Intelligent detection method and system for parts based on three-dimensional machine vision

Abstract

The present application relates to the field of image processing technology, and in particular to a method and system for intelligent detection of parts based on three-dimensional machine vision, the method comprising: calculating the horizontal complexity and vertical complexity of each data point in the standard point cloud of the parts, each data point corresponding to at least one scanning posture; calculating the scanning complexity of each scanning posture based on the horizontal complexity and vertical complexity of each data point; in the process of scanning along a preset path, determining the scanning parameters of each scanning posture according to the scanning complexity, and obtaining the scanning point cloud of the parts to be detected; after the scanning point cloud and the standard point cloud are point cloud registered, the data points in the scanning point cloud with the registration error greater than the error threshold are marked as defective points. Through the technical solution of the present application, accurate point cloud data of parts can be obtained, and accurate quality inspection results can be obtained.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method and system for intelligent detection of parts based on three-dimensional machine vision. Background Art

[0002] In the field of mechanical manufacturing, it is often necessary to conduct quality inspections on finished parts to ensure that they meet the size requirements. Due to the different three-dimensional structures of different parts, it is impossible to get an accurate judgment result by judging whether the parts meet the size requirements on a two-dimensional image.

[0003] At present, a patent application document with publication number CN116539619A discloses a product defect detection method, system, device and storage medium, wherein the method includes: obtaining point cloud data, audio data and image data of the product to be tested; inputting the point cloud data, audio data and image data into a first defect detection model to obtain a first prediction result; the first prediction result is used to characterize the probability that the product to be tested is a defective product based on the multimodal data of the product to be tested; inputting the point cloud data into a second defect detection model to obtain a second prediction result; inputting the image data into a third defect detection model to obtain a third prediction result; weighting the first prediction result, the second prediction result and the third prediction result to obtain a detection classification result of the product to be tested; the detection classification result is used to characterize whether the product to be tested is a defective product or a normal product.

[0004] The above method combines point cloud data, audio data and image data to realize product defect detection. However, it ignores the influence of the structure of components on defect detection. For components with complex structures, accurate point cloud data cannot be collected, which leads to inaccurate quality inspection results of components. Summary of the invention

[0005] In order to solve the technical problem of inaccurate quality inspection results of parts and components, the present application provides a method and system for intelligent inspection of parts and components based on three-dimensional machine vision, which can obtain accurate point cloud data of parts and components, and thus obtain accurate quality inspection results.

[0006] In a first aspect of the present application, a method for intelligent detection of parts based on three-dimensional machine vision is provided, the detection method comprising: calculating the horizontal complexity of each data point in any elevation plane of a standard point cloud of a part, the standard point cloud comprising a plurality of data points, one data point corresponding to at least one scanning posture, the scanning posture comprising a scanning height and a scanning angle; obtaining a vertical plane perpendicular to the elevation plane where the data point is located, and calculating the vertical complexity of the data point on the vertical plane; calculating the scanning complexity of each scanning posture based on the horizontal complexity and vertical complexity of each data point, comprising: calculating the sum of the squares of the horizontal complexity and the vertical complexity as the size complexity of the data point; calculating the number of centers of all data points in any scanning posture According to the Euclidean distance from each data point corresponding to the scanning posture to the central data point, a weight coefficient of each data point is calculated, and the weight coefficient is negatively correlated with the Euclidean distance; according to the weight coefficient, the size complexity of each data point is weighted and summed to obtain the scanning complexity of the scanning posture; in the process of scanning along the preset path, the scanning parameters of each scanning posture are determined according to the scanning complexity to obtain the scanning point cloud of the component to be inspected, and the scanning parameters include scanning density and scanning speed, and the scanning density is positively correlated with the scanning complexity, and the scanning speed is negatively correlated with the scanning complexity; after the scanning point cloud and the standard point cloud are aligned, the data points in the scanning point cloud whose alignment error is greater than the error threshold are marked as defective points.

[0007] Preferably, the data points in the scan point cloud The error threshold for:

[0008] , is the error reference value, For data points The number of point clouds within the neighborhood, is the average number of point clouds in the neighborhood of each data point in the scanned point cloud.

[0009] The larger the number of point clouds within the neighborhood of a data point, the greater the dimensional complexity within the neighborhood of the data point. To ensure that the size of the parts meets the requirements, a smaller error threshold is assigned to the data points with greater dimensional complexity, and the dimensional requirements for the data points with greater dimensional complexity are increased to ensure the accuracy of the detection results in the dimensional complexity area.

[0010] Preferably, the data points The calculation method of the horizontal complexity includes: calculating the data points Elevation plane Gradient direction of each data point in; Get the data point The horizontal neighboring points of The horizontal complexity of ; where the horizontal neighboring points include the data points And data points The distance from the data point in the elevation plane Most recent data points.

[0011] Horizontal complexity can accurately reflect the size change of data points in the elevation plane. The greater the horizontal complexity, the more data points There are large size changes.

[0012] Preferably, the data points The level of complexity Satisfies the relationship: , is the gradient direction, and the value range of the gradient direction is arrive , is the gradient direction in the horizontal adjacent points frequency.

[0013] Preferably, the data points The size complexity of for: , For data points The level of complexity, For data points The vertical complexity of .

[0014] Preferably, the data points The weight coefficient satisfy: , is the preset coefficient, For data points Euclidean distance to the central data point.

[0015] Preferably, the scanning posture The scanning complexity for: , Scanning pose The corresponding number of data points, is the scanning height, is the scanning angle, For data points The weight coefficient of For data points The size complexity.

[0016] The horizontal complexity and vertical complexity of each data point in the standard point cloud are converted into the scanning complexity of each scanning posture. The scanning complexity is used to quantify the difficulty of obtaining accurate point cloud data under the scanning posture.

[0017] Preferably, the scanning parameters of each scanning posture are determined according to the scanning complexity, including: scanning posture Scanning density satisfy:

[0018] , is the minimum scanning density, is the maximum scanning density, is the density adjustment coefficient; scanning posture Scanning speed satisfy: , is the maximum scanning speed, is the minimum scanning speed, is the speed adjustment coefficient, Scanning pose The scanning complexity.

[0019] Scanning pose The greater the scanning complexity, the greater the scanning density, the lower the scanning speed, and the smaller the scanning posture. The longer the dwell time, the more Sufficient data points can be collected to improve the accuracy of the scanned point cloud.

[0020] Preferably, determining the scanning parameters of each scanning posture according to the scanning complexity includes: obtaining the real-time scanning parameters of the previous scanning posture; scanning posture Scanning density satisfy:

[0021] , is the minimum scanning density, is the maximum scanning density, is the density adjustment coefficient, is the real-time scan density of the previous scan pose, Express Round up; scan pose Scanning speed satisfy: , is the maximum scanning speed, is the minimum scanning speed, is the speed adjustment coefficient, is the real-time scanning speed of the previous scanning pose, Scanning pose The scanning complexity, is the smoothing coefficient.

[0022] In order to prevent sudden changes in scanning parameters from causing damage to the laser radar and reducing its service life, when determining the scanning parameters of each scanning posture based on the scanning complexity, it is also necessary to consider the scanning parameters of the previous scanning posture in the preset path. By introducing a smoothing coefficient, smooth control of the scanning parameters can be achieved.

[0023] In the second aspect of the present application, there is also provided a system for intelligent detection of parts based on three-dimensional machine vision, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent detection method of parts based on three-dimensional machine vision according to the first aspect of the present application is implemented.

[0024] The technical solution of this application has the following beneficial technical effects:

[0025] Before inspecting the parts to be inspected, obtain the standard point cloud of the parts to be inspected, in which each data point in the standard point cloud corresponds to at least one scanning posture, so as to realize the mapping from the data point to the scanning posture; further, calculate the horizontal complexity and vertical complexity of the data point on the elevation plane and the vertical plane perpendicular to the elevation plane respectively, and calculate the scanning complexity of the corresponding scanning posture according to the horizontal complexity and vertical complexity of the data point; when the scanning complexity is large, increase the scanning density and reduce the scanning speed to increase the density of the scanning point cloud, ensure that more data points can be collected at the data points with large complexity, and improve the accuracy of the scanning point cloud; after further performing point cloud registration between the scanning point cloud and the standard point cloud, mark the data points in the scanning point cloud with the registration error greater than the error threshold as defective points to obtain accurate quality inspection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easy to understand. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0027] Figure 1 is a flow chart of a component intelligent detection method based on three-dimensional machine vision according to an embodiment of the present application;

[0028] Figure 2 is a schematic diagram of a laser radar according to an embodiment of the present application;

[0029] Figure 3 is a schematic diagram of a standard point cloud, a scanned point cloud, and a point cloud registration result according to an embodiment of the present application;

[0030] Figure 4 It is a structural block diagram of a component intelligent detection system based on three-dimensional machine vision according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] 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 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 the present application.

[0032] It should be understood that when the terms "first", "second", etc. are used in the claims, specification and drawings of the present application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections.

[0033] According to the first aspect of the present application, the present application provides a method for intelligent detection of parts based on three-dimensional machine vision, which is used to perform quality inspection on the finished parts to determine whether the finished parts meet the size. During the detection process, the parts to be detected are first placed on the detection platform, and the detection platform is provided with a positioning groove for fixing the parts to be detected on the detection platform at a set angle; then, at a scanning angle, the laser radar is used to scan all elevation planes of the parts to be detected from top to bottom or from bottom to top in sequence, and then the detection platform is rotated to change the scanning angle and scan again until the 360-degree scan of the parts to be detected is completed, and the scanning point cloud of the parts to be detected is obtained.

[0034] That is to say, in the process of scanning the parts to be inspected, the laser radar moves in a direction perpendicular to the elevation plane to collect data points of all elevation planes at any scanning angle, and changes the scanning angle by rotating the detection platform; with the mutual cooperation of the detection platform and the laser radar, the collection of the scanning point cloud is completed.

[0035] Figure 1 FIG. 1 is a flow chart of a method for intelligent detection of components based on three-dimensional machine vision according to an embodiment of the present application. Figure 1 As shown, the intelligent component detection method based on three-dimensional machine vision includes steps S101 to S105, which are described in detail below.

[0036] S101, calculating the horizontal complexity of each data point in an arbitrary elevation plane of a standard point cloud of a component, wherein the standard point cloud includes a plurality of data points, one data point corresponds to at least one scanning posture, and the scanning posture includes a scanning height and a scanning angle.

[0037] In one embodiment, the standard point cloud is the point cloud data of a normal component. Specifically, a normal component is placed on a detection platform, and the scanned point cloud collected at the maximum scanning density and the minimum scanning speed is the standard point cloud of the component. The standard point cloud includes multiple data points, each of which corresponds to at least one scanning posture, and under this scanning posture, the scanning result of the corresponding data point can be collected; the scanning posture includes a scanning height and a scanning angle, wherein the scanning height is the scanning height of the laser radar, and the scanning angle is the rotation angle of the detection platform.

[0038] At the same time, each data point corresponds to a three-dimensional coordinate , corresponding to the horizontal coordinate value, vertical coordinate value and elevation value respectively; one elevation value corresponds to an elevation plane, and the elevation plane is the XY plane. For example, the elevation plane Includes elevation values All data points of .

[0039] Data Points The calculation method of the horizontal complexity includes: calculating the data points Elevation plane Gradient direction of each data point in; Get the data point The horizontal neighboring points of The horizontal complexity of ; where the horizontal neighboring points include the data points And data points The distance from the data point in the elevation plane Most recent data points.

[0040] Among them, the data points The level of complexity Satisfies the relationship:

[0041] , is the gradient direction, and the value range of the gradient direction is arrive , is the gradient direction in the horizontal adjacent points frequency. is the maximum value of information entropy in the gradient direction, which is used to constrain horizontal complexity. A value between 0 and 1.

[0042] Understandably, when When the value is 4, it means that the horizontal adjacent points contain 5 data points in total. If the gradient directions of the 5 data points are consistent, it means that the data points The horizontal adjacent points of form a straight line, and there is no size change. At this time, the data point The level of complexity The value is 0; if the gradient directions of the five data points are quite different, it means that the data points The horizontal adjacent points of form a complex curve, which has a large size change. At this time, the data point The level of complexity is a larger value.

[0043] In this way, the horizontal complexity of each data point in the standard point cloud of the component is obtained, and the horizontal complexity is used to reflect the degree of size change of the data point in the elevation plane. At the same time, the scanning posture corresponding to each data point is obtained, under which the scanning result of the corresponding data point can be collected, and one data point corresponds to at least one scanning posture.

[0044] S102, obtaining a vertical plane perpendicular to the elevation plane where the data point is located, and calculating the vertical complexity of the data point on the vertical plane.

[0045] In one embodiment, after obtaining the horizontal complexity of the data point on the elevation plane, it is also necessary to calculate the vertical complexity of the data point on the vertical plane, where the vertical complexity is used to reflect the size change of the data point in the vertical direction.

[0046] Passing data point Make a vertical plane The vertical plane , and the data points In elevation plane Understandably, the data points There are multiple vertical planes , just need to get any vertical plane Get the vertical plane Then, you can Internal calculation data points The vertical complexity of .

[0047] Specifically, the data points The vertical complexity of the calculation method includes: calculating the data points Vertical plane Gradient direction of each data point in; Get the data point The vertical neighboring points of the vertical neighboring points are used as data points. The vertical complexity of ; where the vertical adjacent points include the data points And data points Vertical plane Distant data points Most recent data points. The value of is 4.

[0048] Data Points Vertical complexity Calculation formula and data points Horizontal complexity The calculation formula is the same and will not be repeated here.

[0049] In this way, the horizontal complexity and vertical complexity of each data point in the standard point cloud are obtained. The horizontal complexity is used to reflect the degree of size change of the data point in the elevation plane, and the vertical complexity is used to reflect the degree of size change of the data point in the vertical plane, thereby achieving accurate quantification of the degree of size change.

[0050] S103, calculating the scanning complexity of each scanning posture based on the horizontal complexity and vertical complexity of each data point.

[0051] In one embodiment, the scanning complexity of each scanning posture is calculated based on the horizontal complexity and vertical complexity of each data point, including: calculating the sum of the squares of the horizontal complexity and the vertical complexity as the size complexity of the data point; calculating the central data point of all data points of any scanning posture, and calculating the weight coefficient of each data point based on the Euclidean distance from each data point corresponding to the scanning posture to the central data point, and the weight coefficient is negatively correlated with the Euclidean distance; weighted summation of the size complexity of each data point according to the weight coefficient to obtain the scanning complexity of the scanning posture.

[0052] Among them, the data points The size complexity of for: , For data points The level of complexity, For data points The vertical complexity of .

[0053] The central data point is the average value of all data points corresponding to the scanning posture, and the central data point includes a horizontal coordinate center value, a vertical coordinate center value, and an elevation center value. Exemplarily, the horizontal coordinate center value is the average value of the horizontal coordinate values ​​of all data points corresponding to the scanning posture.

[0054] Among them, the data points The weight coefficient satisfy: , is the preset coefficient, For data points The Euclidean distance to the central data point; the default coefficient value is 1.

[0055] Among them, the scanning posture The scanning complexity for: , Scanning pose The corresponding number of data points, is the scanning height, is the scanning angle, For data points The weight coefficient of For data points The size complexity.

[0056] In this way, the horizontal complexity and vertical complexity of each data point in the standard point cloud are converted into the scanning complexity of each scanning posture. The scanning complexity is used to quantify the difficulty of obtaining accurate point cloud data under the scanning posture.

[0057] S104, during scanning along a preset path, the scanning parameters of each scanning posture are determined according to the scanning complexity to obtain a scanning point cloud of the component to be inspected. The scanning parameters include scanning density and scanning speed. The scanning density is positively correlated with the scanning complexity, and the scanning speed is negatively correlated with the scanning complexity.

[0058] In one embodiment, the parts to be inspected are scanned along a preset path, and the preset path is: the parts to be inspected are fixed on the inspection platform at a set angle, the scanning angle at this time is recorded as 0, the scanning height of the laser radar is adjusted from top to bottom or from bottom to top, and the scanning results of each scanning height when the scanning angle is 0 are obtained; the inspection platform is rotated to adjust the scanning angle, and the scanning results of each scanning height under each scanning angle are obtained, thereby completing the collection of the scanning point cloud.

[0059] Specifically, the scanning parameters of each scanning posture determined according to the scanning complexity include: Scanning density satisfy:

[0060] , is the minimum scanning density, is the maximum scanning density, is the density adjustment coefficient; scanning posture Scanning speed satisfy:

[0061] , is the maximum scanning speed, is the minimum scanning speed, is the speed adjustment coefficient, Scanning pose The scanning complexity.

[0062] Among them, the value of the density adjustment coefficient is 2, and the value of the speed adjustment coefficient is 1.

[0063] Among them, the scanning posture The greater the scanning complexity, the greater the scanning density. A large number of data points can be collected and the point cloud data under the scanning posture can be accurately obtained. The greater the scanning complexity, the lower the scanning speed. The longer the dwell time, the more Collect a large number of data points; comprehensively adjust the scanning density and scanning speed to ensure that the scanning posture with greater scanning complexity In this case, enough data points can be obtained to improve the accuracy of the scanned point cloud.

[0064] Understandably, the scan density The value of is a positive integer, see Figure 2 , is a schematic diagram of a laser radar according to an embodiment of the present application, wherein the laser radar is a dot matrix laser, including multiple laser light sources, and the scanning density is adjusted by controlling the number of switches of the laser light sources in the dot matrix laser; the greater the scanning density, the greater the number of laser light sources turned on in the dot matrix laser. In this embodiment, the scanning density is divided into three gears, and different gears correspond to different numbers of switches of the laser light sources.

[0065] In another embodiment, in order to prevent sudden changes in scanning parameters from causing damage to the laser radar and reducing the service life of the laser radar, in the process of determining the scanning parameters of each scanning posture based on the scanning complexity, the scanning parameters of the previous scanning posture in the preset path also need to be considered.

[0066] Specifically, the scanning parameters of each scanning posture are determined according to the scanning complexity, including: obtaining the real-time scanning parameters of the previous scanning posture; scanning posture Scanning density satisfy:

[0067] , is the minimum scanning density, is the maximum scanning density, is the density adjustment coefficient, is the real-time scan density of the previous scan pose, Express Round up; scan pose Scanning speed satisfy:

[0068] , is the maximum scanning speed, is the minimum scanning speed, is the speed adjustment coefficient, is the real-time scanning speed of the previous scanning pose, Scanning pose The scanning complexity, is the smoothing coefficient.

[0069] The smoothing coefficient is set to 0.5.

[0070] In this way, the parts to be inspected are scanned according to the scanning parameters of each scanning posture, and adaptive adjustment of the scanning parameters is achieved. Under the premise of ensuring the accuracy of the scanning point cloud, the scanning efficiency is improved to obtain the scanning point cloud of the parts to be inspected.

[0071] S105, after performing point cloud registration on the scanned point cloud and the standard point cloud, mark the data points in the scanned point cloud whose registration errors are greater than an error threshold as defective points.

[0072] In one embodiment, a point cloud registration algorithm is used to perform point cloud registration on the scanned point cloud and the standard point cloud. The point cloud registration algorithm may be an IPC algorithm (iterative closest point registration algorithm) or an NDT algorithm (normal distribution registration algorithm).

[0073] Scanning data points in point cloud The registration error is: Get data points Neighborhood range, calculate the data point The Euclidean distance between the standard point cloud data point in the neighborhood; the minimum Euclidean distance is taken as the data point The registration error is the neighborhood range of the data points A spherical area with a preset radius and a center of .

[0074] The error threshold is negatively correlated with the number of point clouds within the neighborhood of the data point. After obtaining the scanned point cloud using the method of steps S101 to S104, if the number of point clouds within the neighborhood of the data point is larger, it means that the size complexity within the neighborhood of the data point is larger. In order to ensure that the size of the parts meets the requirements, a smaller error threshold is assigned to the data points with larger size complexity, thereby increasing the size requirements of the data points with larger size complexity. The error threshold for:

[0075] , is the error reference value, For data points The number of point clouds within the neighborhood, is the average number of point clouds in the neighborhood of each data point in the scanned point cloud. The value of is 0.5.

[0076] In this way, data points with registration errors greater than the error threshold are obtained in the scanning point cloud. These data points are defective points with unqualified dimensions, thus realizing intelligent detection of parts.

[0077] For example, see Figure 3 , is a schematic diagram of a standard point cloud, a scanned point cloud, and a point cloud registration result according to an embodiment of the present application. In the point cloud registration result, defect points and normal points are rendered in different colors to intuitively reflect the inspection results of the parts to be inspected.

[0078] The above introduces the technical principle and implementation details of a method for intelligent detection of parts based on three-dimensional machine vision in the present application through specific embodiments. Before detecting the parts to be detected, obtain the standard point cloud of the parts to be detected, and each data point in the standard point cloud corresponds to at least one scanning posture to achieve the mapping of data points to scanning postures; further, calculate the horizontal complexity and vertical complexity of the data points on the elevation plane and the vertical plane perpendicular to the elevation plane respectively, and calculate the scanning complexity of the corresponding scanning posture according to the horizontal complexity and vertical complexity of the data points. When the scanning complexity is large, increase the scanning density and reduce the scanning speed to increase the density of the scanning point cloud, ensure that more data points can be collected at data points with greater complexity, and improve the accuracy of the scanning point cloud; further, after the scanning point cloud and the standard point cloud are point cloud aligned, mark the data points in the scanning point cloud with an alignment error greater than the error threshold as defective points to obtain accurate quality inspection results.

[0079] According to the second aspect of the present application, the present application also provides an intelligent component detection system based on three-dimensional machine vision. Figure 4 FIG. 1 is a block diagram of a component intelligent detection system based on three-dimensional machine vision according to an embodiment of the present application. Figure 4 As shown, the system 50 includes a processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent detection method of parts based on three-dimensional machine vision according to the first aspect of the present application is implemented. The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, whose settings and functions are known in the art, so they are not described here.

[0080] In the present application, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in this application may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0081] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0082] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the patent application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent application shall be subject to the attached claims.

Claims

1. An intelligent detection method for parts based on three-dimensional machine vision, characterized in that: The detection method comprises: Calculate the horizontal complexity of each data point in any elevation plane of the standard point cloud of the component, the standard point cloud includes multiple data points, one data point corresponds to at least one scanning posture, and the scanning posture includes a scanning height and a scanning angle; obtain a vertical plane perpendicular to the elevation plane where the data point is located, and calculate the vertical complexity of the data point on the vertical plane; The scanning complexity of each scanning posture is calculated based on the horizontal complexity and vertical complexity of each data point, including: calculating the sum of the squares of the horizontal complexity and the vertical complexity as the size complexity of the data point; calculating the central data point of all data points of any scanning posture, and calculating the weight coefficient of each data point according to the Euclidean distance from each data point corresponding to the scanning posture to the central data point, wherein the weight coefficient is negatively correlated with the Euclidean distance; weighted summing the size complexity of each data point according to the weight coefficient to obtain the scanning complexity of the scanning posture; During scanning along the preset path, the scanning parameters of each scanning posture are determined according to the scanning complexity to obtain the scanning point cloud of the component to be inspected. The scanning parameters include scanning density and scanning speed. The scanning density is positively correlated with the scanning complexity, and the scanning speed is negatively correlated with the scanning complexity. After performing point cloud registration on the scanned point cloud and the standard point cloud, data points in the scanned point cloud with a registration error greater than an error threshold are marked as defective points; Data Points The calculation method of the horizontal complexity includes: calculating the data points Elevation plane Gradient direction of each data point in; Get the data point The horizontal neighboring points of The horizontal complexity of ; where the horizontal neighboring points include the data points And data points The distance from the data point in the elevation plane Most recent data points.

2. The intelligent detection method of parts based on three-dimensional machine vision according to claim 1 is characterized in that: Scanning data points in point cloud The error threshold for: , is the error reference value, For data points The number of point clouds within the neighborhood, is the average number of point clouds in the neighborhood of each data point in the scanned point cloud.

3. The intelligent detection method of parts based on three-dimensional machine vision according to claim 1 is characterized in that: Data Points The level of complexity Satisfies the relationship: , is the gradient direction, and the value range of the gradient direction is arrive , is the gradient direction in the horizontal adjacent points frequency.

4. The intelligent detection method of parts based on three-dimensional machine vision according to claim 1 is characterized in that: Data Points The size complexity of for: , For data points The level of complexity, For data points The vertical complexity of .

5. The intelligent detection method of parts based on three-dimensional machine vision according to claim 1 is characterized in that: Data Points The weight coefficient satisfy: , is the preset coefficient, For data points Euclidean distance to the central data point.

6. The intelligent detection method of parts based on three-dimensional machine vision according to claim 1 is characterized in that: Scanning pose The scanning complexity for: , Scanning pose The corresponding number of data points, is the scanning height, is the scanning angle, For data points The weight coefficient of For data points The size complexity.

7. The intelligent detection method of parts based on three-dimensional machine vision according to claim 1 is characterized in that: The scanning parameters of each scanning posture determined according to the scanning complexity include: Scanning pose Scanning density satisfy: , is the minimum scanning density, is the maximum scanning density, is the density adjustment coefficient; Scanning pose Scanning speed satisfy: , is the maximum scanning speed, is the minimum scanning speed, is the speed adjustment coefficient, Scanning pose The scanning complexity.

8. The intelligent detection method for parts based on three-dimensional machine vision according to claim 1 is characterized in that: The scanning parameters of each scanning posture determined according to the scanning complexity include: Get the real-time scanning parameters of the previous scanning pose; Scanning pose Scanning density satisfy: , is the minimum scanning density, is the maximum scanning density, is the density adjustment coefficient, is the real-time scan density of the previous scan pose, Express Round up; Scanning pose Scanning speed satisfy: , is the maximum scanning speed, is the minimum scanning speed, is the speed adjustment coefficient, is the real-time scanning speed of the previous scanning pose, Scanning pose The scanning complexity, is the smoothing coefficient.

9. The intelligent detection system of parts based on three-dimensional machine vision is characterized by: The method comprises a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent component detection method based on three-dimensional machine vision according to any one of claims 1 to 8 is implemented.

Images