is predicted; based on the predicted outer contour change trend coefficient, the spatial position points in various dimensions in the second optimization model are anomaly identified; the identified abnormal spatial points are adjusted to obtain the optimized virtual model.

[0060] First, it is necessary to collect various data about the target object under different conditions, including but not limited to historical monitoring data, dimensional specifications in design drawings, relevant process parameters, etc. For example, for a component of industrial equipment, collect its actual dimensional measurement data under different working environments such as temperature and pressure, as well as the standard dimensions and tolerance range specified during design.

[0061] Based on the collected data, analyze the target object in various dimensions ( , , By mathematical modeling, the relationship between the proportional change and related factors (such as temperature, time, etc.) can be found. For example, a linear regression model can be established to describe the functional relationship between the proportional change of the length of a component in a certain dimension and the temperature when the temperature changes.

[0062] Clarify the dimensional relationship between the various parts of the target object, such as the matching dimensions between components, relative position dimensions, etc. These dimensional relationships may remain stable or change under different working conditions. Analyze the changing rules of dimensional relationships and determine the constraints between various dimensions under different conditions.

[0063] Taking into account the proportional change model and the size relationship, the change trend coefficient of the target object's shape contour in the second optimization model is predicted. This coefficient can be a vector containing the change trend information in each dimension. For example, for a three-dimensional target object, the change trend coefficients are mapped to , , The coefficient of change trend of the shape profile in dimension.

[0064] Then, according to the predicted contour change trend coefficient, combined with the design standard of the target object and the experience in practical application, a normal range is set for the spatial position points in each dimension in the second optimization model. This normal range can be expressed by a mathematical formula, for example, In terms of dimension, for a certain point , its normal coordinate range can be expressed as ,in and It is calculated based on the change trend coefficient and other related factors.

[0065] For each spatial position point in the second optimization model = , calculate the deviation between the actual coordinate value in each dimension and the expected coordinate value predicted by the change trend coefficient. The deviation value in the dimension is the point predicted based on the change trend coefficient exist The expected coordinate value along the dimension.

[0066] The calculated deviation value is compared with the set normal range. If the deviation value exceeds the normal range, the location point is judged as an abnormal point. For example, if Greater than and The difference between exist It is abnormal in dimension. By comprehensively judging the deviations in each dimension, it is determined whether each location point is an abnormal point.

[0067] Then, for the identified abnormal spatial points, the adjustment strategy is determined according to their deviation and the actual needs of the target object. If the abnormality is caused by a change in proportion, it can be adjusted according to the predicted trend of the proportion change; if the abnormality is caused by a dimensional relationship that does not meet the requirements, it can be adjusted according to the correct dimensional relationship.

[0068] Adjust the coordinates of the abnormal space points in each dimension. Abnormal points in dimension , if its actual coordinate value Greater than the expected maximum , then adjust its coordinate value to ; If its actual coordinate value Less than the expected minimum , then adjust to By adjusting all abnormal spatial points, the outer contour of the second optimized model is made more consistent with the actual situation of the target object, thereby obtaining an optimized virtual model.

[0069] After completing the adjustment of the abnormal space points, the optimized virtual model is verified to check whether it meets the design requirements and actual application needs of the target object. If there are still situations that do not meet the requirements, further adjustments and optimizations can be made until a satisfactory optimized virtual model is obtained.

[0070] Exemplarily, in the above steps, based on the predicted contour change trend coefficient, abnormal identification of the spatial position points in each dimension in the second optimization model can be implemented as follows: For the spatial position points in the second optimization model , determine the spatial location point The average distance between the adjacent spatial position points; the product of the contour change trend coefficient and the average distance, and the spatial position point In the Add the coordinate reference values ​​in each dimension to get the spatial position point In the dimensions; after solving the expected coordinate value of each spatial position point in the second optimization model in the corresponding dimension, calculate the abnormal parameters corresponding to each spatial position point according to the expected coordinate value of each spatial position point and the real-time coordinate value of each spatial position point in the second optimization model; compare the abnormal parameters of each spatial position point with the dynamic threshold to screen out the abnormal spatial points.

[0071] Through the above steps, the spatial position points in the second optimization model can be effectively identified as abnormal based on the predicted contour change trend coefficient, providing a basis for subsequent adjustment of abnormal points and further optimization of the model.

[0072] In the above steps, the spatial position point in the second optimization model The calculation process of the corresponding abnormal parameters is expressed as the following formula: ; in, Represents a spatial location point Corresponding abnormal parameters, weight factors To adjust the The contribution of dimensions in abnormal judgment, Represents a spatial location point In the The actual coordinate value in dimension, Represents a spatial location point In the The expected coordinate value in dimension, , , express In the The projection coordinate value on the dimension, Indicates that all spatial position points in the second optimization model are The average coordinate value in the dimension, is the number of spatial position points in the second optimization model, For indicating that each spatial position point in the second optimization model is The average fluctuation range in the dimension.

[0073] First, the formula for the spatial position point exist , , The method of comprehensively calculating and summing the deviations in three dimensions can fully reflect the overall abnormality of the point in three-dimensional space. In industrial monitoring scenarios, the spatial morphological changes of target objects are often multi-dimensional. This comprehensive calculation method avoids considering only a single dimension and ignoring the impact of changes in other dimensions on abnormal judgment, making the abnormal identification results more accurate and reliable. For example, for a complex mechanical part, its dimensional changes in different dimensions may affect the function and performance of the part. Comprehensive multi-dimensional information can more comprehensively capture the abnormalities caused by these changes.

[0074] Secondly, the expected coordinate value is predicted based on factors such as the contour change trend coefficient, reflecting the position information of the target object under normal circumstances. The greater the deviation between the actual coordinates and the expected coordinates, the more likely the point is an abnormal point. This calculation method can effectively capture the abnormal changes of spatial position points and provide an important basis for abnormality identification.

[0075] Third, by comparing the actual coordinates with the average coordinates and combining them with the average fluctuation range, it is possible to determine whether the coordinate change of the point in dimension exceeds the normal fluctuation range. In industrial production, due to the influence of various factors, the coordinates of spatial position points will fluctuate to a certain extent, but if the coordinate change of a certain point exceeds the average fluctuation range, then it is likely to be an abnormal point. This way of considering the average fluctuation range can reduce misjudgments caused by normal fluctuations and improve the accuracy of abnormal identification.

[0076] This abnormal parameter calculation method can adapt to different data distribution and industrial monitoring scenarios. Whether the data points are distributed evenly or there is local dense or sparse distribution, anomalies can be effectively identified by comprehensively considering factors such as multi-dimensional information, weight factors, actual and expected coordinate deviations, and average fluctuation range. At the same time, for different types of target objects and monitoring tasks, by reasonably adjusting weight factors and other parameters, the abnormal parameter calculation method can better adapt to specific needs and has strong versatility and flexibility.

[0077] In summary, the above abnormal parameter calculation process can more accurately and effectively identify abnormal spatial points in the second optimization model by integrating multiple factors, providing a reliable basis for subsequent model optimization and industrial monitoring analysis.

[0078] Exemplarily, based on the above formula, in the above steps, comparing the abnormal parameters of each spatial position point with the dynamic threshold to screen out the abnormal spatial point can be implemented as follows: Based on the preset adjustment strategy, the dynamic threshold corresponding to each spatial position point is obtained in real time; if the abnormal parameter is greater than the corresponding dynamic threshold, the spatial position point corresponding to the current abnormal parameter is determined to be an abnormal spatial point; if the abnormal parameter is not greater than the corresponding dynamic threshold, the spatial position point corresponding to the current abnormal parameter is determined to be a normal spatial point.

[0079] The following is a detailed description of the implementation of the preset adjustment strategy and the technical effects of the above steps: For example, we can collect a large amount of abnormal parameter data of spatial location points in historical data and analyze the distribution of these data, such as calculating their mean, standard deviation and other statistics. Based on these statistical information, we can determine the dynamic threshold. For example, we can set the dynamic threshold to be the mean plus several times the standard deviation, so that most normal data can fall within the threshold range, and abnormal data can be effectively identified. With the continuous addition of new data, these statistics are continuously updated, so as to achieve real-time adjustment of the dynamic threshold.

[0080] For example, if the monitoring environment changes, such as changes in the operating conditions of the production equipment, changes in the material or process of the monitored object, etc., the dynamic threshold is adjusted accordingly. For example, when the operating speed of the production equipment increases, the fluctuation of the spatial position point may increase. At this time, the dynamic threshold can be appropriately increased to avoid misjudging normal fluctuations as abnormalities. By establishing a mapping relationship between environmental parameters and dynamic thresholds, the threshold can be adjusted according to the environmental parameters monitored in real time.

[0081] Alternatively, you can also use classification algorithms in machine learning, such as support vector machines and decision trees, to train known normal and abnormal spatial location point data and establish a classification model. Then dynamically adjust the threshold based on the output results of the model. For example, when the model prediction accuracy is high, it means that the current threshold setting is relatively reasonable and can be maintained or fine-tuned; when the model has many misjudgments, adjust the threshold based on the misjudgment situation to continuously optimize the performance of the model. You can also use a clustering algorithm to cluster the abnormal parameters of the spatial location points and determine the thresholds of different categories based on the clustering results.

[0082] In addition, combined with the professional knowledge and practical experience of the domain expert experience library, suggestions are provided for dynamic threshold settings in different situations. For example, based on the experience of previous similar projects, experts know that under certain specific conditions, the range of abnormal parameters of spatial location points is normal and the range beyond which they are abnormal. Combining these experiences with the results of data statistics, algorithm analysis, etc., a more reasonable preset adjustment strategy is formulated to adjust the dynamic threshold in real time.

[0083] Therefore, by comparing the abnormal parameters with the dynamic threshold, it is possible to clearly determine whether the spatial location point is abnormal, accurately screen out the abnormal spatial points, and provide accurate information for subsequent model optimization and problem handling, avoiding erroneous operations or ignoring real abnormal situations caused by misjudgment. The setting of dynamic thresholds and preset adjustment strategies can enable the system to adapt to various dynamic changes, whether it is changes in the monitored object itself or changes in the monitored environment, to ensure the effectiveness and accuracy of abnormal identification and improve the stability and reliability of the system. The preset adjustment strategy enables the system to automatically adjust the dynamic threshold according to different situations without frequent manual intervention, which improves the system's adaptability and intelligence level, reduces labor costs, and also reduces errors and mistakes that may be caused by human factors. After accurately screening out abnormal spatial points, the abnormal points can be processed in a targeted manner, and the virtual model can be further optimized so that the model can more accurately reflect the real situation of the target object, improve the accuracy and credibility of the model, and provide more reliable support for subsequent analysis and decision-making.

[0084] In practical applications, timely detection of abnormal spatial points helps to discover potential problems and risks in advance and take appropriate measures to deal with them, thereby ensuring the safe and stable operation of the entire system and avoiding adverse consequences such as equipment failure and production accidents caused by failure to detect abnormalities in time. It has important practical significance and economic value.

[0085] Further optionally, in the above steps, the first optimization model is subjected to surface smoothing processing using a surface fitting algorithm to obtain a second optimization model, which can be implemented as follows: The first optimization model is processed by minimizing the curvature change; the first optimization model is processed by surface fitting through an energy function; the optimization result of minimizing the curvature and the surface fitting result are fused to obtain the second optimization model.

[0086] Specifically, first, analyze each part of the first optimized model surface. For each small area of ​​the model surface, calculate its curvature. The curvature reflects the degree of curvature of the surface in that area. Then, set a goal to minimize the change in curvature of the entire model surface. To achieve this goal, adjust those areas where the curvature changes greatly. For example, if it is found that the curvature of a certain part of the model surface suddenly increases or decreases, the position of the points in that area is fine-tuned to make it closer to the curvature of the surrounding area, so that the curvature of the entire model surface changes more gently. During the adjustment process, make sure that the overall shape and key features of the model are not destroyed, and only the slight degree of local curvature is optimized.

[0087] Furthermore, an energy function is introduced, which can measure the degree of difference between the model surface and the ideal surface. The ideal surface is a relatively smooth and regular surface shape that is pre-set according to the characteristics and requirements of the target object. The surface of the first optimized model is analyzed, and the points on the model surface are associated with the energy function. The value of the energy function is minimized by adjusting the position of the model surface points. When the value of the energy function is the smallest, it means that the difference between the model surface and the ideal surface is the smallest, and a better surface fitting is achieved. When adjusting the position of the points, the relationship between the points and their position information in three-dimensional space will be considered to ensure that the adjusted surface meets the requirements of the energy function and maintains the rationality of the model.

[0088] Next, the model state after minimizing the curvature change and the model state after the surface fitting are obtained respectively. Taking these two results into consideration, the model is further adjusted and optimized. For example, the adjustment results of the model point positions by the two processing methods can be combined according to a certain weight ratio. If the minimizing curvature change processing is considered more important, its weight is appropriately increased; if the effect of the surface fitting processing is more critical, its weight is increased accordingly. Through this fusion processing, the model can maintain the smoothness of the surface curvature change and be as close to the ideal surface as possible, and finally a smoother and more accurate second optimization model is obtained.

[0089] In this way, after minimizing the curvature change and the surface fitting, the unevenness and mutation of the model surface are significantly improved, making the model surface smoother. This is of great significance for some application scenarios with high requirements on surface quality, such as the appearance design of industrial products, fluid mechanics simulation, etc., and can improve the accuracy of simulation and analysis. The second optimization model after fusion processing is closer to the real form of the target object. Because it comprehensively considers the curvature change and the fitting of the ideal surface, it reduces the error and deviation of the model and can more accurately reflect the geometric characteristics of the target object. In industrial monitoring, such a model can be more reliably used to detect and analyze the state of the target object, and improve the accuracy and reliability of monitoring. The smooth and accurate second optimization model can play a better role in various applications. Whether it is used in fields such as three-dimensional reconstruction, virtual reality or computer-aided design, it can provide a better data foundation, making related operations and analysis smoother and more efficient. The good quality of the second optimization model brings convenience to subsequent processing and analysis. For example, when extracting 3D point cloud data, a smooth model can reduce noise and interference and improve the quality of point cloud data; when further optimizing and deforming the model, it is also easier to implement and control, saving processing time and computing resources.

[0090] Exemplarily, in 105, the three-dimensional point cloud data of the target object is extracted from the optimized virtual model to complete the intelligent extraction of the three-dimensional point cloud data. The three-dimensional point cloud data of the target object is extracted from the optimized virtual model. The point coordinates of the model surface can be directly obtained, or a certain number of points can be uniformly selected on the model surface by sampling to generate point cloud data. The extracted point cloud data is cleaned and processed to remove noise points, outliers, etc. Statistical filtering, radius filtering and other algorithms can be used to remove noise, and clustering algorithms can be used to separate different objects. The processed three-dimensional point cloud data is output as the final result to provide data support for subsequent industrial monitoring, analysis and decision-making.

[0091] The embodiments of the present application can realize automated three-dimensional image analysis and extraction of target objects in industrial monitoring scenarios, improve the image processing accuracy and efficiency of industrial monitoring, and provide users with more comprehensive and timely automated image processing services.

[0092] Figure 2 A schematic diagram of the structure of a 3D point cloud data intelligent processing system provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system comprises the following steps: An acquisition unit, configured to acquire initial image data to be processed; the initial image data includes a target object to be analyzed in an industrial monitoring scene; A stitching unit, used for performing three-dimensional stitching on the initial image data to obtain a stitched image corresponding to the target object; A modeling unit, configured to perform virtual modeling based on the stitched image to obtain an initial virtual model of the target object; the initial virtual model is a virtual three-dimensional model of the target object in the stitched image space; An optimization unit, configured to optimize the spatial form of the initial virtual model through an adaptive optimization model to obtain an optimized virtual model; the optimized virtual model is a virtual three-dimensional model of the target object in the image optimization space; The extraction unit is used to extract the three-dimensional point cloud data of the target object from the optimized virtual model to complete the intelligent extraction of the three-dimensional point cloud data.

[0093] Further optionally, the optimization unit optimizes the spatial form of the initial virtual model through an adaptive optimization model to obtain an optimized virtual model, which is specifically used for: By using the adaptive optimization model, adjusting the positions of various position points in the initial virtual model according to the geometric features of the target object and the preset geometric constraints, so as to obtain a first optimization model; Using a surface fitting algorithm, the first optimization model is subjected to surface smoothing processing to obtain a second optimization model; According to the proportion changes and size relationships of the target object in various dimensions, the position of the second optimized model is adjusted to obtain the optimized virtual model.

[0094] Further optionally, the optimization unit adjusts the position of the second optimization model according to the proportion change and size relationship of the target object in various dimensions to obtain the optimized virtual model, which is specifically used to: Predicting a change trend coefficient of an outer contour of the target object in the second optimization model according to a change in proportion and a size relationship of the target object in various dimensions; Based on the predicted contour change trend coefficient, anomaly identification is performed on the spatial position points in each dimension in the second optimization model; The identified abnormal spatial points are adjusted to obtain the optimized virtual model.

[0095] Further optionally, the optimization unit, based on the predicted contour change trend coefficient, performs abnormality identification on the spatial position points in each dimension in the second optimization model, specifically for: For the spatial position points in the second optimization model , determine the spatial location point The average distance between adjacent spatial locations; The product of the contour change trend coefficient and the average distance is calculated and compared with the spatial position point In the Add the coordinate reference values ​​in each dimension to get the spatial position point In the Expected coordinate values ​​in dimensions; After solving and obtaining the expected coordinate value of each spatial position point in the corresponding dimension in the second optimization model, the abnormal parameters corresponding to each spatial position point are calculated according to the expected coordinate value of each spatial position point and the real-time coordinate value of each spatial position point in the second optimization model; The abnormal parameters of each spatial position point are compared with the dynamic threshold to screen out the abnormal spatial point.

[0096] Further optionally, the spatial position point in the second optimization model The calculation process of the corresponding abnormal parameters is expressed as the following formula: ; in, Represents a spatial location point Corresponding abnormal parameters, weight factors To adjust the The contribution of dimensions in abnormal judgment, Represents a spatial location point In the The actual coordinate value in dimension, Represents a spatial location point In the The expected coordinate value in dimension, , , express In the The projection coordinate value on the dimension, Indicates that all spatial position points in the second optimization model are The average coordinate value in the dimension, is the number of spatial position points in the second optimization model, For indicating that each spatial position point in the second optimization model is The average fluctuation range in the dimension.

[0097] Further optionally, the optimization unit compares the abnormal parameters of each spatial position point with the dynamic threshold to screen out the abnormal spatial point, specifically for: Based on the preset adjustment strategy, the dynamic threshold corresponding to each spatial position point is obtained in real time; If the abnormal parameter is greater than the corresponding dynamic threshold, the spatial position point corresponding to the current abnormal parameter is determined to be an abnormal spatial point; If the abnormal parameter is not greater than the corresponding dynamic threshold, the spatial position point corresponding to the current abnormal parameter is determined to be a normal spatial point.

[0098] Further optionally, the optimization unit uses a surface fitting algorithm to perform surface smoothing on the first optimization model to obtain a second optimization model, which is specifically used for: Performing curvature change minimization processing on the first optimization model; Performing surface fitting processing on the first optimization model through an energy function; The curvature minimization optimization result and the surface fitting result are fused to obtain the second optimization model.

[0099] See also Figure 3 , Figure 3 Schematic diagram of an electronic device provided in an embodiment of the present application. Figure 3 As shown, an embodiment of the present application provides an electronic device 500, including a memory 510, a processor 520, and a computer program 511 stored in the memory 510 and executable on the processor 520. When the processor 520 executes the computer program 511, the aforementioned embodiment is implemented.

[0100] See also Figure 4 , Figure 4 A schematic diagram of an embodiment of a computer-readable storage medium provided in an embodiment of the present application. Figure 4 As shown, this embodiment provides a computer-readable storage medium 600 on which a computer program 611 is stored. When the computer program 611 is executed by a processor, the aforementioned embodiment is implemented.

[0101] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0102] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0103] The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0104] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0105] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0106] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0107] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A 3D point cloud data intelligent processing method, characterized in that: The method at least comprises: Acquire initial image data to be processed; the initial image data includes a target object to be analyzed in an industrial monitoring scene; Performing three-dimensional stitching on the initial image data to obtain a stitched image corresponding to the target object; Perform virtual modeling based on the stitched images to obtain an initial virtual model of the target object; the initial virtual model is a virtual three-dimensional model of the target object in the stitched image space; The initial virtual model is optimized in spatial form by an adaptive optimization model to obtain an optimized virtual model; the optimized virtual model is a virtual three-dimensional model of the target object in the image optimization space; The three-dimensional point cloud data of the target object is extracted from the optimized virtual model to complete the intelligent extraction of the three-dimensional point cloud data.

2. The 3D point cloud data intelligent processing method according to claim 1, characterized in that: The step of optimizing the spatial form of the initial virtual model by using the adaptive optimization model to obtain the optimized virtual model includes: By using the adaptive optimization model, adjusting the positions of various position points in the initial virtual model according to the geometric features of the target object and the preset geometric constraints, so as to obtain a first optimization model; Using a surface fitting algorithm, the first optimization model is subjected to surface smoothing processing to obtain a second optimization model; According to the proportion changes and size relationships of the target object in various dimensions, the position of the second optimized model is adjusted to obtain the optimized virtual model.

3. The 3D point cloud data intelligent processing method according to claim 2, characterized in that: The step of adjusting the position of the second optimized model according to the proportion change and size relationship of the target object in various dimensions to obtain the optimized virtual model includes: Predicting a change trend coefficient of an outer contour of the target object in the second optimization model according to a change in proportion and a size relationship of the target object in various dimensions; Based on the predicted contour change trend coefficient, anomaly identification is performed on the spatial position points in each dimension in the second optimization model; The identified abnormal spatial points are adjusted to obtain the optimized virtual model.

4. The 3D point cloud data intelligent processing method according to claim 3, characterized in that: The method of identifying abnormalities of spatial position points in each dimension in the second optimization model based on the predicted contour change trend coefficient includes: For the spatial position points in the second optimization model , determine the spatial location point The average distance between adjacent spatial locations; The product of the contour change trend coefficient and the average distance is calculated and compared with the spatial position point In the Add the coordinate reference values ​​in each dimension to get the spatial position point In the Expected coordinate values ​​in dimensions; After solving and obtaining the expected coordinate value of each spatial position point in the corresponding dimension in the second optimization model, the abnormal parameters corresponding to each spatial position point are calculated according to the expected coordinate value of each spatial position point and the real-time coordinate value of each spatial position point in the second optimization model; The abnormal parameters of each spatial position point are compared with the dynamic threshold to screen out the abnormal spatial point.

5. The 3D point cloud data intelligent processing method according to claim 4, characterized in that: The spatial position points in the second optimization model The calculation process of the corresponding abnormal parameters is expressed as the following formula: ； in, Represents a spatial location point Corresponding abnormal parameters, weight factors To adjust the The contribution of dimensions in abnormal judgment, Represents a spatial location point In the The actual coordinate value in dimension, Represents a spatial location point In the The expected coordinate value in dimension, , , express In the The projection coordinate value on the dimension, Indicates that all spatial position points in the second optimization model are The average coordinate value in the dimension, is the number of spatial position points in the second optimization model, For indicating that each spatial position point in the second optimization model is The average fluctuation range in dimension.

6. The 3D point cloud data intelligent processing method according to claim 5, characterized in that: The step of comparing the abnormal parameters of each spatial position point with the dynamic threshold to screen out the abnormal spatial point includes: Based on the preset adjustment strategy, the dynamic threshold corresponding to each spatial position point is obtained in real time; If the abnormal parameter is greater than the corresponding dynamic threshold, the spatial position point corresponding to the current abnormal parameter is determined to be an abnormal spatial point; If the abnormal parameter is not greater than the corresponding dynamic threshold, the spatial position point corresponding to the current abnormal parameter is determined to be a normal spatial point.

7. The 3D point cloud data intelligent processing method according to claim 2, characterized in that: The method of using a surface fitting algorithm to perform surface smoothing processing on the first optimization model to obtain a second optimization model includes: Performing curvature change minimization processing on the first optimization model; Performing surface fitting processing on the first optimization model through an energy function; The curvature minimization optimization result and the surface fitting result are fused to obtain the second optimization model.

8. A 3D point cloud data intelligent processing system, characterized in that: The system comprises at least the following units: An acquisition unit, configured to acquire initial image data to be processed; the initial image data includes a target object to be analyzed in an industrial monitoring scene; A stitching unit, used for performing three-dimensional stitching on the initial image data to obtain a stitched image corresponding to the target object; A modeling unit, configured to perform virtual modeling based on the stitched image to obtain an initial virtual model of the target object; The initial virtual model is a virtual three-dimensional model of the target object in the spliced ​​image space; An optimization unit, configured to optimize the spatial form of the initial virtual model through an adaptive optimization model to obtain an optimized virtual model; the optimized virtual model is a virtual three-dimensional model of the target object in the image optimization space; The extraction unit is used to extract the three-dimensional point cloud data of the target object from the optimized virtual model to complete the intelligent extraction of the three-dimensional point cloud data.

9. An electronic device, characterized in that: including a memory for storing a computer software program; A processor is used to read and execute the computer software program, thereby implementing the 3D point cloud data intelligent processing method described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium, characterized in that: The storage medium stores a computer software program, and when the computer software program is executed by the processor, the 3D point cloud data intelligent processing method according to any one of claims 1 to 7 is implemented.

Images