Full-automatic laser bag opening system

Abstract

The invention relates to the technical field of laser processing and intelligent manufacturing, in particular to a full-automatic laser bag opening system. Through the three-dimensional visual recognition module and the multi-dimensional image feature extraction technology and in combination with the support vector machine radial basis accounting method, high-accuracy recognition of bag body material types is achieved. A material identification result directly influences laser parameter self-adaptive matching; according to the method, the cutting track and the energy density distribution are optimized according to the constructed two-dimensional Gaussian laser energy distribution model. By dynamically adjusting the output power and the residence time, the problems of deformation, carbonization, incomplete cutting and the like caused by thermal diffusion are effectively inhibited, and the continuity and uniformity of a cutting boundary are improved; the multispectral imaging technology is introduced to detect the cut area in real time, closed-loop feedback and an intelligent quality judgment mechanism in the cutting process are achieved, and a technical basis is provided for subsequent reprocessing or abnormal alarm.

Description

Technical Field

[0001] The present invention relates to the field of laser processing and intelligent manufacturing technology, and in particular to a fully automatic laser bag opening system. Background Art

[0002] With the continuous development of intelligent manufacturing technology, laser processing, due to its advantages such as high precision, non-contact, high speed, and high degree of automation, is widely used in industrial scenarios such as material cutting, punching, marking, and welding. In particular, in the packaging, textile, medical, and logistics fields, the demand for fine processing of bags made of flexible materials (such as plastic film, paper, and non-woven fabrics) continues to increase, prompting the evolution of laser processing towards high stability, intelligent recognition, and adaptive control.

[0003] Traditional laser cutting systems rely heavily on manual calibration and static path planning, making them difficult to adapt to the complex and varied shapes of flexible bags made of diverse materials. To achieve truly flexible production and automated applications, it is imperative to deeply integrate laser processing technology with 3D visual recognition, material sensing, intelligent path planning, and online quality inspection to build an intelligent laser processing system with highly autonomous decision-making capabilities.

[0004] In view of the above problems, it is necessary to propose a fully automatic laser bag opening system. Summary of the Invention

[0005] The purpose of the present invention is to solve the problems existing in the background technology and to provide a fully automatic laser bag opening system.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] The present invention provides a fully automatic laser bag opening system, which comprises a three-dimensional visual positioning module, a laser path simulation module, a high-precision laser execution module and a real-time quality detection module.

[0008] The 3D visual positioning and recognition module accesses an industrial camera to capture images of the bag and uses the images to identify the bag material, bag edges, and bag opening boundaries. It acquires 3D cloud point data of the bag fabric in real time and obtains the spatial coordinates of all cloud points on the bag opening boundary.

[0009] As a preferred embodiment of the present invention, an image of the bag captured by an industrial camera is acquired to locate the bag opening boundary and identify the bag material.

[0010] The specific process of bag opening boundary positioning is as follows:

[0011] With the industrial camera as the origin (0, 0, 0), a three-dimensional rectangular coordinate system with mutually perpendicular x, y, and z axes is established. Image recognition is used to capture a preset number of bag edge coordinate points {(xi, yi, zi)|i=1,2,3,n}, where i is the sequential number of the coordinate points and n is the preset number of bag edge coordinate points to capture.

[0012] As a preferred embodiment of the present invention, all captured bag edge coordinate points are outputted through the display screen, and all continuous coordinate points selected by the user are recorded as the bag opening boundary, i.e., the target point for subsequent laser bag opening procedure.

[0013] The specific process of bag material identification is as follows:

[0014] Obtain the area domain S formed by all coordinate points on the bag surface, and obtain the three-channel pixel values ​​of all coordinate points (x, y, z) in the area domain S on the three color channels of red R, green G, and blue B, including the red channel pixel value R(x, y, z), the green channel pixel value G(x, y, z), and the blue channel pixel value B(x, y, z).

[0015] As a preferred embodiment of the present invention, texture features and reflection features are extracted from the three-channel pixel values ​​of each coordinate point, and the preset bag material recognition results are matched through supervised learning based on SVM support vector machine. The specific process is as follows:

[0016] The RGB three-channel pixel values ​​of each coordinate point are fused into a grayscale value I (x, y, z) for texture feature extraction. The fusion formula is:

[0017] I(x,y,z)=0.299×R(x,y,z)+0.587×G(x,y,z)+0.114×B(x,y,z).

[0018] By default formula:

[0019] For each coordinate point (x, y, z) in the area domain S, feature extraction of energy E, contrast C and entropy D is performed in the neighborhood of the area domain S, and the energy companion matrix E(x, y, z), contrast companion matrix C(x, y, z) and entropy companion matrix D(x, y, z) of each coordinate point (x, y, z) are obtained, where (x, y, z) belongs to the area domain S.

[0020] The neighborhood of the coordinate point (x, y, z) is defined as the intersection of a three-dimensional sphere with a preset radius and the area domain S.

[0021] Among them, P I(x, y, z) represents the probability of occurrence of coordinate points with the same grayscale value as (x, y, z) in the area domain S; where I1 and I2 are grayscale indices, P I1,I2 (x, y, z) is the probability that grayscale I1 and I2 co-occur in the neighborhood of the coordinate point (x, y, z) in the area domain S. ε is a preset positive decimal used to avoid the occurrence of 0 in the logarithmic operation.

[0022] Among them, each coordinate point (x, y, z) has an energy companion matrix, a contrast companion matrix and an entropy companion matrix with respect to all coordinate points in the area domain S.

[0023] The energy half-life matrix E(x, y, z), contrast companion matrix C and entropy companion matrix D(x, y, z) of all coordinate points are input into the SVM support vector machine for radial basis kernel classification and matched to the preset bag material recognition results.

[0024] The bag material identification result is specifically a set of preset bag material determination symbols j.

[0025] Specifically, the value of the bag material discriminator j is 1, which represents paper;

[0026] The value of the bag material identifier j is 2, which represents plastic film;

[0027] The value of the bag material identifier j is 3, which represents an aluminum-plastic composite material;

[0028] The value of the bag material identifier j is 4, which represents a polymer composite material;

[0029] The value of the bag material identifier j is 5, which represents a coated woven material;

[0030] The value of the bag material identifier j is 6, which represents textile fabric material;

[0031] The value of the bag material identifier j is 7, which represents non-woven fabric material.

[0032] The radial basis kernel classification formula of SVM support vector machine is as follows:

[0033]

[0034] Among them, f j (x, y, z) is the bag material identification prediction result of bag material j calculated for (x, y, z);

[0035] When f j When the value of (x, y, z) is 1, it means that the bag material determined based on the neighborhood of (x, y, z) is j;

[0036] When the value is f jWhen the value of (x, y, z) is 0, it means that it is impossible to determine whether the bag material is j based on the neighborhood of (x, y, z);

[0037] When f j When the value of (x, y, z) is -1, it means that the bag material determined based on the neighborhood of (x, y, z) is not j;

[0038] where j = 1, 2, 3, 4, 5, 6, 7;

[0039] The sign function outputs +1 when the value in the sign function is positive, -1 when the value in the sign function is negative, and 0 when the value in the sign function is 0.

[0040] Among them, α j is the Lagrange multiplier corresponding to the trained support vector; b is the bias term obtained through training, T(x, y, z) is the input feature set, and T(x, y, z) = [E(x, y, z), C(x, y, z), D(x, y, z)] T ; where exp(-γ||T(x,y,z)-T j || 2 ) is a radial kernel function, which calculates the input feature set T and the preset feature set T obtained through training for each bag material discriminator j. j The similarity of γ is the preset influence factor.

[0041] The radial basis kernel classification of the SVM support vector machine traverses all coordinate points (x, y, z) to obtain the recognition prediction results f of each coordinate point for different bag material discriminators j j (x, y, z). By pre-set formula Calculate the overall recognition prediction result Fj of the coordinate points (x, y, z) on the entire bag body with respect to each different bag material discriminator j. Select the maximum value in Fj and extract its corresponding bag material discriminator j as the final recognition result of the entire bag material.

[0042] The laser path simulation module captures the user-selected bag boundary as the target point for the subsequent laser bag opening process. It obtains the coordinates of each coordinate point within the bag boundary. Based on the visual positioning data and bag material identification results, it generates the laser cutting intensity distribution with the optimal energy density.

[0043] Get all coordinate points Pk=(xk, yk, zk) on all bag opening boundaries;

[0044] Obtain the final recognition result of the entire bag material, that is, the bag material identifier j corresponding to the maximum value in Fj, call the preset thermal conductivity, density, specific heat capacity and laser absorption coefficient, and match it to the corresponding preset laser intensity.

[0045] For each coordinate point on the boundary of the bag, a two-dimensional Gaussian light intensity model is used:

[0046] A cutting area prediction model is established for the coordinate points (xk, yk, zk) on the boundaries of each pocket. Here, x, y, and z are the x-axis, y-axis, and z-axis components of the estimated cutting range of a laser with an intensity of I(xk, yk, zk) emitted at point (xk, yk, zk).

[0047] The optimization goal of establishing a two-dimensional Gaussian light intensity model is to ensure penetrating cutting of the bag boundary under the premise of minimum energy consumption.

[0048] The optimization objective function is: Where E(xk, yk, zk) is the laser energy per unit area received by the coordinate point (xk, yk, zk) on the bag boundary; tk is the residence time of the laser at the coordinate point (xk, yk, zk) on the bag boundary; d0 is the preset material thickness, and d(xk, yk, zk) is the predicted thickness of the coordinate point (xk, yk, zk) on the bag boundary.

[0049] Taking the optimization objective function as the optimization goal, the laser residence time tk and laser emission intensity I(xk, yk, zk) of the coordinate points (xk, yk, zk) on the boundary of each bag opening are solved to satisfy the comprehensive minimum of I(xk, yk, zk).

[0050] The high-precision laser execution module generates laser path control parameters according to the laser dwell time and laser emission intensity of the coordinate points on the bag opening boundary to perform high-precision laser cutting.

[0051] Get the path set L formed by the coordinate points (xk, yk, zk) on the bag boundary.

[0052] Retrieve the preset laser galvanometer dynamic equation:

[0053]

[0054] where θ x，y，z (t) is the rotation angle of the galvanometer at time t about the x, y, and z axes, corresponding to the laser polarization position; J is the system moment of inertia. C is the damping coefficient, K is the stiffness coefficient, and K c is the current torsion constant; u(t) is the control voltage at time t.

[0055] Where xk(t), yk(t), and zk(t) are the cumulative times from the start of laser cutting until it reaches the point (xk, yk, zk). fx, fy, and fz are the focal lengths of the galvanometer about the x, y, and z axes, respectively.

[0056] Substitute the laser dwell time tk of the coordinate point (xk, yk, zk) on the bag boundary, solve the rotation control angle and control voltage u(t) of the galvanometer control angle about the x, y, and z axes at time t, and send them to the laser galvanometer control device to complete the laser cutting of the bag boundary.

[0057] The real-time quality inspection module uses multispectral imaging to detect the quality of the cutting seam, evaluate the cutting results, and the deformation of the bag edge caused by laser heat diffusion during the cutting process.

[0058] After the cutting is completed, the detection light with a preset wavelength is emitted to the coordinate point (xk, yk, zk) on the boundary of the bag mouth, and the spectrum after laser burning is extracted and the spectral characteristics of each coordinate point (xk, yk, zk) after cutting are analyzed.

[0059] After the cutting is completed, the spectrum Sk(λk) after laser burning and the reflected light wavelength λk of all coordinate points (xk, yk, zk) on the bag opening boundary are obtained, and the maximum value λmax and the minimum value λmin of the reflected light wavelength are obtained.

[0060] By preset formula Calculate the spectral mean μk, spectral standard deviation σk and spectral average wavelength Ek of the cutting area.

[0061] When it is identified that the spectrum average wavelength Ek belongs to the preset value range and the spectrum mean μk and the spectrum standard deviation σk are both less than the preset threshold, the bag boundary cutting result is judged to be qualified;

[0062] When it is identified that the spectrum average wavelength Ek belongs to the preset value range, the spectrum mean μk is less than or equal to the preset threshold, and the spectrum standard deviation σk is less than the preset threshold, it is determined that the bag edge cutting result has burrs;

[0063] When it is identified that the spectrum average wavelength Ek does not belong to the preset value range, it is determined that the bag opening boundary cutting result is carbonized or not cut through.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] 1. This invention uses a three-dimensional visual recognition module and multi-dimensional image feature extraction technology, combined with a support vector machine radial basis kernel algorithm, to achieve highly accurate identification of bag material types. The material identification results directly affect the adaptive matching of laser parameters, achieving personalized laser cutting strategies for different materials, significantly improving cutting quality and efficiency.

[0066] 2. This invention constructs a two-dimensional Gaussian laser energy distribution model to optimize the cutting trajectory and energy density distribution. By dynamically adjusting the output power and dwell time, it effectively suppresses deformation, carbonization, and incomplete cutting caused by thermal diffusion, and improves the continuity and uniformity of the cutting edge.

[0067] 3. This invention incorporates multispectral imaging technology to perform real-time inspection of the cut area, extracting the reflectance spectrum curve at each coordinate point and accurately distinguishing between acceptable cuts, burrs, carbonization, and incomplete cuts based on judgment rules. This achieves closed-loop feedback and intelligent quality judgment during the cutting process, providing a technical foundation for subsequent reprocessing or abnormality alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings:

[0069] Figure 1 is a system block diagram of the present invention;

[0070] Figure 2 Schematic diagram of identifying the bag body boundary and bag opening edge proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0072] See also Figure 1 As shown, a fully automatic laser bag opening system includes a three-dimensional visual positioning module, a laser path simulation module, a high-precision laser execution module and a real-time quality detection module.

[0073] The 3D visual positioning and recognition module accesses an industrial camera to capture images of the bag and uses the images to identify the bag material, bag edges, and bag opening boundaries. It acquires 3D cloud point data of the bag fabric in real time and obtains the spatial coordinates of all cloud points on the bag opening boundary.

[0074] See also Figure 2As shown, the bag image captured by the industrial camera is obtained to locate the bag opening boundary and identify the bag material.

[0075] The specific process of bag opening boundary positioning is as follows:

[0076] With the industrial camera as the origin (0, 0, 0), a three-dimensional rectangular coordinate system with mutually perpendicular x, y, and z axes is established. Image recognition is used to capture a preset number of bag edge coordinate points {(xi, yi, zi)|i=1,2,3,n}, where i is the sequential number of the coordinate points and n is the preset number of bag edge coordinate points to capture.

[0077] Furthermore, all captured bag edge coordinate points are output through the display screen, and all continuous coordinate points selected by the user are recorded as the bag opening boundary, which is the target point for the subsequent laser bag opening program.

[0078] The specific process of bag material identification is as follows:

[0079] Obtain the area domain S formed by all coordinate points on the bag surface, and obtain the three-channel pixel values ​​of all coordinate points (x, y, z) in the area domain S on the three color channels of red R, green G, and blue B, including the red channel pixel value R(x, y, z), the green channel pixel value G(x, y, z), and the blue channel pixel value B(x, y, z).

[0080] Furthermore, the texture features and reflection features of the three-channel pixel values ​​of each coordinate point are extracted, and the preset bag material recognition results are matched through supervised learning based on the SVM support vector machine. The specific process is as follows:

[0081] The RGB three-channel pixel values ​​of each coordinate point are fused into a grayscale value I (x, y, z) for texture feature extraction. The fusion formula is:

[0082] I(x,y,z)=0.299×R(x,y,z)+0.587×G(x,y,z)+0.114×B(x,y,z).

[0083] By default formula:

[0084] For each coordinate point (x, y, z) in the area domain S, feature extraction of energy E, contrast C and entropy D is performed in the neighborhood of the area domain S, and the energy companion matrix E(x, y, z), contrast companion matrix C(x, y, z) and entropy companion matrix D(x, y, z) of each coordinate point (x, y, z) are obtained, where (x, y, z) belongs to the area domain S.

[0085] The neighborhood of the coordinate point (x, y, z) is defined as the intersection of a three-dimensional sphere with a preset radius and the area domain S.

[0086] Among them, P I (x, y, z) represents the probability of occurrence of coordinate points with the same grayscale value as (x, y, z) in the area domain S; where I1 and I2 are grayscale indices, P I1,I2 (x, y, z) is the probability that grayscale I1 and I2 co-occur in the neighborhood of the coordinate point (x, y, z) in the area domain S. ε is a preset positive decimal used to avoid the occurrence of 0 in the logarithmic operation.

[0087] Among them, each coordinate point (x, y, z) has an energy companion matrix, a contrast companion matrix and an entropy companion matrix with respect to all coordinate points in the area domain S.

[0088] The energy half-life matrix E(x, y, z), contrast companion matrix C and entropy companion matrix D(x, y, z) of all coordinate points are input into the SVM support vector machine for radial basis kernel classification and matched to the preset bag material recognition results.

[0089] The bag material identification result is specifically a set of preset bag material determination symbols j.

[0090] Specifically, the value of the bag material discriminator j is 1, which represents paper;

[0091] The value of the bag material identifier j is 2, which represents plastic film;

[0092] The value of the bag material identifier j is 3, which represents an aluminum-plastic composite material;

[0093] The value of the bag material identifier j is 4, which represents a polymer composite material;

[0094] The value of the bag material identifier j is 5, which represents a coated woven material;

[0095] The value of the bag material identifier j is 6, which represents textile fabric material;

[0096] The value of the bag material identifier j is 7, which represents non-woven fabric material.

[0097] The radial basis kernel classification formula of SVM support vector machine is as follows:

[0098]

[0099] Among them, f j (x, y, z) is the bag material identification prediction result of bag material j calculated for (x, y, z);

[0100] When f jWhen the value of (x, y, z) is 1, it means that the bag material determined based on the neighborhood of (x, y, z) is j;

[0101] When the value is f j When the value of (x, y, z) is 0, it means that it is impossible to determine whether the bag material is j based on the neighborhood of (x, y, z);

[0102] When f j When the value of (x, y, z) is -1, it means that the bag material determined based on the neighborhood of (x, y, z) is not j;

[0103] where j = 1, 2, 3, 4, 5, 6, 7;

[0104] The sign function outputs +1 when the value in the sign function is positive, -1 when the value in the sign function is negative, and 0 when the value in the sign function is 0.

[0105] Among them, α j is the Lagrange multiplier corresponding to the trained support vector; b is the bias term obtained through training, T(x, y, z) is the input feature set, and T(x, y, z) = [E(x, y, z), C(x, y, z), D(x, y, z)] T ; where exp(-γ||T(x,y,z)-T j || 2 ) is a radial kernel function, which calculates the input feature set T and the preset feature set T obtained through training for each bag material discriminator j. j The similarity of γ is the preset influence factor.

[0106] The radial basis kernel classification of the SVM support vector machine traverses all coordinate points (x, y, z) to obtain the recognition prediction results f of each coordinate point for different bag material discriminators j j (x, y, z). By pre-set formula Calculate the overall recognition prediction result Fj of the coordinate points (x, y, z) on the entire bag body with respect to each different bag material discriminator j. Select the maximum value in Fj and extract its corresponding bag material discriminator j as the final recognition result of the entire bag material.

[0107] The laser path simulation module captures the user-selected bag boundary as the target point for the subsequent laser bag opening process. It obtains the coordinates of each coordinate point within the bag boundary. Based on the visual positioning data and bag material identification results, it generates the laser cutting intensity distribution with the optimal energy density.

[0108] Get all coordinate points Pk=(xk, yk, zk) on all bag opening boundaries;

[0109] Obtain the final recognition result of the entire bag material, that is, the bag material identifier j corresponding to the maximum value in Fj, call the preset thermal conductivity, density, specific heat capacity and laser absorption coefficient, and match it to the corresponding preset laser intensity.

[0110] For each coordinate point on the boundary of the bag, a two-dimensional Gaussian light intensity model is used:

[0111] A cutting area prediction model is established for the coordinate points (xk, yk, zk) on the boundaries of each pocket. Here, x, y, and z are the x-axis, y-axis, and z-axis components of the estimated cutting range of a laser with an intensity of I(xk, yk, zk) emitted at point (xk, yk, zk).

[0112] The optimization goal of establishing a two-dimensional Gaussian light intensity model is to ensure penetrating cutting of the bag boundary under the premise of minimum energy consumption.

[0113] The optimization objective function is: Where E(xk, yk, zk) is the laser energy per unit area received by the coordinate point (xk, yk, zk) on the bag boundary; tk is the residence time of the laser at the coordinate point (xk, yk, zk) on the bag boundary; d0 is the preset material thickness, and d(xk, yk, zk) is the predicted thickness of the coordinate point (xk, yk, zk) on the bag boundary.

[0114] Taking the optimization objective function as the optimization goal, the laser residence time tk and laser emission intensity I(xk, yk, zk) of the coordinate points (xk, yk, zk) on the boundary of each bag opening are solved to satisfy the comprehensive minimum of I(xk, yk, zk).

[0115] The high-precision laser execution module generates laser path control parameters according to the laser dwell time and laser emission intensity of the coordinate points on the bag opening boundary to perform high-precision laser cutting.

[0116] Get the path set L formed by the coordinate points (xk, yk, zk) on the bag boundary.

[0117] Retrieve the preset laser galvanometer dynamic equation:

[0118]

[0119] where θ x，y，z (t) is the rotation angle of the galvanometer at time t about the x, y, and z axes, corresponding to the laser polarization position; J is the system moment of inertia. C is the damping coefficient, K is the stiffness coefficient, and K cis the current torsion constant; u(t) is the control voltage at time t.

[0120] Where xk(t), yk(t), and zk(t) are the cumulative times from the start of laser cutting until it reaches the point (xk, yk, zk). fx, fy, and fz are the focal lengths of the galvanometer about the x, y, and z axes, respectively.

[0121] Substitute the laser dwell time tk of the coordinate point (xk, yk, zk) on the bag boundary, solve the rotation control angle and control voltage u(t) of the galvanometer control angle about the x, y, and z axes at time t, and send them to the laser galvanometer control device to complete the laser cutting of the bag boundary.

[0122] The real-time quality inspection module uses multispectral imaging to detect the quality of the cutting seam, evaluate the cutting results, and the deformation of the bag edge caused by laser heat diffusion during the cutting process.

[0123] After the cutting is completed, the detection light with a preset wavelength is emitted to the coordinate point (xk, yk, zk) on the boundary of the bag mouth, and the spectrum after laser burning is extracted and the spectral characteristics of each coordinate point (xk, yk, zk) after cutting are analyzed.

[0124] After the cutting is completed, the spectrum Sk(λk) after laser burning and the reflected light wavelength λk of all coordinate points (xk, yk, zk) on the bag opening boundary are obtained, and the maximum value λmax and the minimum value λmin of the reflected light wavelength are obtained.

[0125] By preset formula Calculate the spectral mean μk, spectral standard deviation σk and spectral average wavelength Ek of the cutting area.

[0126] When it is identified that the spectrum average wavelength Ek belongs to the preset value range and the spectrum mean μk and the spectrum standard deviation σk are both less than the preset threshold, the bag boundary cutting result is judged to be qualified;

[0127] When it is identified that the spectrum average wavelength Ek belongs to the preset value range, the spectrum mean μk is less than or equal to the preset threshold, and the spectrum standard deviation σk is less than the preset threshold, it is determined that the bag edge cutting result has burrs;

[0128] When it is identified that the spectrum average wavelength Ek does not belong to the preset value range, it is determined that the bag opening boundary cutting result is carbonized or not cut through.

[0129] It should be understood that the terms “include” and “comprising” used in the specification and claims of the present disclosure indicate the presence of described features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0130] It should also be understood that the terms used in this disclosure are for the purpose of describing particular embodiments only and are not intended to limit the disclosure. As used in this disclosure and the claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in this disclosure and the claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations;

[0131] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A fully automatic laser bag opening system, including a three-dimensional visual positioning recognition module, a light path simulation module and a high-precision laser execution module, characterized in that ； The 3D visual positioning and recognition module accesses the industrial camera to capture images of the bag and uses the images to identify the bag material, bag edge, and bag opening boundary. It also acquires the 3D cloud point data of the bag fabric in real time and obtains the spatial coordinates of all cloud points on the bag opening boundary. It also obtains the bag image captured by the industrial camera to locate the bag opening boundary and identify the bag material. The laser path simulation module obtains the bag opening boundary selected by the user as the target point for the subsequent laser bag opening process; obtains the coordinates of each coordinate point in the bag opening boundary; and generates the laser cutting intensity distribution with the optimal energy density based on the visual positioning data and the identification results of the bag material; The high-precision laser execution module generates laser path control parameters according to the laser dwell time and laser emission intensity of the coordinate points on the bag opening boundary to perform high-precision laser cutting.

2. The fully automatic laser bag opening system according to claim 1, characterized in that: Also includes real-time quality detection module: The real-time quality inspection module uses multispectral imaging to detect the quality of the cutting seam, evaluate the cutting results, and the deformation of the bag edge caused by laser heat diffusion during the cutting process.

3. The fully automatic laser bag opening system according to claim 1, characterized in that: The specific process of bag opening boundary positioning is as follows: With the industrial camera as the origin (0, 0, 0), a three-dimensional rectangular coordinate system with mutually perpendicular x, y, and z axes is established. Image recognition is used to capture a preset number of coordinate points {(xi, yi, zi)|i=1,2,3,n} on the edge of the bag. Here, i is the sequential number of the coordinate points, and n is the preset number of bag edge coordinate points to capture. The display screen outputs all captured bag edge coordinate points, and records all continuous coordinate points selected by the user as the bag opening boundary, which is the target point for the subsequent laser bag opening process.

4. The fully automatic laser bag opening system according to claim 1, characterized in that: The specific process of bag material identification is as follows: Obtain the area domain S formed by all coordinate points on the bag surface, and obtain the three-channel pixel values ​​of all coordinate points (x, y, z) in the area domain S on the three color channels of red R, green G, and blue B, including the red channel pixel value R(x, y, z), the green channel pixel value G(x, y, z), and the blue channel pixel value B(x, y, z); The texture and reflection features of the three-channel pixel values ​​at each coordinate point are extracted, and the preset bag material recognition results are matched through supervised learning based on the SVM support vector machine. The specific process is as follows: The RGB three-channel pixel values ​​of each coordinate point are fused into a grayscale value I (x, y, z) for texture feature extraction. The fusion formula is: I(x,y,z)=0.299×R(x,y,z)+0.587×G(x,y,z)+0.114×B(x,y,z); By default formula: For each coordinate point (x, y, z) in the area domain S, the energy E, contrast C and entropy D features are extracted in the neighborhood of the area domain S, and the energy companion matrix E(x, y, z), contrast companion matrix C(x, y, z) and entropy companion matrix D(x, y, z) of each coordinate point (x, y, z) are obtained, where (x, y, z) belongs to the area domain S; Define the neighborhood of the coordinate point (x, y, z) as the intersection of a three-dimensional sphere with a preset radius and the area domain S; Among them, P I (x, y, z) represents the probability of occurrence of coordinate points with the same grayscale value as (x, y, z) in the area domain S; where I1 and I2 are grayscale indices, P I1,I2 (x, y, z) is the probability that grayscale I1 and I2 co-occur in the neighborhood of the coordinate point (x, y, z) in the area domain S; where ε is a preset positive decimal used to avoid the occurrence of 0 in the logarithmic operation; Among them, each coordinate point (x, y, z) has an energy companion matrix, a contrast companion matrix, and an entropy companion matrix with respect to all coordinate points in the area domain S; The energy half-life matrix E(x, y, z), contrast companion matrix C and entropy companion matrix D(x, y, z) of all coordinate points are input into the SVM support vector machine for radial basis kernel classification and matched to the preset bag material recognition results.

5. The fully automatic laser bag opening system according to claim 4, characterized in that: The specific process of performing radial basis kernel classification and matching the preset bag material recognition results is as follows: Get the preset bag material identifier j; The value of the bag material identifier j is 1, which represents paper; The value of the bag material identifier j is 2, which represents plastic film; The value of the bag material identifier j is 3, which represents an aluminum-plastic composite material; The value of the bag material identifier j is 4, which represents a polymer composite material; The value of the bag material identifier j is 5, which represents a coated woven material; The value of the bag material identifier j is 6, which represents textile fabric material; The value of the bag material identifier j is 7, which represents non-woven fabric material; The radial basis kernel classification formula of SVM support vector machine is as follows: Among them, f j (x, y, z) is the bag material identification prediction result of bag material j calculated for (x, y, z); When f j When the value of (x, y, z) is 1, it means that the bag material determined based on the neighborhood of (x, y, z) is j; When the value is f j When the value of (x, y, z) is 0, it means that it is impossible to determine whether the bag material is j based on the neighborhood of (x, y, z); When f j When the value of (x, y, z) is -1, it means that the bag material determined based on the neighborhood of (x, y, z) is not j; where j = 1, 2, 3, 4, 5, 6, 7; The sign function outputs +1 when the value in the sign function is positive; -1 when the value in the sign function is negative; and 0 when the value in the sign function is 0. Among them, α j is the Lagrange multiplier corresponding to the trained support vector; b is the bias term obtained through training, T(x, y, z) is the input feature set, and T(x, y, z) = [E(x, y, z), C(x, y, z), D(x, y, z)] T ; where exp(-γ||T(x,y,z)-T j || 2 ) is a radial kernel function, which calculates the input feature set T and the preset feature set T obtained through training for each bag material discriminator j. j The similarity of ; where γ is the preset impact factor; The radial basis kernel classification of the SVM support vector machine traverses all coordinate points (x, y, z) to obtain the recognition prediction results f of each coordinate point for different bag material discriminators j j (x, y, z); by preset formula Calculate the overall recognition prediction result Fj of the coordinate points (x, y, z) on the entire bag body with respect to each different bag material discriminator j; select the maximum value in Fj and extract its corresponding bag material discriminator j as the final recognition result of the entire bag material.

6. The fully automatic laser bag opening system according to claim 1, characterized in that: The specific process of generating the laser cutting intensity distribution with the optimal energy density is as follows: Get all coordinate points Pk=(xk, yk, zk) on all bag opening boundaries; Obtain the final identification result of the entire bag material, that is, the bag material identifier j corresponding to the maximum value in Fj, retrieve the preset thermal conductivity, density, specific heat capacity and laser absorption coefficient, and match it to the corresponding preset laser intensity; For each coordinate point on the boundary of the bag, a two-dimensional Gaussian light intensity model is used: Establish a cutting area prediction model for the coordinate points (xk, yk, zk) on the boundary of each pocket; where x, y, and z are the x-axis, y-axis, and z-axis components of the estimated cutting range of a laser with an intensity of I(xk, yk, zk) emitted at the point (xk, yk, zk); The optimization goal of establishing a two-dimensional Gaussian light intensity model is to ensure penetrating cutting of the bag edge under the premise of minimum energy consumption; The optimization objective function is: Where E(xk, yk, zk) is the laser energy per unit area received by the coordinate point (xk, yk, zk) on the bag boundary; tk is the dwell time of the laser at the coordinate point (xk, yk, zk) on the bag boundary; d0 is the preset material thickness, and d(xk, yk, zk) is the predicted thickness at the coordinate point (xk, yk, zk) on the bag boundary; Taking the optimization objective function as the optimization goal, the laser residence time tk and laser emission intensity I(xk, yk, zk) of the coordinate points (xk, yk, zk) on the boundary of each bag opening are solved to satisfy the comprehensive minimum of I(xk, yk, zk).

7. The fully automatic laser bag opening system according to claim 1, characterized in that: The specific process of generating laser path control parameters is as follows: Get the path set L formed by the coordinate points (xk, yk, zk) on the bag boundary; Retrieve the preset laser galvanometer dynamic equation: where θ x，y，z (t) is the rotation control angle of the galvanometer at time t about the x, y, and z axes, corresponding to the laser polarization position; J is the system moment of inertia; C is the damping coefficient, K is the stiffness coefficient, and K c is the current torsion constant; u(t) is the control voltage at time t; Where xk(t), yk(t), and zk(t) are the cumulative times from the start of laser cutting until it reaches the point (xk, yk, zk); fx, fy, and fz are the focal lengths of the galvanometer about the x, y, and z axes, respectively; Substitute the laser dwell time tk of the coordinate point (xk, yk, zk) on the bag boundary, solve the rotation control angle and control voltage u(t) of the galvanometer control angle about the x, y, and z axes at time t, and send them to the laser galvanometer control device to complete the laser cutting of the bag boundary.

8. The fully automatic laser bag opening system according to claim 2, characterized in that: The specific process of evaluating the cutting results is: After the cutting is completed, the detection light with a preset wavelength is emitted to the coordinate point (xk, yk, zk) on the boundary of the bag opening, and the spectrum after laser ablation is extracted and the spectrum characteristics of each coordinate point (xk, yk, zk) after cutting are analyzed; Obtain the spectrum Sk(λk) after laser ablation at all coordinate points (xk, yk, zk) on the bag boundary after cutting, as well as the reflected light wavelength λk therein, and obtain the maximum value λmax and the minimum value λmin of the reflected light wavelength; By preset formula Calculate the spectral mean μk, spectral standard deviation σk and spectral average wavelength Ek of the cutting area; When it is identified that the spectrum average wavelength Ek belongs to the preset value range and the spectrum mean μk and the spectrum standard deviation σk are both less than the preset threshold, the bag boundary cutting result is judged to be qualified; When it is identified that the spectrum average wavelength Ek belongs to the preset value range, the spectrum mean μk is less than or equal to the preset threshold, and the spectrum standard deviation σk is less than the preset threshold, it is determined that the bag edge cutting result has burrs; When it is identified that the spectrum average wavelength Ek does not belong to the preset value range, it is determined that the bag opening boundary cutting result is carbonized or not cut through.

Images