A full-size detection method and equipment for power battery modules

Abstract

The present invention relates to a full-size detection method and equipment for power battery modules. The method comprises the following steps: collecting images of battery modules on a conveyor belt, and transmitting them to an industrial computer for image data analysis and processing. The image data analysis comprises: comprehensively evaluating light source parameters, mainly through calculation and feedback by a complex lighting evaluation algorithm. If the light adaptability is poor, the angle, posture and brightness of the light source are adjusted and re-evaluated before re-collecting the image. The image data processing comprises: grayscale processing of the module mounting surface, threshold segmentation, edge sub-pixel extraction, and edge fitting to obtain the actual module length, width and mounting hole diameter. Scanning module three-dimensional point cloud information. The steps comprise: filtering and denoising the point cloud information, fitting the point cloud plane, and obtaining the actual height of the module and the flatness of the mounting surface. The present invention has the advantages of high efficiency and good precision, and can meet actual production needs.

Description

Technical Field

[0001] The present invention belongs to the field of power battery module detection, and in particular relates to a full-size detection method and equipment for a power battery module. Background Art

[0002] New energy vehicles are the direction of the new round of scientific and technological revolution and industrial transformation. Developing new energy vehicles is a key component in achieving the transformation and upgrading of my country's automotive industry and achieving leapfrog development. It is also crucial for cultivating new drivers of growth, developing a new economy, and advancing the industry towards mid-to-high-end markets. As a semi-finished product of a power battery pack, the dimensional accuracy of the battery module not only affects the quality of the module itself but also the subsequent processing and assembly of the power battery pack. Therefore, achieving precise online dimensional measurement of battery modules is essential in power battery production. Traditional measurement methods such as vernier calipers, micrometers, and gauges often require manual measurement and cannot meet the high measurement efficiency requirements of battery modules.

[0003] In recent years, with the further development of various technical fields such as computer technology, digital image signal processing technology, and optoelectronic technology, three-dimensional dimension measurement technology based on machine vision has gradually developed. Traditional three-dimensional dimension measurement technology generally adopts contact measurement, the most representative of which is the three-coordinate measuring machine. However, this method has certain bottlenecks: (1) Contact measurement using a probe will inevitably cause certain damage to the surface of the object being measured due to the need for direct contact with the object being measured; (2) Measurement data needs to be collected point by point manually, which has slow measurement speed, low efficiency, poor real-time performance, and low degree of automation; (3) Data sampling density is low, there are measurement blind spots, and it cannot adapt to high-precision measurement of large surfaces. Three-dimensional measurement technology based on machine vision has the characteristics of non-contact, high precision, and high degree of automation. Its measurement process does not directly contact the object being measured and will not cause any damage to the object being measured. It is suitable for various applications with complex scenes and the need for rapid measurement. Therefore, more and more companies are beginning to choose three-dimensional vision measurement technology to measure the geometric dimensions of products. Summary of the Invention

[0004] The purpose of the present invention is to provide a full-size detection method and equipment for power battery modules, which has the advantages of high efficiency and good precision and can meet actual production needs.

[0005] To achieve the above objectives, the technical solution of the present invention is: a full-size detection method for a power battery module, comprising the following steps:

[0006] S1, collect images of battery modules on the conveyor belt and transmit them to the industrial computer;

[0007] S2. The industrial computer performs image data analysis on the image collected in step S1, that is, the global complex illumination adaptability number S is used to evaluate the comprehensive impact of factors including overall light intensity change, local light intensity distribution uniformity, light source illumination angle, and the distance from the light source to the battery module on the full-size measurement of the battery module. If S is less than the threshold value S0, the industrial computer controls the light source to adjust the light intensity and posture, and returns to step S1 to re-collect the image; if S is greater than or equal to the threshold value S0, the process proceeds to step S3;

[0008] S3. The industrial computer processes the image data collected in step S1, namely: grayscale processing, threshold segmentation, edge sub-pixel extraction of the battery module mounting surface, and edge fitting to obtain the actual battery module length, width and mounting hole diameter; then, scan the three-dimensional point cloud information of the battery module, filter and denoise the point cloud information, and fit the point cloud plane to obtain the actual height of the battery module and the flatness of the mounting surface.

[0009] In one embodiment of the present invention, in step S1, a CCD camera is used to capture images of the battery modules on the conveyor belt.

[0010] In one embodiment of the present invention, the specific implementation steps of step S2 are as follows:

[0011] Step S21: For a CCD camera, the relationship between the pixel grayscale value G(x,y) and the light intensity I(x,y) is:

[0012]

[0013] Where x and y are sub-pixel coordinate values;

[0014] Its light intensity I is expressed as

[0015]

[0016] Where I0 is the output light intensity of the lighting source, i is the irradiation angle of the light source, d0 is the distance from the light source to the battery module, ρ d (x, y) is the reflectivity distribution of the battery module surface;

[0017] Step S22: Assume that the image size is M×N, the segmentation threshold of the battery module and the background is denoted as T, and the ratio of the number of pixels belonging to the battery module to the entire image is denoted as ω. a , whose average grayscale is μ a ; The ratio of background pixels to the entire image is ω b , whose average grayscale is μ b The total grayscale of the image is recorded as μ, the inter-class variance is recorded as g; the number of pixels in the image whose grayscale value is less than the threshold T is recorded as N0, and the number of pixels in the image whose grayscale value is greater than the threshold T is recorded as N1, then:

[0018] ω a =N0 / M×N (1)

[0019] ω b =N1 / M×N (2)

[0020] N0+N1=M×N (3)

[0021] ω a +ω b =1 (4)

[0022] μ=ω a *μ a +ω b *μ b (5)

[0023] g=ω a (μ a -μ)^2+ω b (μ b -μ)^2 (6)

[0024] Substituting formula (5) into formula (6), we get the equivalent formula:

[0025] g=ω a ω b (μ a -μ b )^2 (7)

[0026] Step S23: Considering the different characteristics of each measurement area under actual measurement conditions, the local edge contour contrast of the battery module under actual illumination is recorded as C a , the between-class variance is denoted as g a , the surface light reflectivity is recorded as ρ d1 The contrast ratio of the local mounting hole of the battery module is recorded as C b , the between-class variance is denoted as g b , the surface light reflectivity is recorded as ρ d2 ; Light intensity is I1; when the inter-class variance g a 、g b When the maximum values are taken respectively, the most appropriate contrast for identifying the local edge contour length of the battery module is recorded as C1, and the most appropriate contrast for identifying all the circular holes in the battery module is recorded as C2. Then:

[0027] C=∑r(i,j)*r(i,j)*I(i,j)

[0028] Where i, j = 0, 1, 2, 3..., r(i, j) = |ij|, which is the grayscale difference between adjacent pixels; p(i, j) is the pixel distribution probability of the grayscale difference between adjacent pixels being r;

[0029] Define the global complex lighting adaptability number S, The larger S is, the better the light adaptability is. If S≥threshold S0, the light adaptability is good and the process goes to step S3. Otherwise, the data information is fed back to the industrial computer for light compensation and the light intensity is adjusted to Adjust the light source position Return to step S1.

[0030] In one embodiment of the present invention, the threshold S0 is 7.4.

[0031] In one embodiment of the present invention, in step S3, the specific implementation method of filtering and denoising the point cloud information and fitting the point cloud plane is:

[0032] (1) Filter and denoise the point cloud information:

[0033] Perform statistical analysis on the neighborhood of each point in the point cloud information. Assume that the distances of all points in the point cloud form a Gaussian distribution, whose shape is determined by the mean μ and standard deviation σ. Let the coordinates of the nth point in the point cloud be Pn(Xn, Yn, Zn), and the distance from this point to any point Pm(Xm, Ym, Zm) is:

[0034]

[0035] The formula for calculating the average distance between each point and any point is:

[0036]

[0037] The standard deviation is:

[0038]

[0039] Assume that the standard deviation multiple is std, input two thresholds k and std, when the average distance of a point to the k points is within the standard range (μ-σ*std, μ+σ*std), the point is retained, and if it is not within the range, it is defined as an outlier and deleted;

[0040] (2) Fitting point cloud plane:

[0041] After filtering and denoising the point cloud information in step (1), two weights, proximity distance and chromaticity difference, are added. Among them, proximity refers to the distance between the proximity distance and the center point cloud cluster; chromaticity difference refers to the absolute value of the difference between the grayscale of the current point under the influence of noise and the grayscale of the center point; the closer the point is to the center point, the greater its weight coefficient; within the neighborhood, the closer the grayscale value is to the grayscale value of the center point, the greater the weight of the point, and the smaller the weight of the point with a large difference in grayscale value; the two weight coefficients are multiplied together to obtain the final convolution template; the mathematical form of the Gaussian function after the weight coefficient kernel is convolved with the image is:

[0042]

[0043] Where (Xi, Yi) is the current point position, (Xc, Yc) is the center point position, g(Xi, Yi) is the current point grayscale value, and g(Xc, Yc) is the center point grayscale value;

[0044] Then, the point cloud plane fitting process can be realized.

[0045] The present invention also provides a full-size inspection device for power battery modules, including a conveyor belt for transporting battery modules, a 2D camera arranged above the conveyor belt, a robot arranged on one side of the conveyor belt for flipping the battery modules, a light source arranged above the conveyor belt, and two 3D cameras installed on both sides of the conveyor belt.

[0046] In one embodiment of the present invention, the hand of the manipulator is equipped with a 3D camera for identifying the battery module and a vacuum suction cup for sucking the battery module; a gantry is provided above the conveyor belt, a transverse ball screw slide is installed on the gantry, a vertical ball screw slide is installed on the slide of the transverse ball screw slide, the 2D camera is installed on the slide of the vertical ball screw slide, and the light source is connected to both sides of the slide of the vertical ball screw slide through adjustable lamp brackets.

[0047] In one embodiment of the present invention, the 3D camera uses a laser profile sensor, the 2D camera uses a CCD camera, and an industrial computer connected to the 2D camera, the 3D camera, the robotic arm, and the light source is also included. The industrial computer has a built-in program that can implement the method described above.

[0048] Compared with the existing technology, the present invention has the following beneficial effects: the present invention has the advantages of high efficiency, good precision, etc., and can meet the actual production needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Flow chart of the method of the present invention.

[0050] Figure 2 Schematic diagram of the battery module profile scan of the present invention.

[0051] Figure 3 Schematic diagram of the battery module visual measurement device of the present invention. DETAILED DESCRIPTION

[0052] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0053] The present invention provides a full-size detection method for a power battery module, comprising the following steps:

[0054] S1, collect images of battery modules on the conveyor belt and transmit them to the industrial computer;

[0055] S2. The industrial computer performs image data analysis on the image collected in step S1, that is, the global complex illumination adaptability number S is used to evaluate the comprehensive impact of factors including overall light intensity change, local light intensity distribution uniformity, light source illumination angle, and the distance from the light source to the battery module on the full-size measurement of the battery module. If S is less than the threshold value S0, the industrial computer controls the light source to adjust the light intensity and posture, and returns to step S1 to re-collect the image; if S is greater than or equal to the threshold value S0, the process proceeds to step S3;

[0056] S3. The industrial computer processes the image data collected in step S1, namely: grayscale processing, threshold segmentation, edge sub-pixel extraction of the battery module mounting surface, and edge fitting to obtain the actual battery module length, width and mounting hole diameter; then, scan the three-dimensional point cloud information of the battery module, filter and denoise the point cloud information, and fit the point cloud plane to obtain the actual height of the battery module and the flatness of the mounting surface.

[0057] The present invention also provides a full-size inspection device for power battery modules, including a conveyor belt, a 2D camera arranged above the conveyor belt, a robot arranged on one side of the conveyor belt for flipping the battery module, a light source arranged above the conveyor belt, two 3D cameras installed on both sides of the conveyor belt, and an industrial computer connected to the 2D camera, 3D camera, robot arm, and light source, wherein the industrial computer has a built-in program that can implement the method described above.

[0058] The following is a specific implementation process of the present invention.

[0059] like Figure 1 As shown, the present invention provides a full-scale inspection method for power battery modules. This method fully considers the impact of factors such as the complex factory lighting conditions, overall light intensity variations, and local light intensity distribution uniformity on full-scale measurement during the image acquisition process of the battery module image. A mathematical model for multi-scale evaluation of the local light intensity of each measurement area and full-scale measurement accuracy under real lighting conditions is established. This improves the effective accuracy of full-scale inspection and is both timely and applicable. The method includes the following steps:

[0060] Use industrial cameras and high-definition lenses to capture images of the modules on the conveyor belt and transmit them to the industrial computer for image data analysis and processing;

[0061] Image data analysis includes: using the global complex illumination adaptability number S to evaluate the combined impact of factors such as overall light intensity changes, local light intensity distribution uniformity, light source illumination angle, and light source-to-module distance on the full-size measurement of the battery module. If S is less than the threshold S0 (7.4 in this example), the light source is controlled by the PLC to adjust the brightness and posture, and then the image is recaptured; if S is greater than or equal to the threshold S0, proceed to the next step.

[0062] Specifically, the image data analysis steps are as follows:

[0063] 1) For a CCD camera, the relationship between the pixel grayscale value G(x,y) and the light intensity I(x,y) is:

[0064]

[0065] Where x and y are sub-pixel coordinate values;

[0066] Its light intensity I is expressed as

[0067]

[0068] Where I0 is the output light intensity of the lighting source, i is the irradiation angle of the light source, d0 is the distance from the light source to the battery module, ρ d (x, y) is the reflectivity distribution of the battery module surface;

[0069] 2) Assume that the image size is M×N, the segmentation threshold of the battery module and the background is recorded as T, and the ratio of the number of pixels belonging to the battery module to the entire image is recorded as ω a , whose average grayscale is μ a ; The ratio of background pixels to the entire image is ω b , whose average grayscale is μ b The total grayscale of the image is recorded as μ, the inter-class variance is recorded as g; the number of pixels in the image whose grayscale value is less than the threshold T is recorded as N0, and the number of pixels in the image whose grayscale value is greater than the threshold T is recorded as N1, then:

[0070] ω a =N0 / M×N (1)

[0071] ω b =N1 / M×N (2)

[0072] N0+N1=M×N (3)

[0073] ω a +ω b =1 (4)

[0074] μ=ω a *μ a +ω b *μ b (5)

[0075] g=ω a (μ a -μ)^2+ω b (μ b -μ)^2 (6)

[0076] Substituting formula (5) into formula (6), we get the equivalent formula:

[0077] g=ω a ω b (μ a -μ b )^2 (7)

[0078] 3) Considering the different characteristics of each measurement area under actual measurement conditions, the local edge contour contrast of the battery module under actual illumination is recorded as C a , the between-class variance is denoted as g a , the surface light reflectivity is recorded as ρ d1 The contrast ratio of the local mounting hole of the battery module is recorded as C b , the between-class variance is denoted as g b , the surface light reflectivity is recorded as ρ d2 ; Light intensity is I1; when the inter-class variance g a 、g b When the maximum values are taken respectively, the most appropriate contrast for identifying the local edge contour length of the battery module is recorded as C1, and the most appropriate contrast for identifying all the circular holes in the battery module is recorded as C2. Then:

[0079] C=∑r(i,j)*r(i,j)*I(i,j)

[0080] Where i, j = 0, 1, 2, 3..., r(i, j) = |ij|, which is the grayscale difference between adjacent pixels; p(i, j) is the pixel distribution probability of the grayscale difference between adjacent pixels being r;

[0081] Define the global complex lighting adaptability number S, The larger S is, the better the light adaptability is. If S≥threshold S0, the light adaptability is good and the process goes to step S3. Otherwise, the data information is fed back to the industrial computer for light compensation and the light intensity is adjusted to Adjust the light source position Then recapture the image.

[0082] The collected battery module image data is processed, namely: the battery module mounting surface is grayscaled, threshold segmented, and edge sub-pixel extracted, and the edge is fitted to obtain the actual battery module length, width, and mounting hole diameter; then, the battery module three-dimensional point cloud information is scanned, the point cloud information is filtered and denoised, and the point cloud plane is fitted to obtain the actual height of the battery module and the flatness of the mounting surface.

[0083] The specific process of filtering and denoising the point cloud information and fitting the point cloud plane is as follows:

[0084] (1) Filter and denoise the point cloud information:

[0085] Perform statistical analysis on the neighborhood of each point in the point cloud information. Assume that the distances of all points in the point cloud form a Gaussian distribution, whose shape is determined by the mean μ and standard deviation σ. Let the coordinates of the nth point in the point cloud be Pn(Xn, Yn, Zn), and the distance from this point to any point Pm(Xm, Ym, Zm) is:

[0086]

[0087] The formula for calculating the average distance between each point and any point is:

[0088]

[0089] The standard deviation is:

[0090]

[0091] Assume that the standard deviation multiple is std, input two thresholds k and std, when the average distance of a point to the k points is within the standard range (μ-σ*std, μ+σ*std), the point is retained, and if it is not within the range, it is defined as an outlier and deleted;

[0092] (2) Fitting point cloud plane:

[0093] After filtering and denoising the point cloud information in step (1), two weights, proximity distance and chromaticity difference, are added. Among them, proximity refers to the distance between the proximity distance and the center point cloud cluster; chromaticity difference refers to the absolute value of the difference between the grayscale of the current point under the influence of noise and the grayscale of the center point; the closer the point is to the center point, the greater its weight coefficient; within the neighborhood, the closer the grayscale value is to the grayscale value of the center point, the greater the weight of the point, and the smaller the weight of the point with a large difference in grayscale value; the two weight coefficients are multiplied together to obtain the final convolution template; the mathematical form of the Gaussian function after the weight coefficient kernel is convolved with the image is:

[0094]

[0095] Where (Xi, Yi) is the current point position, (Xc, Yc) is the center point position, g(Xi, Yi) is the current point grayscale value, and g(Xc, Yc) is the center point grayscale value;

[0096] Then, the point cloud plane fitting process can be realized.

[0097] like Figure 3As shown, the present invention provides a full-size detection device for power battery modules in response to the above method, including a conveyor belt 100 for conveying battery modules 700, a 2D camera 200 arranged above the conveyor belt, a manipulator 300 arranged on one side of the conveyor belt for flipping the battery module 700, a light source 400 arranged above the conveyor belt, and two 3D cameras 500 installed on both sides of the conveyor belt.

[0098] In one embodiment of the present invention, the hand of the manipulator is equipped with a 3D camera for identifying battery modules and a vacuum suction cup 310 for sucking battery modules; a gantry 600 is provided above the conveyor belt, and a horizontal ball screw slide 610 is installed on the gantry, and a vertical ball screw slide 620 is installed on the slide of the horizontal ball screw slide, and the 2D camera is installed on the slide of the vertical ball screw slide, and the light source is connected to both sides of the slide of the vertical ball screw slide through an adjustable lamp holder 630; the adjustable lamp holder is hinged into a chain shape by a number of connecting rods.

[0099] In one embodiment of the present invention, a 3D camera uses a laser profile sensor, and a 2D camera uses a CCD camera. The invention also includes an industrial computer 800 connected to the 2D camera, the 3D camera, the robotic arm, and the light source. The industrial computer has a built-in program that can implement the method described above. To avoid mutual interference between the two 3D cameras on both sides of the conveyor belt, the two 3D cameras are usually not completely opposite or symmetrical, but are staggered a certain distance to prevent laser interference.

[0100] Two laser profile sensors of the same model and specifications are installed on both sides of the conveyor belt to scan the top and bottom surfaces of the battery module, respectively obtaining point cloud data of the two measurement surfaces for fitting. Then, the edge detection accuracy is improved based on the grayscale distribution characteristics, geometric shape characteristics, and geometric and grayscale coupling characteristics of the power battery module to obtain higher dimensional accuracy. Figure 2 As shown in the figure. Since the data acquired by the two laser profile sensors are independent of each other, each laser profile sensor only needs to scan the measurement surface to obtain complete point cloud data. Then, using an affine relationship, the point cloud data acquired by both laser profile sensors is affine-mapped to the world coordinate system, achieving a unified coordinate system for the point cloud data of the dual laser profile sensors. The point cloud data of the module's top surface is then fitted to derive the top surface equation. Finally, the average distance from the bottom surface point cloud data to the fitted top surface is calculated to obtain the measured module thickness. This measurement method offers the advantages of high efficiency and high accuracy, meeting actual production needs.

[0101] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions and effects do not exceed the scope of the technical solution of the present invention, shall fall within the scope of protection of the present invention.

Claims

1. A full-size detection method for a power battery module, characterized in that: The steps include: S1, collect images of battery modules on the conveyor belt and transmit them to the industrial computer; S2: The industrial computer performs image data analysis on the image collected in S1: The global complex illumination adaptability number S is used to evaluate the comprehensive impact of factors including overall light intensity changes, local light intensity distribution uniformity, light source illumination angle, and the distance from the light source to the battery module on the full-size measurement of the battery module. If S is less than the threshold value S0, the industrial computer controls the light source to adjust the light intensity and posture, and returns to S1 to re-collect the image; if S is greater than or equal to the threshold value S0, the process proceeds to S3. The specific implementation is as follows: S21. For a CCD camera, the relationship between the pixel grayscale value G(x,y) and the light intensity I(x,y) is: x and y are the coordinate values of the sub-pixel; the light intensity I is expressed as I0 is the output light intensity of the lighting source, i0 is the irradiation angle of the light source, d0 is the distance from the light source to the battery module, ρ d (x, y) is the reflectivity distribution of the battery module surface; S22. Assume that the image size is M×N, the segmentation threshold of the battery module and the background is denoted as T, and the proportion of the number of pixels belonging to the battery module in the entire image is denoted as ω a , the average grayscale is μ a ; The ratio of background pixels to the entire image is ω b , the average grayscale is μ b The total grayscale of the image is recorded as μ, the inter-class variance is recorded as g; the number of pixels in the image whose grayscale value is less than the threshold T is recorded as N0, and the number of pixels in the image whose grayscale value is greater than the threshold T is recorded as N1, then: oh a =N0 / (M×N) (1) oh b =N1 / (M×N) (2) N0+N1=M×N (3) oh a +oh b =1 (4) m = oh a *m a +oh b *m b (5) g=ω a (m a -μ)^2+ω b (m b -μ)^2 (6) Substituting formula (5) into formula (6), we get the equivalent formula: g=ω a oh b (m a -m b )^2 (7) S23. Considering the different characteristics of each measurement area under actual measurement conditions, the local edge contour contrast of the battery module under actual illumination is recorded as C a , the between-class variance is denoted as g a , the surface light reflectivity is recorded as ρ d1 The contrast ratio of the local mounting hole of the battery module is recorded as C b , the between-class variance is denoted as g b , the surface light reflectivity is recorded as ρ d2 ; Light intensity is I1; when the inter-class variance g a 、g b When the maximum values are taken respectively, the most appropriate contrast for identifying the local edge contour length of the battery module is recorded as C1, and the most appropriate contrast for identifying all the circular holes in the battery module is recorded as C2. Then: C=∑r(i,j)*r(i,j)*I(i,j) i, j = 0, 1, 2, 3..., n-1, r(i, j) = |ij| is the grayscale difference between adjacent pixels; p(i, j) is the pixel distribution probability of the grayscale difference between adjacent pixels being r; Define the global complex lighting adaptability number S, The larger S is, the better the light adaptability is. If S ≥ threshold S0, the light adaptability is good and enters S3; Otherwise, the data information is fed back to the industrial computer to perform light compensation and adjust the light intensity to Adjust the light source position , return to S1; S3: The industrial computer processes the image data collected in S1: grayscale processing, threshold segmentation, edge sub-pixel extraction, and edge fitting are performed on the battery module installation surface to obtain the actual battery module length, width, and installation hole diameter; then, the three-dimensional point cloud information of the battery module is scanned, the point cloud information is filtered and denoised, and the point cloud plane is fitted to obtain the actual height of the battery module and the flatness of the installation surface.

2. A power battery module full-size detection method according to claim 1, characterized in that: In step S1, a CCD camera is used to capture images of the battery modules on the conveyor belt.

3. A power battery module full-size detection method according to claim 1, characterized in that: The threshold S0 is set to 7.

4.

4. A power battery module full-size detection method according to claim 1, characterized in that: In step S3, the specific implementation method of filtering and denoising the point cloud information and fitting the point cloud plane is as follows: (1) Filter and denoise the point cloud information: Perform statistical analysis on the neighborhood of each point in the point cloud information. Assume that the distances of all points in the point cloud form a Gaussian distribution, whose shape is determined by the mean μ0 and standard deviation σ. Let the coordinates of the nth point in the point cloud be Pn(Xn, Yn, Zn), and the distance from this point to any point Pm(Xm, Ym, Zm) is: The formula for calculating the average distance between each point and any point is: The standard deviation is: Assume that the standard deviation multiple is std, input two thresholds k0 and std, when the average distance of a point close to k0 points is within the standard range (μ0-σ*std, μ0+σ*std), the point is retained, if it is not within the range, it is defined as an outlier and deleted; (2) Fitting point cloud plane: After filtering and denoising the point cloud information in step (1), two weights, proximity and chromaticity difference, are added. Proximity refers to the distance between the proximity distance and the central point cloud cluster; chromaticity difference refers to the absolute value of the difference between the grayscale of the current point under the influence of noise and the grayscale of the central point; the closer the point is to the central point, the greater its weight coefficient; within the neighborhood, the closer the grayscale value is to the grayscale value of the central point, the greater the weight of the point, and the smaller the weight of the point with a large difference in grayscale value; the two weight coefficients are multiplied together to obtain the final convolution template; the mathematical form of the Gaussian function after the weight coefficient kernel is convolved with the image is: Where (Xi, Yi) is the current point position, (Xc, Yc) is the center point position, gray (Xi, Yi) is the gray value of the current point, gray(Xc,Yc) is the gray value of the center point; Then, the point cloud plane fitting process is realized.

5. A full-size testing device for power battery modules, characterized by: It includes a conveyor belt for transporting battery modules, a 2D camera arranged above the conveyor belt, a robot arranged on one side of the conveyor belt for flipping the battery modules, a light source arranged above the conveyor belt, and two 3D cameras installed on both sides of the conveyor belt; it also includes an industrial computer, which has a built-in program that can implement the method as described in any one of claims 1 to 4.

6. The power battery module full-size detection equipment according to claim 5, characterized in that: The hand of the manipulator is equipped with a 3D camera for identifying the battery module and a vacuum suction cup for sucking the battery module; a gantry is provided above the conveyor belt, a horizontal ball screw slide is installed on the gantry, a vertical ball screw slide is installed on the slide of the horizontal ball screw slide, the 2D camera is installed on the slide of the vertical ball screw slide, and the light source is connected to both sides of the slide of the vertical ball screw slide through an adjustable lamp holder.

7. The power battery module full-size detection equipment according to claim 5, characterized in that: The 3D camera uses a laser profile sensor, and the 2D camera uses a CCD camera.

Images