Stereoscopic vision length dimension accurate measurement method and system

Abstract

The invention discloses a stereoscopic vision length dimension accurate measurement method and system, and relates to the technical field of three-dimensional measurement, and the method comprises the steps: evaluating the point cloud density and visual angle consistency; and determining a first measurement method or a second measurement method based on the point cloud density and the view angle consistency, and measuring the length dimension of the object. According to the accurate measurement method for the stereoscopic vision length dimension, the problem that the measurement accuracy is obviously reduced under the condition that the point cloud density is not uniform or noise exists in a traditional single-view-angle measurement method is solved.

Description

Technical Field

[0001] The present invention relates to the field of three-dimensional measurement technology, and in particular to a method and system for accurately measuring a length dimension using stereoscopic vision. Background Art

[0002] Accurately measuring the dimensions of an object is critical to many industrial and scientific applications. Traditional measurement methods often rely on contact instruments, which may affect the measured object or are not suitable for objects with small or complex shapes.

[0003] Inside the generator, accurate 3D feature point extraction is essential for efficient and accurate internal inspection and maintenance. The stereo vision system can extract 3D feature points in a non-contact manner, improving detection accuracy and efficiency. We are developing a generator inspection robot based on a distributed multi-dimensional system. The robot uses advanced image processing and artificial intelligence technology to achieve visual inspection of the generator, stator slot wedge tightness test, and electromagnetic core defect detection. This places higher requirements on the robot's 3D feature point extraction technology. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is to solve the problem of large single-view measurement error and unstable precision: the traditional single-view measurement method is easily affected by the limitations of the viewing angle when facing objects with complex shapes, resulting in increased measurement errors. The present invention effectively reduces the error caused by the single view through a multi-view fusion measurement method, and improves the stability and precision of the measurement.

[0006] Solve the problem of inaccurate measurement when point cloud data quality is uneven or there is noise: When point cloud data is sparse or there is noise interference, traditional methods cannot effectively handle the unevenness of data, resulting in inaccurate measurement results. The present invention improves the accuracy of measurement by evaluating the point cloud density and perspective consistency, selecting a suitable weighting method or confidence weighting processing, and is particularly suitable for scenes with uneven data quality.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for accurately measuring the length dimension of a stereoscopic vision, comprising:

[0008] Evaluate point cloud density and view consistency;

[0009] The first measurement method or the second measurement method is determined based on the point cloud density and the viewing angle consistency to measure the length size of the object.

[0010] As a preferred solution of the method for accurately measuring the length dimension of stereoscopic vision described in the present invention, wherein: the point cloud density is expressed as,

[0011]

[0012] in, Indicates point position, V voxel Indicates volume.

[0013] As a preferred solution of the stereoscopic vision length dimension accurate measurement method of the present invention, wherein: the perspective consistency includes measuring the first feature point and the second feature point on the same object through any two perspectives;

[0014] The coordinates of the first feature point are A1=(x A1 ,y A1 ,z A1 )), (B1=(x B1 ,y B1 ,z B1 ); the coordinates of the second feature point are A2=(x A2 ,y A2 ,z A2 )), (B2=(x B2 ,y B2 ,z B2 ).

[0015] As a preferred solution of the method for accurately measuring the length dimension of stereoscopic vision described in the present invention, wherein: the viewing angle consistency is expressed as:

[0016]

[0017] ΔL=|L1-L2|.

[0018] As a preferred solution of the method for accurately measuring the length dimension of stereoscopic vision of the present invention, wherein: the determining of the first measurement method or the second measurement method includes selecting the first measurement method when D meets a preset threshold and ΔL meets a preset threshold;

[0019] When D does not meet the preset threshold or ΔL does not meet the preset threshold, the second measurement method is selected.

[0020] As a preferred solution of the method for accurately measuring the length of stereoscopic vision described in the present invention, the first measuring method includes:

[0021] The point cloud data P of each perspective is standardized and expressed as:

[0022]

[0023] Among them, μ and σ are the mean and standard deviation of the point cloud data respectively;

[0024] Use feature extraction algorithm to extract key points K from each view i i and the eigenvector V i ;

[0025] For each pair of key points P A and P B , calculate the Euclidean distance L in each view i i , expressed as,

[0026]

[0027] Define the quality Q of each view i , combined with the coverage of the viewing angle C i and point cloud density D i , expressed as,

[0028] Q i =λ×C i +(1-λ)×D i

[0029] Among them, C i and D i are the coverage rate and point cloud density of the i-th viewing angle, respectively, and λ represents the balance coefficient;

[0030] Calculate the weight W for each view i It is expressed as,

[0031]

[0032] Among them, γ is the weighting coefficient;

[0033] Calculate the final object length L final It is expressed as,

[0034]

[0035] As a preferred solution of the method for accurately measuring the length dimension of stereoscopic vision described in the present invention, wherein: the second measurement method includes standardizing the point cloud data P of each viewing angle;

[0036] Use feature extraction algorithm to extract key points K from each view angle i i and the eigenvector V i ;

[0037] For each pair of key points P A and P B , calculate the Euclidean distance L in each view i i , expressed as,

[0038]

[0039] Among them, xAi ,y Ai ,z Ai and x Bi ,y Bi ,z Bi They are the key points P A and P B Coordinates in the i-th viewing angle;

[0040] Calculate all viewing distance values ​​L i The standard deviation σ is expressed as,

[0041]

[0042] in, All L i The average value of , n is the number of viewing angles;

[0043] Calculate the confidence S for each view i , expressed as,

[0044]

[0045] Among them, κ is a positive proportionality coefficient;

[0046] Calculate the weight W for each view i It is expressed as,

[0047]

[0048] Among them, S i is the confidence of other perspectives;

[0049] Calculate the final object length L final It is expressed as,

[0050]

[0051] Another object of the present invention is to provide a stereoscopic vision length dimension precise measurement system, which can solve the problem that the measurement accuracy of the traditional single-view measurement method is significantly reduced when the point cloud density is uneven or there is noise by constructing a stereoscopic vision length dimension precise measurement system.

[0052] To solve the above technical problems, the present invention provides the following technical solutions: a stereoscopic vision length dimension precise measurement system, comprising: an evaluation module and a measurement module; the evaluation module is used to evaluate the point cloud density and the perspective consistency; based on the point cloud density and the perspective consistency, a first measurement method or a second measurement method is determined; the measurement module performs object length dimension measurement.

[0053] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned method for accurately measuring the length dimension of stereoscopic vision when executing the computer program.

[0054] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for accurately measuring the length dimension of stereoscopic vision.

[0055] Beneficial effects of the invention: The stereoscopic vision length dimension accurate measurement method provided by the invention has obvious technical advantages over the traditional measurement method in terms of data quality adaptability, multi-view fusion and measurement accuracy improvement, which makes the method have broad application prospects in the fields of industrial automation, quality control and three-dimensional modeling.

[0056] The present invention introduces a point cloud density and view consistency evaluation mechanism, which can automatically select the appropriate measurement method according to the data quality and view coverage. For scenes with uneven data density or insufficient view coverage, the measurement accuracy is improved through weighted averaging or confidence weighting processing, which adapts to different data quality conditions.

[0057] Through the multi-view data fusion method, the present invention can effectively reduce the single-view measurement error and enhance the reliability of the measurement result. In the presence of noise, the confidence weighting can significantly reduce the measurement error and ensure high-precision measurement in complex environments.

[0058] The present invention significantly improves the measurement accuracy and consistency by evaluating the quality and perspective consistency of point cloud data and using different weighting methods under different conditions. Experimental data show that the measurement error of the present method is significantly smaller than that of the traditional single-view measurement method, especially when the point cloud data is sparse or noisy.

[0059] The present invention is not only suitable for high-density, high-quality point cloud data scenarios, but can also handle low-density, insufficient coverage and noise interference scenarios, showing strong adaptability, and is particularly suitable for the measurement of complex-shaped objects in industrial and scientific research fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0061] Figure 1An overall flow chart of a method for accurately measuring length dimensions using stereoscopic vision provided by one embodiment of the present invention.

[0062] Figure 2 An overall structural diagram of a stereoscopic vision length dimension precision measurement system provided for the second embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0064] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0065] Example 1

[0066] Reference Figure 1 , as an embodiment of the present invention, provides a method for accurately measuring length dimensions using stereoscopic vision, comprising:

[0067] Evaluate point cloud density and view consistency;

[0068] The first measurement method or the second measurement method is determined based on the point cloud density and the viewing angle consistency to measure the length size of the object.

[0069] The point cloud density is expressed as,

[0070]

[0071] in, Indicates point position, V voxel Indicates volume.

[0072] The perspective consistency includes measuring a first feature point and a second feature point on the same object through any two perspectives;

[0073] The coordinates of the first feature point are A1=(x A1 ,y A1 ,z A1 )), (B1=(x B1 ,y B1 ,z B1 ); the coordinates of the second feature point are A2=(xA2 ,y A2 ,z A2 )), (B2=(x B2 ,y B2 ,z B2 ).

[0074] The perspective consistency is expressed as,

[0075]

[0076] ΔL=|L1-L2|.

[0077] The determining of the first measurement method or the second measurement method includes selecting the first measurement method when D satisfies a preset threshold and ΔL satisfies a preset threshold;

[0078] When D does not meet the preset threshold or ΔL does not meet the preset threshold, the second measurement method is selected.

[0079] The first measurement method includes

[0080] The point cloud data P of each perspective is standardized and expressed as:

[0081]

[0082] Among them, μ and σ are the mean and standard deviation of the point cloud data respectively;

[0083] Use feature extraction algorithm to extract key points K from each view i i and the eigenvector V i ;

[0084] For each pair of key points P A and P B , calculate the Euclidean distance L in each view i i , expressed as,

[0085]

[0086] Define the quality Q of each view i , combined with the coverage of the viewing angle C i and point cloud density D i , expressed as,

[0087] Q i =λ×C i +(1-λ)×D i

[0088] Among them, C i and D i are the coverage rate and point cloud density of the i-th viewing angle, respectively, and λ represents the balance coefficient;

[0089] Calculate the weight W for each view i It is expressed as,

[0090]

[0091] Among them, γ is the weighting coefficient;

[0092] Calculate the final object length L final It is expressed as,

[0093]

[0094] It should be noted that in order to ensure the certainty and scientificity of the balance coefficient λ and the weighting coefficient γ in the first measurement method, we need to provide a clear algorithm to determine these coefficients. The following is a specific process:

[0095] Collect multiple sample datasets, each containing point cloud data from different perspectives and actual measurements as a benchmark.

[0096] Initially set the values ​​of λ and γ. For example, λ = 0.5 and γ = 1 as a starting point.

[0097] An iterative optimization algorithm (such as gradient descent or genetic algorithm) is used to adjust the values ​​of λ and γ. In each iteration:

[0098] Apply the current λ and γ values ​​and follow the steps in Method 1 to calculate the object length L for each data set. final .

[0099] Calculate L final The error between the measured value and the actual value is used as the optimization target.

[0100] Use mean square error (MSE) as the evaluation indicator:

[0101]

[0102] Where n is the number of samples, L final,i is the calculated length, and L actual,i is the actual length.

[0103] λ and γ are adjusted through an iterative process to minimize the MSE.

[0104] After each iteration, λ and γ are updated according to the rules of the optimization algorithm.

[0105] When the predetermined number of iterations is reached or the error change is less than a certain threshold, the iteration is stopped.

[0106] The final determined λ and γ are a set of values ​​that minimize the MSE.

[0107] Through this data-driven approach, the values ​​of λ and γ are determined to ensure the effectiveness and accuracy of the algorithm. This method takes into account data from multiple different scenarios to ensure that the parameters are universal and adaptable, thereby improving the accuracy and reliability of the algorithm in practical applications.

[0108] The second measurement method includes standardizing the point cloud data P of each viewing angle;

[0109] Use feature extraction algorithm to extract key points K from each view angle i i and the eigenvector V i ;

[0110] For each pair of key points P A and P B , calculate the Euclidean distance L in each view i i , expressed as,

[0111]

[0112] Among them, x Ai ,y Ai ,z Ai and x Bi ,y Bi ,z Bi They are the key points P A and P B Coordinates in the i-th viewing angle;

[0113] Calculate all viewing distance values ​​L i The standard deviation σ is expressed as,

[0114]

[0115] in, All L i The average value of , n is the number of viewing angles;

[0116] Calculate the confidence S for each view i , expressed as,

[0117]

[0118] Among them, κ is a positive proportionality coefficient;

[0119] Calculate the weight W for each view i It is expressed as,

[0120]

[0121] Among them, S i is the confidence of other perspectives;

[0122] Calculate the final object length L final It is expressed as,

[0123]

[0124] The calculation process for determining κ includes,data collection, collecting multiple sample data sets, each data set containing point cloud data collected from different,viewing angles and the actual object length as a reference value.

[0125] Parameter initialization, initially setting the value of κ, the preferred setting in the present invention is κ=1.

[0126] An iterative optimization algorithm is used to adjust the value of κ. In each iteration:

[0127] Apply the current κ value and calculate the object length L for each data set according to the steps of the second measurement method. final .

[0128] Calculate L final The error between the measured value and the actual value is used as the optimization target.

[0129] Use mean square error MSE as the evaluation indicator:

[0130]

[0131] Where n is the number of samples, L final,i is the calculated length, and L actual,i is the actual length.

[0132] κ is adjusted through an iterative process to minimize the MSE.

[0133] After each iteration, κ is updated according to the rules of the optimization algorithm.

[0134] When the predetermined number of iterations is reached or the error change is less than a certain threshold, the iteration is stopped.

[0135] The final κ is the value that minimizes the MSE.

[0136] In this way, an optimal κ value is determined to ensure that the confidence calculation can effectively reflect the measurement differences between different viewpoints. This data-driven optimization method ensures that the selection of κ is both statistically based and adaptable to practical applications.

[0137] Both methods improve the accuracy and reliability of measurement by fusing multiple views. The first measurement method focuses on the coverage and point cloud density of the view, which is suitable for situations where the point cloud data quality and coverage are uneven. The second measurement method emphasizes measurement consistency and is suitable for scenarios where the measurement results vary greatly under different view angles. Through these methods, we can combine information from different view angles to provide more accurate and comprehensive measurement results.

[0138] It should be noted that before evaluating the point cloud density and perspective consistency, two precisely calibrated cameras are deployed to capture environmental images in parallel to create a stereoscopic visual effect. Ensure that the position and angle of the camera are calibrated to obtain images with high spatial consistency. Stereo image matching technology is used to process the images captured by the dual cameras to generate a depth map. The depth value of each pixel represents the distance between the corresponding environmental point and the camera. The depth map is converted into three-dimensional point cloud data to accurately reflect the spatial position of the object. Three-dimensional reconstruction technology is used to optimize the accuracy and density of the point cloud. Edge detection is performed on the located objects to clarify their boundaries and key feature points. Image processing technology is used to accurately extract the outline of the object.

[0139] The first measurement method focuses on measuring based on view coverage and point cloud density through weighted averaging, which can provide more reliable measurement results when the point cloud data quality is uneven. The second measurement method emphasizes the consistency of the measurement results and processes the data in a confidence-weighted manner to reduce the impact of view differences on the final measurement results. This method is particularly suitable for environments with high noise or poor data consistency.

[0140] Example 2

[0141] Reference Figure 2 , as an embodiment of the present invention, provides a stereoscopic vision length dimension accurate measurement system, comprising:

[0142] Evaluation module 100 and measurement module 200;

[0143] The evaluation module 100 is used to evaluate the point cloud density and the perspective consistency; determine the first measurement method or the second measurement method based on the point cloud density and the perspective consistency;

[0144] The measuring module 200 measures the length of the object.

[0145] The present invention adopts a dual camera system to obtain stereoscopic visual images from multiple perspectives, generates depth maps and three-dimensional point cloud data, and combines a multi-perspective fusion method to reduce measurement errors and improve measurement accuracy.

[0146] By evaluating the density and perspective consistency of point cloud data, the present invention can automatically select a measurement method suitable for the current data situation. For uneven point cloud data, a density-based weighted average method is used; for poor perspective consistency, a confidence weighted method is used to reduce the impact of noise and data inconsistency.

[0147] The present invention effectively solves the problems of large single-view measurement error, noise affecting measurement accuracy, and uneven point cloud data by introducing multi-view data fusion and adaptive weighted processing, thereby improving the accuracy and reliability of measurement.

[0148] Example 3

[0149] An embodiment of the present invention is different from the first two embodiments in that:

[0150] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0151] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0152] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0153] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0154] Example 4

[0155] As an embodiment of the present invention, a method for accurately measuring length dimensions in stereoscopic vision is provided. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation experiments.

[0156] The measurement accuracy and consistency of the multi-view weighted fusion measurement method in the present invention and the traditional single-view measurement method under different conditions are compared.

[0157] The test objects are industrial parts with complex geometric shapes (the present invention uses workpieces with multiple curved surfaces and concave and convex details) with a size range of between 100 mm and 500 mm.

[0158] The traditional method uses a single-view measurement system to measure the length of parts, and only measures the point cloud data from one perspective without multi-view fusion.

[0159] The method of the present invention uses a multi-view measurement system to collect point cloud data from three different viewpoints, respectively, adopts a weighted fusion method to integrate the data, and performs measurement consistency analysis and weighting.

[0160] Scan the parts in three dimensions from multiple angles to obtain point cloud data. Use the traditional method to measure the length of a single view. Use the method of the present invention to collect data from multiple viewpoints, calculate the density, coverage and consistency of each viewpoint, and use the weighted average method to obtain the final length measurement value. Measure the length of the real object as a reference value.

[0161] The experimental scenarios include scenario 1: high-density point cloud data with sufficient viewing angle coverage. Scenario 2: medium-density point cloud data with low point cloud density in some viewing angles. Scenario 3: low-density point cloud data with insufficient viewing angle coverage and noise interference. The experimental results are shown in Table 1.

[0162] Table 1 Comparison of experimental results

[0163]

[0164]

[0165] In scenario 1, both the method of the present invention and the traditional method have small errors, but the error of the method of the present invention is lower, showing better measurement accuracy. In scenario 2, the error of the traditional method increases, while the method of the present invention significantly reduces the error and maintains a high accuracy through multi-view weighted fusion. In scenario 3, the error of the traditional method increases significantly due to data sparsity and noise, while the method of the present invention eliminates noise data through perspective consistency evaluation and controls the error within a low range.

[0166] Through experimental comparison, the method of the present invention has shown better measurement accuracy and consistency in different density and noise environments, especially in low-density and noise interference scenes, the error is reduced by 1.5% to 2%. The traditional single-view measurement method has a significant decrease in measurement accuracy when the point cloud density is uneven or there is noise, and the advantages of multi-view weighted fusion are particularly prominent in these scenes.

[0167] Scene 1: High-density point cloud data with sufficient viewing angle coverage. In this scene, the point cloud data is of high quality and has sufficient coverage. This corresponds to the situation mentioned in the first measurement method where the point cloud data is dense and the viewing angle coverage is good. The experimental results show that the method of the present invention has higher accuracy and less error than the traditional method, indicating that the advantages of the method of the present invention are more obvious under the conditions of high point cloud density and sufficient viewing angle coverage.

[0168] In scene 2, the density of point cloud data is medium, and the density of point cloud is low in some viewpoints. In this scene, the density of point cloud data in some viewpoints is low, and there is a certain degree of unevenness in the measured data. This scene is very suitable for the first measurement method, because the first measurement method uses weighted averaging to deal with the problem of uneven point cloud density, thereby reducing the measurement error caused by uneven data quality. The experimental results show that the error of the method of the present invention in this scene is significantly lower than that of the traditional method, proving that the weighted fusion method is effective in dealing with uneven data quality.

[0169] Scene 3: Low-density point cloud data, insufficient viewing angle coverage, and noise interference. In this scene, the point cloud data is sparse and noisy, resulting in large differences in measurement results under different viewing angles. The second measurement method can be applied in this scene. Through measurement consistency analysis, confidence weighting is used to reduce the influence of noise and viewing angle differences. The experimental results show that the method of the present invention significantly reduces the measurement error when the noise interference is large and the viewing angle coverage is insufficient, indicating that the second measurement method has advantages in measurement consistency processing under low-quality data.

[0170] In scenario 1, when the data quality is high and the coverage is sufficient, the basic weighted average method is sufficient, so there is no need for complex error correction, and the difference between the first and second measurement methods is not obvious.

[0171] In scenario 2, the data quality is uneven, so the first measurement method is suitable. The point cloud density is weighted to reduce the impact of uneven data and improve measurement accuracy.

[0172] Scene 3 has large noise and obvious differences in measurement results. It is suitable to use the second measurement method, which reduces the measurement errors under different viewing angles by confidence weighting, especially showing its advantages under low-density data and noise conditions.

[0173] The experiment demonstrates the adaptability of the method of the present invention in various data scenarios through comparisons under different conditions. In particular, when the data quality is poor or the noise is large, the method of the present invention has a significantly greater advantage than the traditional single-view measurement method. The data in the description directly verifies the applicability of the first measurement method and the second measurement method for different measurement scenarios, as well as the superiority of the present invention in improving measurement accuracy and consistency.

[0174] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for accurately measuring length dimensions using stereoscopic vision, characterized in that: include: Evaluate point cloud density and view consistency; The first measurement method or the second measurement method is determined based on the point cloud density and the viewing angle consistency to measure the length size of the object.

2. The method for accurately measuring the length dimension of stereoscopic vision according to claim 1, characterized in that: The point cloud density is expressed as, in, Indicates point position, V voxel Indicates volume.

3. The method for accurately measuring the length dimension of stereoscopic vision according to claim 2, characterized in that: The perspective consistency includes measuring a first feature point and a second feature point on the same object through any two perspectives; The coordinates of the first feature point are A1=(x A1 ,y A1 ,z A1 )), (B1=(x B1 ,y B1 ,z B1 ); the coordinates of the second feature point are A2=(x A2 ,y A2 ,z A2 )), (B2=(x B2 ,y B2 ,z B2 ).

4. The method for accurately measuring the length dimension of stereoscopic vision according to claim 3, characterized in that: The perspective consistency is expressed as, ΔL=|L1-L2|.

5. The method for accurate measurement of length dimension using stereoscopic vision according to claim 4, characterized in that: The determining of the first measurement method or the second measurement method includes selecting the first measurement method when D satisfies a preset threshold and ΔL satisfies a preset threshold; When D does not meet the preset threshold or ΔL does not meet the preset threshold, the second measurement method is selected.

6. The method for accurate measurement of length dimension using stereoscopic vision according to claim 5, characterized in that: The first measurement method includes The point cloud data P of each perspective is standardized and expressed as: Among them, μ and σ are the mean and standard deviation of the point cloud data respectively; Use feature extraction algorithm to extract key points K from each view i i and the eigenvector V i ; For each pair of key points P A and P B , calculate the Euclidean distance L in each view i i , expressed as, Define the quality Q of each view i , combined with the coverage of the viewing angle C i and point cloud density D i , expressed as, Among them, C i and D i are the coverage rate and point cloud density of the i-th viewing angle, respectively, and λ represents the balance coefficient; Calculate the weight W for each view i It is expressed as, Among them, γ is the weighting coefficient; Calculate the final object length L final It is expressed as, 7. The method for accurately measuring the length dimension of stereoscopic vision according to claim 6, characterized in that: The second measurement method includes standardizing the point cloud data P of each viewing angle; Use feature extraction algorithm to extract key points K from each view angle i i and the eigenvector V i ; For each pair of key points P A and P B , calculate the Euclidean distance L in each view i i , expressed as, Among them, x Ai ,y Ai ,z Ai and x Bi ,y Bi ,z Bi They are the key points P A and P B Coordinates in the i-th perspective; Calculate all viewing distance values ​​L i The standard deviation σ is expressed as, in, All L i The average value of , n is the number of viewing angles; Calculate the confidence S for each view i , expressed as, Among them, κ is a positive proportionality coefficient; Calculate the weight W for each view i It is expressed as, Among them, S i is the confidence of other perspectives; Calculate the final object length L final It is expressed as, 8. A system using the stereoscopic vision length dimension accurate measurement method as claimed in any one of claims 1 to 7, characterized in that: include: An evaluation module (100) and a measurement module (200); The evaluation module (100) is used to evaluate the point cloud density and the perspective consistency; and determine the first measurement method or the second measurement method based on the point cloud density and the perspective consistency; The measuring module (200) measures the length dimension of an object.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for accurately measuring the length dimension of stereoscopic vision described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for accurately measuring the length dimension of stereoscopic vision described in any one of claims 1 to 7 are implemented.

Images