Baking temperature adjusting method, system and equipment for coating production line and medium

Abstract

The invention discloses a baking temperature adjusting method, system and equipment of a coating production line and a medium, and belongs to the technical field of engineering machinery intelligent manufacturing technologies. The baking temperature adjusting method of the coating production line comprises the following steps: acquiring a two-dimensional image of a workpiece suspension area of the coating production line, and generating a workpiece category detection result of each workpiece in the two-dimensional image by using a target detection algorithm model; obtaining a three-dimensional point cloud of the workpiece suspension area, and generating a corresponding size type detection result for each workpiece in the three-dimensional point cloud; determining the total mass of the batch of workpieces according to the workpiece type detection result and / or the size type detection result of each workpiece in the workpiece suspension area; and determining a corresponding target baking temperature according to the total mass of the batch of workpieces, and controlling the coating production line to execute corresponding baking operation according to the target baking temperature. High-precision and automatic baking temperature adjustment can be achieved.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent manufacturing of engineering machinery, and in particular to a method, system, equipment and medium for regulating the baking temperature of a coating production line. Background Art

[0002] In coating production lines, workpieces are baked. In related technologies, the baking temperature for coating production lines is typically set manually based on experience. However, since manual temperature settings cannot be standardized, this leads to energy waste and increased factory emissions.

[0003] Therefore, how to achieve high-precision and automated baking temperature regulation is a technical problem that those skilled in the art currently need to solve. Summary of the Invention

[0004] The purpose of this application is to provide a baking temperature regulation method, system, equipment and medium for a coating production line, which can achieve high-precision and automated baking temperature regulation.

[0005] To solve the above technical problems, the present application provides a method for adjusting the baking temperature of a coating production line, comprising:

[0006] Obtain a two-dimensional image of the workpiece hanging area of ​​the coating production line, and use the target detection algorithm model to generate a workpiece category detection result for each workpiece in the two-dimensional image;

[0007] Acquire a three-dimensional point cloud of the workpiece suspension area, and generate a corresponding size type detection result for each workpiece in the three-dimensional point cloud;

[0008] determining the total mass of the batch of workpieces according to a workpiece category detection result and / or a size type detection result of each workpiece in the workpiece suspension area;

[0009] The corresponding target baking temperature is determined according to the total mass of the workpieces in the batch, and the coating production line is controlled to perform the corresponding baking operation according to the target baking temperature.

[0010] Optionally, determining the total mass of the batch of workpieces according to a workpiece category detection result and / or a size type detection result of each workpiece in the workpiece suspension area includes:

[0011] Selecting a target workpiece from the workpiece hanging area;

[0012] Querying a database for a size type corresponding to a workpiece category detection result of the target workpiece to obtain a size type query result;

[0013] Determining whether the size type detection result of the target workpiece is the same as the size type query result;

[0014] If yes, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type detection result;

[0015] If not, determining a two-dimensional detection confidence level and a three-dimensional detection confidence level of the target workpiece; wherein the two-dimensional detection confidence level is the confidence level of the workpiece category detection result, and the three-dimensional detection confidence level is the confidence level of the size type detection result;

[0016] If the two-dimensional detection confidence is greater than the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type query result;

[0017] If the two-dimensional detection confidence is less than the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the size type detection result;

[0018] If the two-dimensional detection confidence is equal to the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type detection result;

[0019] Determine whether all workpieces in the workpiece hanging area have been selected; if so, calculate the total mass of this batch of workpieces based on the mass of all workpieces in the workpiece hanging area; if not, enter the step of selecting a target workpiece from the workpiece hanging area.

[0020] Optionally, after selecting the target workpiece from the workpiece hanging area, the method further includes:

[0021] If the target workpiece has a corresponding workpiece category detection result, but does not have a corresponding size type detection result, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type query result;

[0022] If a corresponding size type detection result exists for the target workpiece, but no corresponding workpiece category detection result exists, the workpiece quality of the target workpiece is determined according to the size type detection result.

[0023] Optionally, determining the workpiece quality of the target workpiece according to the size type detection result includes:

[0024] Searching a database for candidate workpieces that match the size type detection result;

[0025] The average mass of all the candidate workpieces is set as the workpiece mass of the target workpiece.

[0026] Optionally, before generating the workpiece category detection result of each workpiece in the two-dimensional image using the target detection algorithm model, the method further includes:

[0027] Obtain training samples from a data set; wherein each training sample includes an original sample, a positive sample, and a negative sample, the original sample is an image containing an artifact, the positive sample is an image within the bounding box of the artifact in the original sample, and the negative sample is an image outside the bounding box of the artifact in the original sample;

[0028] Using the training samples to train the target detection algorithm model;

[0029] Among them, the target detection algorithm model includes a skeleton network, a neck network, a detection head, a contrastive learning module and a loss calculation module. The contrastive learning module includes a global average pooling layer, a multi-layer perceptron and a spatial attention layer. The contrastive learning module is used to convert the feature map input by the skeleton network into a positive sample global enhanced feature map.

[0030] Optionally, the process of converting the feature map input by the skeleton network into a positive sample global enhanced feature map by the contrastive learning module includes:

[0031] Receiving the original sample feature map, the positive sample feature map and the negative sample feature map output by the skeleton network;

[0032] Inputting the original sample feature map, the positive sample feature map, and the negative sample feature map into the global average pooling layer to obtain a global channel weight; wherein the global channel weight includes the original sample global channel weight, the positive sample global channel weight, and the negative sample global channel weight;

[0033] Performing a dot multiplication of the original sample feature map and the original sample global channel weight to obtain a global enhanced feature map;

[0034] Inputting the original sample global channel weight, the positive sample global channel weight, and the negative sample global channel weight into a multilayer perceptron to obtain a processing result;

[0035] Inputting the original sample feature map, the positive sample feature map, and the negative sample feature map into the spatial attention layer to obtain spatial local feature weights;

[0036] Performing dot multiplication of the original sample feature map, the positive sample feature map, and the negative sample feature map with the corresponding spatial local feature weights to obtain a local enhanced feature map; wherein the local enhanced feature map includes the original sample local enhanced feature map, the positive sample local enhanced feature map, and the negative sample local enhanced feature map;

[0037] Performing element-wise subtraction operation on the local enhanced feature map of the original sample from the local enhanced feature map of the negative sample to obtain a local information attention feature;

[0038] Calculating a contrast loss function value based on the processing result and the local enhancement feature map, and calculating an approximation loss function value based on the positive sample local enhancement feature map and the local information focus feature, so as to optimize the parameters of the contrastive learning module using the contrast loss function value and the approximation loss function value;

[0039] An element-wise addition operation is performed on the global enhancement feature map and the local information attention feature to obtain a positive sample global enhancement feature map; wherein the positive sample global enhancement feature map is used to input into the neck network.

[0040] Optionally, generating a corresponding size type detection result for each workpiece in the three-dimensional point cloud includes:

[0041] Performing point cloud filtering and segmentation operations on the three-dimensional point cloud to obtain a workpiece point cloud for each workpiece;

[0042] Determining an inclination angle of a point cloud plane equation of each workpiece according to the workpiece point cloud;

[0043] Determining a minimum bounding box of the workpiece point cloud according to the point cloud plane equation, and calculating the volume of the workpiece point cloud according to the minimum bounding box;

[0044] A size type detection result of each workpiece is determined according to the area of ​​the workpiece point cloud, the volume of the workpiece point cloud and the inclination angle of the point cloud plane equation.

[0045] Optionally, the coating production line is provided with a workpiece hanging rack, a three-dimensional laser camera and a plurality of two-dimensional cameras, the workpiece hanging area is the area below the workpiece hanging rack, all the two-dimensional cameras are in the same plane, and the shooting direction of the three-dimensional laser camera is perpendicular to the shooting direction of each of the two-dimensional cameras;

[0046] Accordingly, a two-dimensional image of the workpiece hanging area of ​​the coating production line is obtained, including:

[0047] Using all of the two-dimensional cameras to capture partial images of the workpiece hanging area of ​​the coating production line from multiple directions, and stitching the partial images to obtain the two-dimensional image;

[0048] Accordingly, obtaining a three-dimensional point cloud of the workpiece suspension area includes:

[0049] Controlling the three-dimensional laser camera to move to preset positions in sequence, and acquiring a local point cloud of the workpiece hanging area at the preset positions;

[0050] The local point clouds are spliced ​​to obtain a three-dimensional point cloud of the workpiece hanging area.

[0051] Optionally, before determining the corresponding target baking temperature according to the total mass of the workpieces in the batch, the method further includes:

[0052] Determine a mapping relationship between the total mass of the workpiece and the baking temperature based on historical data, establish a temperature prediction model based on the mapping relationship, and optimize the temperature prediction model using a gradient descent algorithm;

[0053] Accordingly, determining a corresponding target baking temperature according to the total mass of the workpieces in the batch, and controlling the coating production line to perform a corresponding baking operation according to the target baking temperature, includes:

[0054] Inputting the total mass of the batch of workpieces into the temperature prediction model to obtain the target baking temperature;

[0055] The opening of the natural gas regulating valve is adjusted according to the target baking temperature, so as to control the hot air blower to perform a baking operation on the workpieces in the coating production line according to the target baking temperature.

[0056] The present application also provides a baking temperature control system for a coating production line, the system comprising:

[0057] A two-dimensional image processing module is used to obtain a two-dimensional image of the workpiece hanging area of ​​the coating production line and generate a workpiece category detection result for each workpiece in the two-dimensional image using a target detection algorithm model;

[0058] a three-dimensional point cloud processing module, configured to obtain a three-dimensional point cloud of the workpiece suspension area and generate a corresponding size type detection result for each workpiece in the three-dimensional point cloud;

[0059] A mass calculation module, configured to determine the total mass of the batch of workpieces based on a workpiece category detection result and / or a size type detection result of each workpiece in the workpiece suspension area;

[0060] The temperature control module is used to determine the corresponding target baking temperature according to the total mass of the workpieces in this batch, and control the coating production line to perform the corresponding baking operation according to the target baking temperature.

[0061] The present application also provides a storage medium on which a computer program is stored. When the computer program is executed, the steps of the baking temperature adjustment method of the coating production line are implemented.

[0062] The present application also provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, the steps of the baking temperature adjustment method of the above-mentioned coating production line are implemented.

[0063] The present application discloses a method for regulating the baking temperature of a coating production line. The method obtains a two-dimensional image and a three-dimensional point cloud of a workpiece hanging area, determines and identifies the category of the workpiece in the workpiece hanging area based on the two-dimensional image, and identifies the size of the workpiece in the workpiece hanging area based on the three-dimensional point cloud. The mass of the workpiece is related to the category and size of the workpiece. On the basis of obtaining the workpiece category detection results and the size type detection results, the present application determines the total mass of the workpieces in this batch, and then determines the corresponding target baking temperature according to the total mass of the workpieces in this batch, so as to perform the corresponding baking operation according to the target baking temperature. The above process does not require manual intervention and can automatically adjust the baking temperature. Therefore, the present application can achieve high-precision and automated baking temperature regulation. The present application also provides a baking temperature regulation system for a coating production line, a storage medium and an electronic device, which have the above-mentioned beneficial effects and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0065] Figure 1 This is a flow chart of a method for adjusting the baking temperature of a coating production line provided in an embodiment of the present application;

[0066] Figure 2 A schematic diagram of the structure of an intelligent energy consumption regulation system for a coating production line provided in an embodiment of the present application;

[0067] Figure 3 A schematic diagram of the structure of a two-dimensional visual inspection module provided in an embodiment of the present application;

[0068] Figure 4 A schematic diagram of the structure of a three-dimensional visual point cloud processing module provided in an embodiment of the present application;

[0069] Figure 5 A schematic diagram of the arrangement of a camera and a lighting device of an image acquisition and processing module provided in an embodiment of the present application;

[0070] Figure 6 A schematic diagram of the structure of a YOLOv7-CSM model provided in an embodiment of the present application;

[0071] Figure 7 A schematic diagram of the structure of a comparative learning module provided in an embodiment of the present application;

[0072] Figure 8 A schematic diagram of a working scenario of a three-dimensional laser camera module provided in an embodiment of the present application;

[0073] Figure 9 A schematic diagram of a number alignment method provided in an embodiment of the present application;

[0074] Figure 10 This is a structural diagram of a baking energy consumption intelligent adjustment module provided in an embodiment of the present application. DETAILED DESCRIPTION

[0075] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0076] See below Figure 1 , Figure 1 This is a flow chart of a method for adjusting the baking temperature of a coating production line provided in an embodiment of the present application.

[0077] Specific steps may include:

[0078] S101: Acquire a two-dimensional image of a workpiece hanging area of ​​a coating production line, and use a target detection algorithm model to generate a workpiece category detection result for each workpiece in the two-dimensional image.

[0079] This embodiment can be applied to the control end of a coating production line, which may include workstations such as loading, pre-treatment, spraying, baking, cooling, and unloading. The workpiece hanging area is a baking station within the coating production line. The coating production line can also be a coating production line for engineering equipment, such as excavators, cranes, and shield machines.

[0080] In this embodiment, a 2D camera can be installed in the coating production line to capture a 2D image of the workpiece hanging area. After obtaining the 2D image, the 2D image can be input into an object detection algorithm model to generate an artifact category detection result for each artifact in the 2D image. The artifact category detection result represents the type of artifact contained in the 2D image generated by the object detection algorithm model, such as boom, arm, cylinder, connecting rod, etc.

[0081] The above-mentioned target detection algorithm model is an artificial intelligence model that can locate and classify artifacts in the image. Before this step, there can also be an operation of training the artificial intelligence model.

[0082] S102: Acquire a three-dimensional point cloud of the workpiece hanging area, and generate a corresponding size type detection result for each workpiece in the three-dimensional point cloud.

[0083] The coating line can also be equipped with a 3D laser camera to capture a 3D point cloud of the workpiece suspension area. To avoid missed workpieces, this embodiment can set the 3D laser camera's shooting direction to be different from that of the 2D camera. After obtaining the 3D point cloud, workpiece dimensions can be detected on the 3D point cloud to obtain the size and type of each workpiece.

[0084] The size type detection result is used to describe the type of workpiece size, such as large-size workpiece, medium-size workpiece, small-size workpiece, etc. This embodiment can also use other rules to classify the workpiece size types.

[0085] S103: Determine the total mass of the batch of workpieces according to the workpiece category detection result and / or size type detection result of each workpiece in the workpiece suspension area.

[0086] In this embodiment, the correspondence between workpiece type, workpiece size, and workpiece mass can be pre-stored in a database. Based on this, the embodiment can determine the mass of a single workpiece based on the workpiece type detection results, the size type detection results, or the combined workpiece type and size type detection results. After obtaining the masses of all workpieces contained in the 2D image and 3D point cloud, the embodiment can add up the masses of all workpieces to obtain the total mass of the batch of workpieces.

[0087] Specifically, if a workpiece exists in a two-dimensional image but not in a three-dimensional point cloud, it is determined that there is a workpiece category detection result corresponding to the workpiece, but there is no size type detection result corresponding to the workpiece. At this time, the quality of the workpiece can be determined based on the workpiece category detection result.

[0088] If a workpiece exists in the 3D point cloud but not in the 2D image, it is determined that there is a size type detection result corresponding to the workpiece, but there is no workpiece category detection result corresponding to the workpiece. At this time, the quality of the workpiece can be determined based on the size type detection result.

[0089] If a workpiece exists in both a two-dimensional image and a three-dimensional point cloud, it is determined that there are workpiece category detection results and size type detection results corresponding to the workpiece; at this time, it can be further determined whether the workpiece category detection result and the size type detection result of the workpiece match (that is, whether they conform to the correspondence between the workpiece category, workpiece size and workpiece quality stored in the database). If they match, the quality of the workpiece is determined based on the workpiece category detection result and the size type detection result. If they do not match, the quality of the workpiece is determined based on the type of information with a higher confidence level between the workpiece category detection result and the size type detection result.

[0090] S104: determining a corresponding target baking temperature according to the total mass of the workpieces in the batch, and controlling the coating production line to perform a corresponding baking operation according to the target baking temperature.

[0091] This embodiment can pre-store the relationship between the total mass of the workpieces and the baking temperature, and determine the target baking temperature corresponding to the total mass of the workpieces in the current batch based on this relationship. Furthermore, this embodiment can pre-train a temperature prediction model that learns the mapping relationship between the total mass of the workpieces and the optimal baking temperature. This embodiment can input the total mass of the workpieces in the current batch into the temperature prediction model to obtain the corresponding target baking temperature. After obtaining the target baking temperature, this embodiment can control the hot air blower to bake the workpieces in the coating production line at the target baking temperature.

[0092] This embodiment acquires a two-dimensional image and a three-dimensional point cloud of the workpiece suspension area, identifies the type of workpieces in the suspension area based on the two-dimensional image, and identifies the size of the workpieces in the suspension area based on the three-dimensional point cloud. Workpiece mass is related to the type and size of the workpieces. Based on the workpiece type and size detection results, this embodiment determines the total mass of the batch of workpieces. Furthermore, the target baking temperature corresponding to the total mass of the batch of workpieces is determined based on the total mass of the batch, so that the corresponding baking operation can be performed at the target baking temperature. This process requires no human intervention and automatically adjusts the baking temperature. Therefore, this embodiment can achieve high-precision and automated baking temperature adjustment.

[0093] As for Figure 1 Further introduction to the corresponding embodiment: the above-mentioned coating production line is provided with a workpiece suspension rack, a three-dimensional laser camera and multiple two-dimensional cameras. The workpiece suspension area is the area below the workpiece suspension rack, all the two-dimensional cameras are in the same plane, and the shooting direction of the three-dimensional laser camera is perpendicular to the shooting direction of each of the two-dimensional cameras.

[0094] The workpiece hanger may be rectangular in shape. In this embodiment, a virtual rectangle having the same shape as the workpiece hanger may be provided at a predetermined distance directly below the workpiece hanger, with the virtual rectangle being parallel to the plane on which the workpiece hanger resides. The two-dimensional cameras used in this embodiment may include an upper left camera, a lower left camera, an upper right camera, a lower right camera, an upper side camera, and a lower side camera. The optical centers of the upper left camera, the lower left camera, the upper right camera, the lower right camera, the upper side camera, and the lower side camera are all located in the plane on which the virtual rectangle resides, and the shooting directions of the upper left camera, the lower left camera, the upper right camera, the lower right camera, the upper side camera, and the lower side camera all pass through the diagonal center of the virtual rectangle.

[0095] The three-dimensional laser camera can be arranged on a movable device below the workpiece suspension frame, and the shooting direction of the three-dimensional laser camera is perpendicular to the plane where the workpiece suspension frame is located.

[0096] Based on the above structure, this embodiment can obtain a two-dimensional image of the workpiece hanging area of ​​the coating production line in the following manner: using all the two-dimensional cameras to capture partial images of the workpiece hanging area of ​​the coating production line from multiple directions, and stitching the partial images to obtain the two-dimensional image;

[0097] Based on the above structure, this embodiment can obtain the three-dimensional point cloud of the workpiece hanging area in the following manner: control the three-dimensional laser camera to move to preset positions in sequence, and obtain the local point cloud of the workpiece hanging area at the preset positions; splice the local point clouds to obtain the three-dimensional point cloud of the workpiece hanging area.

[0098] The above method can realize multi-view imaging, thereby avoiding the problem of mutual occlusion between workpieces due to the difference in viewing angles.

[0099] As for Figure 1 Further introduction to the corresponding embodiment: based on the three-dimensional point cloud, this embodiment can generate a corresponding size type detection result for each workpiece in the three-dimensional point cloud in the following manner: perform point cloud filtering and segmentation operations on the three-dimensional point cloud to obtain a workpiece point cloud for each workpiece; calculate the point cloud plane equation of each workpiece based on the workpiece point cloud, and determine the inclination angle of the point cloud plane equation; determine the minimum bounding box of the workpiece point cloud based on the point cloud plane equation, and calculate the volume of the workpiece point cloud based on the minimum bounding box; determine the size type detection result of each workpiece based on the area of ​​the workpiece point cloud, the volume of the workpiece point cloud and the inclination angle of the point cloud plane equation.

[0100] In a 3D point cloud, a single workpiece can have any number of planes. This embodiment generates a corresponding point cloud plane equation for each plane of the workpiece, and then combines all corresponding point cloud plane equations to determine the minimum bounding box of the workpiece point cloud. This minimum bounding box is the smallest circumscribed cube that can enclose the workpiece point cloud. Based on the area of ​​the workpiece point cloud, the volume of the workpiece point cloud, and the tilt angle of the point cloud plane equation, the size estimate of each workpiece is determined, resulting in the corresponding size type detection result.

[0101] Specifically, this embodiment can acquire three-dimensional point clouds from multiple locations, and the process of performing point cloud filtering and segmentation operations on the three-dimensional point clouds is as follows: performing point cloud registration and splicing preprocessing operations on the three-dimensional point clouds collected from multiple locations to obtain an integrated point cloud of the workpiece hanging area; performing point cloud filtering on the integrated point cloud to filter out noise point clouds and retain the point clouds of the driving plane and the workpiece; using the three-dimensional camera coordinate system as the reference system, setting a Z-axis threshold, partitioning the filtered point cloud in the Z-axis direction according to the Z-axis threshold, obtaining a point cloud near the driving plane, and using a plane fitting algorithm to fit the point cloud near the driving plane to obtain the driving plane equation. Setting a distance threshold, calculating the distance from the integrated point cloud to the driving plane equation, and filtering point cloud coordinate indexes within the distance threshold range; using the filtered driving plane point cloud coordinate indexes, filtering the point clouds of the driving plane and the hook on the integrated point cloud to obtain the workpiece point cloud; filtering the workpiece point cloud to filter out point cloud noise, and using a point cloud clustering algorithm to segment the workpiece point cloud to obtain independent point clouds for each workpiece. The Z-axis direction is perpendicular to the plane where the workpiece hanger is located. The crane is a device for transporting workpieces in the coating production line.

[0102] As for Figure 1 As a further introduction to the corresponding embodiment, in this embodiment, the total mass of the workpieces in this batch can be determined in the following manner.

[0103] Step A1: Select a target workpiece from the workpiece hanging area.

[0104] Step A2: querying the database for the size type corresponding to the workpiece category detection result of the target workpiece to obtain a size type query result.

[0105] The database stores the size type corresponding to each workpiece category. In this step, the size type queried from the database is used as the size type query result.

[0106] Step A3: Determine whether the size type detection result of the target workpiece is the same as the size type query result; if so, proceed to step A4; if not, proceed to step A5.

[0107] Step A4: determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type detection result.

[0108] Step A5: Determine the two-dimensional detection confidence and the three-dimensional detection confidence of the target workpiece.

[0109] The two-dimensional detection confidence level is the confidence level of the workpiece category detection result, and the three-dimensional detection confidence level is the confidence level of the size type detection result.

[0110] Step A6: If the two-dimensional detection confidence is greater than the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type query result.

[0111] Step A7: If the two-dimensional detection confidence is less than the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the size type detection result.

[0112] Step A8: If the two-dimensional detection confidence is equal to the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type detection result.

[0113] Specifically, if the two-dimensional detection confidence is equal to the three-dimensional detection confidence, the first alternative mass of the target workpiece is determined based on the workpiece category detection result and the size type query result, and the second alternative mass of the target workpiece is determined based on the size type detection result, and the average value of the first alternative mass and the second alternative mass is set as the workpiece mass of the target workpiece.

[0114] Step A9: Determine whether all workpieces in the workpiece suspension area have been selected; if so, calculate the total mass of this batch of workpieces based on the mass of all workpieces in the workpiece suspension area; if not, enter step A1 of selecting the target workpiece from the workpiece suspension area.

[0115] As a feasible implementation method, after selecting the target workpiece from the workpiece hanging area, if the target workpiece has a corresponding workpiece category detection result, but does not have a corresponding size type detection result, the workpiece quality of the target workpiece is determined based on the workpiece category detection result and the size type query result; if the target workpiece has a corresponding size type detection result, but does not have a corresponding workpiece category detection result, the workpiece quality of the target workpiece is determined based on the size type detection result.

[0116] As a feasible implementation, the process of determining the workpiece quality of the target workpiece based only on the size type detection result includes: querying the database for alternative workpieces that match the size type detection result; and setting the average quality of all the alternative workpieces as the workpiece quality of the target workpiece.

[0117] As for Figure 1In a further description of the corresponding embodiment, before using the object detection algorithm model to generate an artifact category detection result for each artifact in the two-dimensional image, training samples can also be obtained from the dataset; and the object detection algorithm model is trained using the training samples. Each training sample includes an original sample, a positive sample, and a negative sample. The original sample is an image containing an artifact, the positive sample is an image within the artifact's bounding box in the original sample, and the negative sample is an image outside the artifact's bounding box in the original sample.

[0118] The above-mentioned target detection algorithm model includes a skeleton network, a neck network, a detection head, a contrastive learning module and a loss calculation module. The contrastive learning module includes a global average pooling layer, a multi-layer perceptron and a spatial attention layer. The contrastive learning module is used to convert the feature map input by the skeleton network into a positive sample global enhanced feature map.

[0119] The process of converting the feature map input by the skeleton network into a positive sample global enhanced feature map by the contrast learning module includes: receiving the original sample feature map F, the positive sample feature map F+ and the negative sample feature map F- output by the skeleton network. The original sample feature map F, the positive sample feature map F+ and the negative sample feature map F- are input into the global average pooling layer GAP to obtain the global channel weight; wherein, the global channel weight includes the original sample global channel weight , positive sample global channel weight and negative sample global channel weights . The original sample feature map F and the original sample global channel weight Perform point multiplication to obtain the global enhanced feature map . The global channel weight of the original sample , the positive sample global channel weight and the negative sample global channel weight Input the multi-layer perceptron to obtain the processing result. Input the original sample feature map F, the positive sample feature map F+ and the negative sample feature map F- into the spatial attention layer SA to obtain the spatial local feature weight; wherein the spatial local feature weight includes the positive sample spatial local feature weight , local feature weights in negative sample space and the local feature weights of the original sample space The original sample feature map F, the positive sample feature map F+ and the negative sample feature map F- are respectively multiplied by the corresponding spatial local feature weights to obtain a local enhanced feature map; wherein the local enhanced feature map includes the original sample local enhanced feature map Fs, the positive sample local enhanced feature map And negative sample local enhanced feature map . The original sample local enhanced feature map Fs is compared with the negative sample local enhanced feature map Perform element-wise subtraction to obtain the local information attention feature Fs1. Calculate the contrast loss function value based on the processing result and the local enhanced feature map, and calculate the approximation loss function value based on the positive sample local enhanced feature map and the local information attention feature, so as to optimize the parameters of the contrast learning module using the contrast loss function value and the approximation loss function value. Perform element-wise addition operation on the local information attention feature Fs1 to obtain the positive sample global enhanced feature map ; Wherein, the positive sample global enhanced feature map is used to input into the neck network.

[0120] The skeleton network can output sample feature maps of multiple spatial resolutions for three types of samples (original samples, positive samples, and negative samples), and use the sample feature maps of the same spatial resolution as a set of original sample feature maps F, positive sample feature maps F+, and negative sample feature maps F-, so that the comparative learning module can process the original sample feature maps F, positive sample feature maps F+, and negative sample feature maps F- with the same spatial resolution.

[0121] As for Figure 1 Further introduction to the corresponding embodiment: before determining the corresponding target baking temperature based on the total mass of the workpieces in this batch, the mapping relationship between the total mass of the workpieces and the baking temperature can also be determined based on historical data, and a temperature prediction model can be established based on the mapping relationship, and the temperature prediction model can be optimized using a gradient descent algorithm.

[0122] Accordingly, the corresponding target baking temperature is determined according to the total mass of the workpieces in this batch, and the coating production line is controlled to perform the corresponding baking operation according to the target baking temperature, including: inputting the total mass of the workpieces in this batch into the temperature prediction model to obtain the target baking temperature; adjusting the opening of the natural gas regulating valve according to the target baking temperature, so as to control the hot air blower to perform the baking operation on the workpieces in the coating production line according to the target baking temperature.

[0123] In the process of baking workpieces, existing solutions usually determine the baking temperature based on the working parameters of the hot air blower. However, in actual applications, due to changes in ambient humidity, differences in the initial temperature of the workpieces, power fluctuations caused by aging of the hot air blower, and uneven thickness of the coating on the workpiece surface, it is difficult to ensure that the workpiece is actually baked at the target baking temperature. To address this problem, the present embodiment has an improvement method in this regard, which is as follows: an infrared thermal imager, a temperature sensor, and a humidity sensor are set on the coating production line, the infrared thermal imager is used to detect the surface temperature of the workpiece, the temperature sensor is used to detect the ambient temperature, and the humidity sensor is used to detect the ambient humidity. The present embodiment can monitor the surface temperature, ambient temperature, and ambient humidity of the workpiece in real time during the baking process of the coating production line, and use Kalman filtering or fuzzy fusion models to process the surface temperature, ambient temperature, and ambient humidity of the workpiece to generate a baking status feedback signal, so as to dynamically adjust the target baking temperature according to the baking status feedback signal through an adaptive algorithm.

[0124] The process described in the above embodiment is explained below through an embodiment in actual application.

[0125] With the rapid development of artificial intelligence and intelligent manufacturing technology, the factory production model of the construction machinery industry is transforming towards automation and unmanned operation. The industry has also put forward higher requirements for the green, energy-saving, intelligent and automated application of intelligent manufacturing technology in the industrial field.

[0126] Excavators are a mainstream product in the construction machinery industry, and their assembly lines are a core component of factory manufacturing. However, under the current production model, the traditional excavator workpiece coating and baking process still relies on manual piece counting and empirical estimation to adjust the baking temperature for each batch of excavator parts. While this method reduces labor costs, it suffers from low production efficiency, insufficient automation, and the inability to standardize manual temperature determination, resulting in energy waste and increased factory emissions.

[0127] The key to solving this problem lies in accurately identifying the type and size of each workpiece and automatically setting the correct baking temperature. Although deep learning-based target detection algorithms (such as the YOLO series) are widely used in industry, their recognition accuracy and generalization capabilities remain challenging in the excavator workpiece painting and baking scenario. For example, when the workpieces enter the baking area via the crane suspension, they vary in size and are relatively close together. The field of view of a single two-dimensional camera is limited, resulting in occlusion of the workpiece image and difficulty in obtaining complete information. In addition, excavator workpieces are cross-scale, diverse in type, and have similar features. YOLO (You Only Look Once), a target detection algorithm based on a two-dimensional monocular camera, has difficulty adapting to these complex situations.

[0128] Existing technologies attempt to overcome the problem of missing field of view through 3D reconstruction using binocular or multi-camera technology. However, this type of technology requires expensive industrial-grade CCD (charge coupled device) cameras and high-precision multi-target localization algorithms. However, in real-world production, the area where the workpiece is suspended by a crane is large, and the space available for camera placement is limited. This results in significant differences in field of view between cameras at distant locations, as well as non-overlapping fields of view, which limits the accuracy of the multi-target localization algorithm. Another type of object recognition technology based on 3D vision directly scans the workpiece area using 3D imaging and utilizes point cloud processing algorithms to identify the workpiece's type and geometric features. However, this technology still faces the following challenges: Traditional point cloud processing solutions based on PCL (Point Cloud Library) rely on template registration technology to determine workpiece type. However, in reality, it is difficult to fully cover the entire 3D (three-dimensional) model library of workpieces, and the production cycle is long and costly. Point cloud recognition algorithms based on deep learning also face the dilemma of high complexity in training dataset generation. In addition, the point cloud volume estimation algorithm based on cross-section integration is difficult to perform well in the case of hollowing and irregular structures inside the workpiece.

[0129] As can be seen, monocular vision-based object detection algorithms struggle to cope with wide detection scenarios, occlusion between workpieces, and multi-scale and diverse detection scenarios. 3D reconstruction solutions based on binocular or multi-camera cameras are expensive and struggle to overcome the lack of reconstruction accuracy caused by non-overlapping fields of view. 3D vision-based object recognition technologies are limited by the variability of workpiece geometry and the difficulty of producing data or 3D models.

[0130] To address the above issues, this embodiment provides an intelligent energy consumption adjustment solution for excavator painting production lines. This solution addresses the problem of accurate and generalized identification of the size, type, and category of excavator workpieces during painting and baking, and overcomes the low automation and energy waste associated with traditional manual baking temperature setting. This embodiment integrates the advantages of two-dimensional and three-dimensional vision, overcoming the limited field of view and insufficient model robustness of monocular object detection algorithms. It economically achieves effective multi-field perception and detection of workpieces. Furthermore, a concise and effective three-dimensional vision point cloud processing algorithm is designed to address the misclassification problem caused by the lack of three-dimensional information between identical workpiece features in two-dimensional imaging. Combining two-dimensional and three-dimensional information enables accurate assessment of workpiece categories in excavator painting and baking scenarios. Furthermore, an intelligent baking temperature adjustment algorithm automatically and accurately adjusts the baking temperature of batches of workpieces. This embodiment designs a network model, YOLOv7-CSM, based on YOLOv7 and contrastive learning. By constructing a CSM module, the model improves its ability to distinguish between workpieces and background in the presence of overlapping workpieces. This embodiment proposes a multi-view imaging solution using multiple two-dimensional cameras to overcome the problem of mutual occlusion caused by differences in viewing angles between workpieces. It also avoids the high cost and limited calibration accuracy problems of three-dimensional reconstruction solutions based on multi-cameras. This embodiment designs a point cloud processing algorithm based on three-dimensional vision, and does not introduce template registration and deep learning models. It obtains the basic size type of the excavator workpiece by evaluating the geometric rules of the workpiece placement and the area and volume. This embodiment proposes a scheme for matching the target detection type of the workpiece under multi-camera two-dimensional camera imaging with the size type of the workpiece point cloud under three-dimensional vision, and a mixed decision on the workpiece category. The two-dimensional detection results are integrated with the three-dimensional detection results and refined and evaluated to output the accurate workpiece category. This embodiment proposes an intelligent baking energy consumption adjustment scheme to realize intelligent and automatic adjustment of the baking temperature of batch workpieces in the excavator painting production line, thereby reducing energy waste.

[0131] See Figure 2 , Figure 2 This is a structural schematic diagram of an intelligent energy consumption adjustment system for a coating production line provided in an embodiment of the present application. The system includes a two-dimensional visual detection module, a three-dimensional visual point cloud processing module, and a baking energy consumption intelligent adjustment module. The control end of the coating production line (such as an industrial computer or server) is connected to the two-dimensional visual detection module and the three-dimensional visual point cloud processing module, respectively.

[0132] See Figure 3 , Figure 3This is a schematic diagram of the structure of a two-dimensional visual inspection module provided in an embodiment of the present application. The two-dimensional visual inspection module includes an image acquisition and processing module and a target detection algorithm model. The image acquisition and processing module includes a lighting device, multiple two-dimensional cameras (such as camera 1, camera 2, ..., camera 6), and an image preprocessing module. The image acquisition and processing module can acquire a two-dimensional image of an excavator workpiece and transmit the 2D image to the target detection algorithm model. An industrial computer or server is connected to the image acquisition and processing module and the target detection algorithm model, respectively.

[0133] See Figure 4 , Figure 4 This is a schematic diagram of the structure of a 3D visual point cloud processing module provided in an embodiment of the present application. The 3D visual point cloud processing module includes a 3D laser camera module, a point cloud processing algorithm module, and a hybrid decision-making module. The 3D laser camera module includes a lighting device, an electric cart, and a 3D laser camera. The 3D laser camera module is used to collect 3D point clouds of excavator workpieces. An industrial computer or server is connected to the 2D visual inspection module, the hybrid decision-making module, the point cloud processing algorithm module, and the 3D laser camera module.

[0134] The image acquisition and processing module is primarily used to construct multi-view 2D images of excavator workpieces, acquiring image information of the workpiece from different fields of view. It also performs preliminary preprocessing on the acquired data, providing data support for the subsequent target detection algorithm model. In this embodiment, the hardware of the image acquisition and processing module primarily consists of six cameras and a lighting device. The cameras are located at the upper left, lower left, upper right, lower right, and upper and lower sides of the workpiece inspection area.

[0135] See Figure 5 , Figure 5 This is a schematic diagram of the arrangement of cameras and lighting devices for an image acquisition and processing module provided in an embodiment of the present application. The two-dimensional cameras used in this embodiment include an upper left camera D1, a lower left camera D2, an upper right camera D3, a lower right camera D4, an upper side camera D5, and a lower side camera D6. The upper left camera D1, the lower left camera D2, the upper right camera D3, and the lower right camera D4 have their corresponding lighting devices, namely, an upper left light source B1, a lower left light source B2, an upper right light source B3, and a lower right light source B4. Figure 5 In the figure, C1 represents the workpiece suspension frame (i.e., the workpiece suspension rack), and C2 represents the rectangle of the same length and width below the workpiece suspension frame.

[0136] The upper left camera D1, the lower left camera D2, the upper right camera D3, and the lower right camera D4 are respectively located at the discharge port and the feed port 500 mm (millimeters) vertically downward of the workpiece suspension frame C1, and the optical center of the camera is facing the diagonal center of a rectangle with the same length and width 500 mm vertically downward of the workpiece suspension frame; the upper side camera D5 and the lower side camera D6 are respectively located at 500 mm from the upper side and 500 mm from the lower side of a rectangle with the same length and width 500 mm vertically downward of the workpiece suspension frame, and the optical center of the camera is facing the diagonal center of the rectangle; the lighting device includes four light sources, wherein each light source is respectively located 500 mm vertically above the upper left camera D1, the lower left camera D2, the upper right camera D3, and the lower right camera D4, and the optical center of the light source is downward 45 degrees toward the diagonal center of a rectangle with the same length and width 500 mm vertically downward of the workpiece suspension frame.

[0137] The steps of the data preprocessing method of the image acquisition and processing module in this embodiment are as follows:

[0138] Step E1: The collected two-dimensional images at the upper left, lower left, upper right and lower right directions are divided into a pair of upper left and lower left, and a pair of upper right and lower right.

[0139] Step E2: For each pair of images mentioned above, the key point features of the image are obtained by using the Harris operator.

[0140] Step E3: Calculate the invariant feature descriptors of the two images using SIFT (Scale-Invariant Feature Transform), and align the two images based on their key point features and invariant feature descriptors.

[0141] Step E4: Use Ransac (Random Sample Consensus) to calculate the homography matrix between the two image features.

[0142] Step E5: Perform affine transformation on the image using the homography matrix, and finally perform splicing and fusion on the current image to obtain the left workpiece image and the right workpiece image.

[0143] The object detection algorithm model is used to calculate the object's category based on the two-dimensional images (i.e., the left and right workpiece images) acquired by the image acquisition and processing module. In this embodiment, the object detection algorithm model is based on the YOLOv7 model (an object detection algorithm) and incorporates a contrastive learning module (CSM) to enhance the model's ability to distinguish between objects and backgrounds, improving its accuracy and generalization.

[0144] In this embodiment, the collected two-dimensional image can be annotated, and then the designed contrastive learning module is inserted into the YOLOv7 model to obtain a YOLOv7-CSM (i.e., an object detection algorithm provided with a contrastive learning module) model. The YOLOv7-CSM model is then trained, and finally the YOLOv7-CSM model is inferred to obtain the category of the artifact in the image.

[0145] In this embodiment, the image dataset collected by the image acquisition and processing module can be divided into a training set, a validation set, and a test set in a 3:1:1 ratio. This embodiment can also annotate the image data. Specifically, the dataset can be categorized into three types: large, medium, and small based on the specific size of the excavator workpieces. Individual workpieces within these three categories are annotated, with the annotation information including the workpiece category, the coordinates of the workpiece's bounding box corners, and the workpiece size type. Each image is divided into positive and negative sample images based on the coordinates of the annotated target bounding box. Specifically, for positive samples, only the pixel values ​​within the target bounding box are retained, while all other pixel values ​​outside the box are assigned a value of 0. For negative samples, only the pixel values ​​outside the target bounding box are retained, while all other pixel values ​​outside the box are assigned a value of 0. The original image (i.e., the original sample), the positive sample, and the negative sample are paired and integrated to generate the excavator workpiece dataset for training. This embodiment can perform image enhancement operations on the dataset before inputting it into model training, including image scaling, grayscale processing, and contrast enhancement.

[0146] This embodiment can be Figure 6 Build an improved YOLOv7 network structure: YOLOv7-CSM.

[0147] Figure 6 A structural diagram of a YOLOv7-CSM model provided in an embodiment of the present application, wherein the YOLOv7-CSM model includes a contrastive learning module, which is located after the feature map P3, feature map P4, and feature map P5 output by the skeleton network of YOLOv7, and before the neck network Neck structure. The above-mentioned model also includes a backbone network Backbone, a detection head Head, and a loss calculation module Loss. The input of the backbone network includes the original sample Img, the positive sample Img+, and the negative sample Img-. After the original sample Img is processed by the backbone network, the original sample feature map F is obtained; after the positive sample Img+ is processed by the backbone network, the positive sample feature map F+ is obtained; after the negative sample Img- is processed by the backbone network, the negative sample feature map F- is obtained. Feature map P3, feature map P4, and feature map P5 are three feature maps with different spatial resolutions.

[0148] Specifically, the contrastive learning module includes a global average pooling layer GAP, a multi-layer perceptron MLP, a spatial attention layer SA, two point multiplication and element addition and subtraction operations, a contrast loss function and an approximation loss function.

[0149] Contrastive loss function used by the contrastive learning module The formula is as follows.

[0150] .

[0151] in, To approximate the loss function, the cosine similarity loss is used here, and N represents the total number of a set of positive sample feature maps, negative sample feature maps, and original sample feature maps. The temperature parameter used for function adjustment.

[0152] represents the contrast loss function value, Express expectations, represents the logarithm, k and i represent the sample number, represents the indicator function, Represents the kth original sample feature map.

[0153] The approximation loss function used The formula is as follows.

[0154] .

[0155] in This is a feature flattening operation that reduces the dimension of the feature map. n represents the total length of the feature map after dimensionality reduction.

[0156] Represents the value of the approximation loss function.

[0157] See Figure 7 , Figure 7 This is a structural diagram of a comparative learning module provided in an embodiment of the present application. represents the dot product operation, represents element-wise subtraction, Represents element addition. The specific steps of the contrastive learning module to process the feature map are as follows:

[0158] Step F1: For each feature map of the original sample (feature map P3, feature map P4 and feature map P5), take the feature maps of the positive samples and negative samples corresponding to the same batch to form a set of original sample feature map F, positive sample feature map F+ and negative sample feature map F-, and send them to the contrastive learning module.

[0159] Step F2: The original sample feature map F, the positive sample feature map F+ and the negative sample feature map F- enter the global average pooling layer GAP and the spatial attention layer SA respectively to extract the global channel weights (including the original sample global channel weights , positive sample global channel weight and negative sample global channel weights ) and spatial local feature weights (including positive sample spatial local feature weights , local feature weights in negative sample space and the local feature weights of the original sample space ).

[0160] Among them, after the original sample feature map F, the positive sample feature map F+ and the negative sample feature map F- are input into the global average pooling layer GAP, the global average pooling layer GAP outputs the original sample global channel weight corresponding to the original sample feature map , the positive sample global channel weight corresponding to the positive sample feature map Negative sample global channel weight corresponding to the negative sample feature map .

[0161] After the original sample feature map F, the positive sample feature map F+ and the negative sample feature map F- are input into the spatial attention layer SA, the spatial attention layer SA outputs the positive sample spatial local feature weight corresponding to the positive sample feature map , the local feature weight of the negative sample space corresponding to the negative sample feature map The local feature weight of the original sample space corresponding to the original sample feature map .

[0162] Step F3: Original sample feature map F and original sample global channel weight After doing the dot multiplication operation, we get the global enhanced feature map , take the original sample global channel weight obtained in step F2 , positive sample global channel weight and negative sample global channel weights Send it to the multi-layer perceptron MLP and then input the contrast loss function Calculation is performed in .

[0163] Step F4: The original sample feature map F, the positive sample feature map F+ and the negative sample feature map F- are respectively compared with the original sample space local feature weights , local feature weights of positive sample space and local feature weights in the negative sample space Perform dot multiplication to obtain the original sample local enhanced feature map Fs and the positive sample local enhanced feature map And negative sample local enhanced feature map , the original sample local enhanced feature map Fs, the positive sample local enhanced feature map And negative sample local enhanced feature map Feed into the contrast loss function Calculation is performed in .

[0164] Specifically, this step combines the original sample feature map F with the original sample space local feature weight Perform a dot multiplication operation to obtain the local enhanced feature map Fs of the original sample; add the positive sample feature map F+ and the local feature weight of the positive sample space Perform point multiplication operation to obtain the local enhanced feature map of the positive sample ; Combine the negative sample feature map F- with the local feature weight of the negative sample space Perform point multiplication operation to obtain the negative sample local enhanced feature map .

[0165] Step F5: Take the original sample local enhanced feature map Fs and the positive sample local enhanced feature map obtained in step F4 And negative sample local enhanced feature map , use the original sample local enhanced feature map Fs to subtract the negative sample local enhanced feature map Get the local information focus feature Fs1, and enhance the local feature map of the positive sample Feed it into the approximation loss function together with Fs1 Calculation is performed in .

[0166] Step F6: Take the value obtained in step F3 Add the element-wise addition operation to Fs1 obtained in step F5 to obtain the positive sample global enhanced feature map , and globally enhance the feature map of the positive sample Into the neck network.

[0167] Train the YOLOv7-CSM model and evaluate its performance using the test set. Perform model inference based on the trained YOLOv7-CSM model to obtain artifact detection results.

[0168] In the present invention, the three-dimensional laser camera module is mainly used to obtain three-dimensional point cloud data of the workpiece. The module includes a lighting device, an electric cart and a three-dimensional laser camera, wherein the lighting device and the three-dimensional laser camera are located on the cart.

[0169] See Figure 8 , Figure 8This is a schematic diagram of a working scenario of a three-dimensional laser camera module provided in an embodiment of the present application. G1 represents a three-dimensional laser camera, G2 represents a lighting device, G3 represents an electric cart, H1 represents the left area, H2 represents the middle area, and H3 represents the right area. The figure also shows the entrance side, exit side and multiple workpieces.

[0170] In this embodiment, the workpiece hanging area can be divided into three acquisition areas, namely the right area, the middle area, and the left area, according to the three hole rectangles on the right, middle, and left sides of the workpiece hanging rack, with the intersection of the rectangular diagonals as the center. Turn on the three-dimensional laser camera, initialize it, and perform parameter calibration. After the YOLOv7-CSM model detects the workpiece, the server transmits a start flag to the control processing unit on the cart. The cart moves at a constant speed, moving from the right area to the left area where the workpiece is suspended. Each time it moves to the center position of each interval, it stops and transmits the point cloud of the area to the server. After completing the shooting of the three intervals, the control unit on the cart transmits the point cloud processing algorithm running flag back to the algorithm module in the server. Then, the cart returns to the initial position at a constant speed with the three-dimensional camera, and waits for the next scanning task.

[0171] The point cloud processing algorithm module is mainly used to process the workpiece point cloud data collected by the 3D laser camera module, evaluate the geometric features of each workpiece point cloud such as volume, area, and inclination, and thereby identify the size and type of each workpiece, such as large, medium, and small.

[0172] The processing process of the point cloud processing algorithm module includes the following steps.

[0173] Step I1: Perform a point cloud registration and splicing preprocessing operation on the point cloud data collected from the three received areas to obtain an integrated point cloud of the workpiece hanging area.

[0174] Step I2: Perform point cloud filtering on the integrated point cloud to remove noise point clouds and retain the point clouds of the driving plane and the workpiece.

[0175] Step I3: Using the 3D camera coordinate system as the reference system, set a Z-axis threshold. Partition the filtered point cloud along the Z-axis according to the threshold. Obtain a point cloud near the vehicle plane. Use the RANSAC plane fitting algorithm to fit the point cloud near the vehicle plane to obtain the vehicle plane equation. Set a distance threshold, calculate the distance from the integrated point cloud to the vehicle plane equation, and select point cloud coordinate indices within the distance threshold.

[0176] Step I4: Using the filtered crane plane point cloud coordinate index, filter the crane and hook point clouds on the integrated point cloud to obtain the workpiece point cloud.

[0177] Step I5: Filter the workpiece point cloud to remove point cloud noise, and use the point cloud clustering algorithm to segment the workpiece point cloud to obtain an independent point cloud for each workpiece.

[0178] Step I6: For each workpiece point cloud, fit the point cloud plane equation, and calculate the inclination angle between the point cloud plane equation of each workpiece and the horizontal reference plane, where the horizontal reference plane is perpendicular to the Z axis of the 3D camera coordinate system.

[0179] Step I7: Estimate the area of ​​each workpiece point cloud, fit the minimum bounding box of each workpiece point cloud, estimate the volume of the point cloud based on the bounding box, and use the obtained area and volume parameters combined with the inclination angle of the workpiece point cloud plane equation obtained in step 6 to predict the size type of the workpiece (such as large, medium, or small), and output the prediction confidence.

[0180] In this embodiment, the hybrid decision module is mainly used to number the independent point cloud of each workpiece obtained by the point cloud processing algorithm, and based on this number, align the corresponding number of the two-dimensional external frame of the workpiece obtained by the target detection algorithm in the two-dimensional visual inspection module, so that the two-dimensional detection information of the workpiece and the three-dimensional prediction information of the three-dimensional point cloud processing algorithm are paired, and then the accurate category and size type of the workpiece are inferred through the fusion of two-dimensional and three-dimensional visual decisions.

[0181] See Figure 9 , Figure 9 A schematic diagram of a numbering alignment method provided for an embodiment of the present application, in which 1, 2, 3, 4, 5, 6, 7, 8, and 9 represent the numbers of the workpieces, X, Y, and Z represent the three-dimensional laser camera coordinate system, the figure shows the left area, middle area, and right area of ​​the workpiece hanger, the figure shows the position of the workpiece from the left side perspective, the right side perspective, the lower side perspective, and the upper side perspective, and the key position relationship includes left, right, far, and near.

[0182] The process of the hybrid decision module to determine the workpiece category and size type is as follows.

[0183] Step J1: According to the 3D camera shooting path, the workpieces are numbered in ascending order from the right area to the middle area to the left area (i.e. according to the Y-axis direction of the 3D laser camera). The starting number sequence is incremented from left to right along the X-axis direction of the 3D laser camera, such as Figure 9 shown.

[0184] Step J2: Number the workpiece point cloud obtained by the point cloud processing algorithm according to the method described in step J1.

[0185] Step J3: Use the depth estimation neural network model to perform depth estimation on the left workpiece image, right workpiece image, upper workpiece image and lower workpiece image data collected by the two-dimensional camera to obtain the distance relationship of the pixels.

[0186] Step J4: Feed the left, right, top, and bottom workpiece images captured by the 2D camera into the YOLOv7-CSM model to obtain the workpiece detection object results and number them. Numbering rule: Using the workpiece point cloud numbering as a reference, the right workpiece image is numbered from left to right and from near to far. The left workpiece image is numbered from right to left and from far to near. The top workpiece image is numbered from left to right and from far to near. The bottom workpiece image is numbered from right to left and from near to far. The right workpiece image is only for workpieces in the right area, and the left workpiece image is only for workpieces in the left area.

[0187] Step J5: Align the workpiece detection object number from the 2D camera with the workpiece point cloud number from the 3D camera. If the YOLOv7-CSM inspection data corresponding to the left, right, top, or bottom workpiece images captured by the 2D camera contains missed or missing workpieces, the inspection results from the other workpiece images are prioritized. If all are missing, the workpiece point cloud number is used as the basis, and only the workpiece size type with that number is retained.

[0188] Step J6: According to the workpiece number in step J5, obtain the point cloud size type estimation corresponding to the workpiece and the target detection category results of the YOLOv7-CSM model in sequence. If the workpiece point cloud type estimation and target detection both have values ​​and the evaluated workpiece size types are consistent, the consistent results are used as the final size of the workpiece, and the target detection category result is used as the category of the workpiece.

[0189] The above “both workpiece point cloud type estimation and target detection have values ​​and the evaluated workpiece size types are consistent” means that the size type obtained by the workpiece point cloud type estimation matches the workpiece category obtained by the target detection.

[0190] Step J7: If both the workpiece point cloud size type estimation and the target detection category of the workpiece in step J6 have values ​​but the workpiece size types are inconsistent, then the result with the largest prediction confidence among the workpiece point cloud size type estimation and the target detection category is taken as the standard. If the point cloud size type estimation confidence is high, only the size type is output; if the target detection category confidence is high, both the workpiece size type and category are output.

[0191] The above “both the workpiece point cloud size type estimation and the target detection category have values ​​but the workpiece size types are inconsistent” means that the size type obtained by the workpiece point cloud type estimation does not match the workpiece category obtained by the target detection.

[0192] Step J8: If either the workpiece point cloud size type estimate or the target detection category in step J6 is missing a value, the object with a value is used as the final output for the workpiece type. For example, if only the point cloud size type is available, only the workpiece size type is output. If only the target detection category is available, the target detection category is used as the output for both the workpiece size and category.

[0193] Among them, the above-mentioned "there is no value in the workpiece point cloud size type estimation and the target detection category": there is no size type obtained by the workpiece point cloud type estimation, or there is no workpiece category obtained by the target detection.

[0194] See Figure 10 , Figure 10 This is a schematic diagram of the structure of an intelligent baking energy consumption regulation module provided in an embodiment of the present application. The module includes a temperature regulation module, a hot air blower, and a natural gas regulating valve. An industrial computer / server is connected to the temperature regulation module and the natural gas regulating valve, respectively. The temperature regulation module's input includes the excavator workpiece type. The temperature regulation module is implemented by two components: a mass estimation algorithm and a temperature regulation algorithm (i.e., a mass-temperature mapping function model). First, based on the workpiece type for each batch obtained by the 2D visual inspection module and the 3D visual point cloud processing module, the corresponding mass values ​​are retrieved from the workpiece quality database and summed to calculate the total mass of the batch. After obtaining the total mass of the batch, a mapping function relationship between mass and temperature is established using gradient descent and linear regression algorithms. Based on the temperature calculated by the mapping function, the industrial computer adjusts the opening and closing of the natural gas regulating valve to control the hot air blower's air temperature, thereby regulating the baking room temperature. This module automatically implements intelligent baking energy consumption regulation, reducing the energy waste caused by manually setting a fixed baking temperature.

[0195] The baking energy consumption intelligent adjustment module can implement the following steps.

[0196] Step K1: Quality estimation.

[0197] Based on the type of workpieces in each batch calculated by the two-dimensional visual inspection module and the three-dimensional visual point cloud processing module, the corresponding quality value of the workpiece is taken from the manually established workpiece quality database. If only the size is available, the average quality value of the size is taken, and the sum is calculated to calculate the total quality of the workpieces in this batch.

[0198] Step K2: Based on expert experience, label the workpiece quality and temperature dataset, and randomly divide the total number of samples in the dataset into a training set, a validation set, and a test set in a ratio of 3:1:1.

[0199] Step K3: mass and temperature mapping function model training.

[0200] An initial model is established using a linear equation to map the workpiece mass to the baking temperature. The training and validation sets from step 2 are then used as input into the model and, using gradient descent optimization, the optimal weighting between mass and temperature is determined. Finally, the model's performance is evaluated using the test set, resulting in a model that can be used for inference.

[0201] Step K4: Temperature estimation inference.

[0202] The baking temperature corresponding to the total mass of the batch of workpieces transmitted in step K1 is inferred based on the mapping function model between mass and temperature established in step K3.

[0203] Step K5: Energy consumption adjustment.

[0204] According to the temperature inferred in step K4, the opening and closing degree of the natural gas regulating valve is adjusted by the industrial computer to control the blowing temperature of the hot air blower, thereby achieving temperature regulation of the baking room.

[0205] This embodiment provides a structural design for an intelligent energy consumption adjustment method for an excavator painting production line; this embodiment provides a structural design for a two-dimensional visual inspection module, with each module handling different functions; this embodiment provides a structural design for a three-dimensional visual point cloud processing module, with each module handling different functions; this embodiment provides a structural design for an intelligent baking energy consumption adjustment module, with each module handling different functions. This embodiment can solve the problem of accurately identifying the category of each workpiece through vision in the excavator workpiece painting production line scenario, where each batch of workpieces has different sizes, the workpiece hanging positions are random, there is occlusion between medium and small-sized workpieces under a single perspective, the workpieces have different shapes, and the workpieces are of various types, so as to automatically adjust the baking temperature of the batch of workpieces. This further promotes the automation and energy conservation of the excavator painting production line, reduces the energy waste and increased emissions caused by manually setting the baking temperature, and promotes the intelligent and green transformation of the production line.

[0206] An embodiment of the present application provides a baking temperature control system for a coating production line, which may include:

[0207] A two-dimensional image processing module is used to obtain a two-dimensional image of the workpiece hanging area of ​​the coating production line and generate a workpiece category detection result for each workpiece in the two-dimensional image using a target detection algorithm model;

[0208] a three-dimensional point cloud processing module, configured to obtain a three-dimensional point cloud of the workpiece suspension area and generate a corresponding size type detection result for each workpiece in the three-dimensional point cloud;

[0209] A mass calculation module, configured to determine the total mass of the batch of workpieces based on a workpiece category detection result and / or a size type detection result of each workpiece in the workpiece suspension area;

[0210] The temperature control module is used to determine the corresponding target baking temperature according to the total mass of the workpieces in this batch, and control the coating production line to perform the corresponding baking operation according to the target baking temperature.

[0211] This embodiment acquires a two-dimensional image and a three-dimensional point cloud of the workpiece suspension area, identifies the type of workpieces in the suspension area based on the two-dimensional image, and identifies the size of the workpieces in the suspension area based on the three-dimensional point cloud. Workpiece mass is related to the type and size of the workpieces. Based on the workpiece type and size detection results, this embodiment determines the total mass of the batch of workpieces. Furthermore, the target baking temperature corresponding to the total mass of the batch of workpieces is determined based on the total mass of the batch, so that the corresponding baking operation can be performed at the target baking temperature. This process requires no human intervention and automatically adjusts the baking temperature. Therefore, this embodiment can achieve high-precision and automated baking temperature adjustment.

[0212] Furthermore, the process of the quality calculation module determining the total mass of the batch of workpieces based on the workpiece category detection result and / or size type detection result of each workpiece in the workpiece hanging area includes: selecting a target workpiece from the workpiece hanging area; querying the size type corresponding to the workpiece category detection result of the target workpiece in the database to obtain the size type query result; judging whether the size type detection result of the target workpiece is the same as the size type query result; if so, determining the workpiece quality of the target workpiece based on the workpiece category detection result and the size type detection result; if not, determining the two-dimensional detection confidence and the three-dimensional detection confidence of the target workpiece; wherein the two-dimensional detection confidence is the confidence of the workpiece category detection result, and the three-dimensional detection confidence is the confidence of the size type detection result; if the two-dimensional detection confidence is greater than the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type query result; if the two-dimensional detection confidence is less than the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the size type detection result; if the two-dimensional detection confidence is equal to the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type detection result; judging whether all workpieces in the workpiece hanging area have been selected; if so, calculating the total quality of the workpieces of this batch according to the workpiece quality of all workpieces in the workpiece hanging area; if not, entering the step of selecting the target workpiece from the workpiece hanging area.

[0213] Furthermore, the quality calculation module is also used to determine the workpiece quality of the target workpiece based on the workpiece category detection result and the size type query result if there is a corresponding workpiece category detection result for the target workpiece and no corresponding size type detection result; and is also used to determine the workpiece quality of the target workpiece based on the size type detection result if there is a corresponding size type detection result for the target workpiece and no corresponding workpiece category detection result.

[0214] Furthermore, the process of the quality calculation module determining the workpiece quality of the target workpiece according to the size type detection result includes: searching the database for candidate workpieces that match the size type detection result; and setting the average quality of all the candidate workpieces as the workpiece quality of the target workpiece.

[0215] Furthermore, it also includes:

[0216] a model training module, configured to obtain training samples from a data set before generating a workpiece category detection result for each workpiece in the two-dimensional image using a target detection algorithm model; wherein each training sample includes an original sample, a positive sample, and a negative sample, wherein the original sample is an image containing the workpiece, the positive sample is an image within a bounding box of the workpiece in the original sample, and the negative sample is an image outside the bounding box of the workpiece in the original sample; and further configured to train the target detection algorithm model using the training samples;

[0217] Among them, the target detection algorithm model includes a skeleton network, a neck network, a detection head, a contrastive learning module and a loss calculation module. The contrastive learning module includes a global average pooling layer, a multi-layer perceptron and a spatial attention layer. The contrastive learning module is used to convert the feature map input by the skeleton network into a positive sample global enhanced feature map.

[0218] Furthermore, the process of converting the feature map input by the skeleton network into a positive sample global enhanced feature map by the contrastive learning module includes: receiving the original sample feature map, the positive sample feature map and the negative sample feature map output by the skeleton network; inputting the original sample feature map, the positive sample feature map and the negative sample feature map into the global average pooling layer to obtain global channel weights; wherein, the global channel weights include the original sample global channel weights, the positive sample global channel weights and the negative sample global channel weights; performing dot multiplication on the original sample feature map and the original sample global channel weights to obtain a global enhanced feature map; inputting the original sample global channel weights, the positive sample global channel weights and the negative sample global channel weights into a multilayer perceptron to obtain a processing result; inputting the original sample feature map, the positive sample feature map and the negative sample feature map into the spatial attention layer to obtain spatial local feature weights; This feature map, the positive sample feature map and the negative sample feature map are respectively multiplied by the corresponding spatial local feature weights to obtain a local enhancement feature map; wherein, the local enhancement feature map includes the original sample local enhancement feature map, the positive sample local enhancement feature map and the negative sample local enhancement feature map; the original sample local enhancement feature map is subjected to element-wise subtraction operation on the negative sample local enhancement feature map to obtain a local information attention feature; the contrast loss function value is calculated based on the processing result and the local enhancement feature map, and the approximation loss function value is calculated based on the positive sample local enhancement feature map and the local information attention feature, so as to optimize the parameters of the contrast learning module using the contrast loss function value and the approximation loss function value; the global enhancement feature map and the local information attention feature are subjected to element-wise addition operation to obtain a positive sample global enhancement feature map; wherein, the positive sample global enhancement feature map is used to be input into the neck network.

[0219] Furthermore, the process of the three-dimensional point cloud processing module generating a corresponding size type detection result for each workpiece in the three-dimensional point cloud includes: performing point cloud filtering and segmentation operations on the three-dimensional point cloud to obtain a workpiece point cloud for each workpiece; calculating the point cloud plane equation of each workpiece based on the workpiece point cloud, and determining the inclination angle of the point cloud plane equation; determining the minimum bounding box of the workpiece point cloud based on the point cloud plane equation, and calculating the volume of the workpiece point cloud based on the minimum bounding box; determining the size type detection result of each workpiece based on the area of ​​the workpiece point cloud, the volume of the workpiece point cloud and the inclination angle of the point cloud plane equation.

[0220] Furthermore, the coating production line is provided with a workpiece suspension rack, a three-dimensional laser camera and multiple two-dimensional cameras. The workpiece suspension area is the area below the workpiece suspension rack. All the two-dimensional cameras are in the same plane. The shooting direction of the three-dimensional laser camera is perpendicular to the shooting direction of each of the two-dimensional cameras.

[0221] Correspondingly, the process of the two-dimensional image processing module obtaining a two-dimensional image of the workpiece hanging area of ​​the coating production line includes: using all the two-dimensional cameras to shoot local images of the workpiece hanging area of ​​the coating production line from multiple directions, and stitching the local images to obtain the two-dimensional image.

[0222] Correspondingly, the process of the three-dimensional point cloud processing module obtaining the three-dimensional point cloud of the workpiece hanging area includes: controlling the three-dimensional laser camera to move to preset positions in sequence, and obtaining the local point cloud of the workpiece hanging area at the preset positions; splicing the local point clouds to obtain the three-dimensional point cloud of the workpiece hanging area.

[0223] Furthermore, it also includes: a temperature prediction model, which is used to determine the mapping relationship between the total mass of the workpiece and the baking temperature based on historical data before determining the corresponding target baking temperature based on the total mass of the workpieces in this batch, establish a temperature prediction model based on the mapping relationship, and optimize the temperature prediction model using a gradient descent algorithm.

[0224] Correspondingly, the temperature control module determines the corresponding target baking temperature based on the total mass of the batch of workpieces, and controls the coating production line to perform the corresponding baking operation according to the target baking temperature. The process includes: inputting the total mass of the batch of workpieces into the temperature prediction model to obtain the target baking temperature; adjusting the opening of the natural gas regulating valve according to the target baking temperature, so as to control the hot air blower to perform the baking operation on the workpieces in the coating production line according to the target baking temperature.

[0225] Since the embodiments of the system part correspond to the embodiments of the method part, please refer to the description of the embodiments of the method part for the embodiments of the system part, and will not be repeated here.

[0226] This application also provides a storage medium having a computer program stored thereon. When executed, the computer program can implement the steps provided in the above embodiments. The storage medium may include: a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, an optical disk, or other medium capable of storing program code.

[0227] The present application also provides an electronic device that may include a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, the steps provided in the above embodiment can be implemented. Of course, the electronic device may also include various network interfaces, a power supply, and other components.

[0228] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of this application, several improvements and modifications can be made to this application, and these improvements and modifications also fall within the scope of protection of this application.

[0229] It should also be noted that, in this specification, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

Claims

1. A method for regulating the baking temperature of a coating production line, characterized in that: include: Obtain a two-dimensional image of the workpiece hanging area of ​​the coating production line, and use the target detection algorithm model to generate a workpiece category detection result for each workpiece in the two-dimensional image; Acquire a three-dimensional point cloud of the workpiece suspension area, and generate a corresponding size type detection result for each workpiece in the three-dimensional point cloud; determining the total mass of the batch of workpieces according to a workpiece category detection result and / or a size type detection result of each workpiece in the workpiece suspension area; The corresponding target baking temperature is determined according to the total mass of the workpieces in the batch, and the coating production line is controlled to perform the corresponding baking operation according to the target baking temperature.

2. The method for adjusting the baking temperature of a coating production line according to claim 1, characterized in that: Determining the total mass of the batch of workpieces according to a workpiece category detection result and / or a size type detection result of each workpiece in the workpiece suspension area includes: Selecting a target workpiece from the workpiece hanging area; Querying a database for a size type corresponding to a workpiece category detection result of the target workpiece to obtain a size type query result; Determining whether the size type detection result of the target workpiece is the same as the size type query result; If yes, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type detection result; If not, determining a two-dimensional detection confidence level and a three-dimensional detection confidence level of the target workpiece; wherein the two-dimensional detection confidence level is the confidence level of the workpiece category detection result, and the three-dimensional detection confidence level is the confidence level of the size type detection result; If the two-dimensional detection confidence is greater than the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type query result; If the two-dimensional detection confidence is less than the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the size type detection result; If the two-dimensional detection confidence is equal to the three-dimensional detection confidence, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type detection result; Determine whether all workpieces in the workpiece hanging area have been selected; if so, calculate the total mass of this batch of workpieces based on the mass of all workpieces in the workpiece hanging area; if not, enter the step of selecting a target workpiece from the workpiece hanging area.

3. The method for adjusting the baking temperature of a coating production line according to claim 2, characterized in that: After selecting the target workpiece from the workpiece hanging area, the method further includes: If the target workpiece has a corresponding workpiece category detection result, but does not have a corresponding size type detection result, determining the workpiece quality of the target workpiece according to the workpiece category detection result and the size type query result; If a corresponding size type detection result exists for the target workpiece, but no corresponding workpiece category detection result exists, the workpiece quality of the target workpiece is determined according to the size type detection result.

4. The baking temperature adjustment method for a coating production line according to claim 2, characterized in that: Determining the workpiece quality of the target workpiece according to the size type detection result includes: Searching a database for candidate workpieces that match the size type detection result; The average mass of all the candidate workpieces is set as the workpiece mass of the target workpiece.

5. The baking temperature adjustment method for a coating production line according to claim 1, characterized in that: Before generating a workpiece category detection result for each workpiece in the two-dimensional image using the target detection algorithm model, the method further includes: Obtain training samples from a data set; wherein each training sample includes an original sample, a positive sample, and a negative sample, the original sample is an image containing an artifact, the positive sample is an image within the bounding box of the artifact in the original sample, and the negative sample is an image outside the bounding box of the artifact in the original sample; Using the training samples to train the target detection algorithm model; Among them, the target detection algorithm model includes a skeleton network, a neck network, a detection head, a contrastive learning module and a loss calculation module. The contrastive learning module includes a global average pooling layer, a multi-layer perceptron and a spatial attention layer. The contrastive learning module is used to convert the feature map input by the skeleton network into a positive sample global enhanced feature map.

6. The method for regulating the baking temperature of a coating production line according to claim 5, characterized in that: The process of converting the feature map input by the skeleton network into a positive sample global enhanced feature map by the contrastive learning module includes: Receiving the original sample feature map, the positive sample feature map and the negative sample feature map output by the skeleton network; Inputting the original sample feature map, the positive sample feature map, and the negative sample feature map into the global average pooling layer to obtain a global channel weight; wherein the global channel weight includes the original sample global channel weight, the positive sample global channel weight, and the negative sample global channel weight; Performing a dot multiplication of the original sample feature map and the original sample global channel weight to obtain a global enhanced feature map; Inputting the original sample global channel weight, the positive sample global channel weight, and the negative sample global channel weight into a multilayer perceptron to obtain a processing result; Inputting the original sample feature map, the positive sample feature map, and the negative sample feature map into the spatial attention layer to obtain spatial local feature weights; Performing dot multiplication of the original sample feature map, the positive sample feature map, and the negative sample feature map with the corresponding spatial local feature weights to obtain a local enhanced feature map; wherein the local enhanced feature map includes the original sample local enhanced feature map, the positive sample local enhanced feature map, and the negative sample local enhanced feature map; Performing element-wise subtraction operation on the local enhanced feature map of the original sample from the local enhanced feature map of the negative sample to obtain a local information attention feature; Calculating a contrast loss function value based on the processing result and the local enhancement feature map, and calculating an approximation loss function value based on the positive sample local enhancement feature map and the local information focus feature, so as to optimize the parameters of the contrastive learning module using the contrast loss function value and the approximation loss function value; An element-wise addition operation is performed on the global enhancement feature map and the local information attention feature to obtain a positive sample global enhancement feature map; wherein the positive sample global enhancement feature map is used to input into the neck network.

7. The method for regulating the baking temperature of a coating production line according to claim 1, characterized in that: Generating a corresponding size type detection result for each workpiece in the three-dimensional point cloud, including: Performing point cloud filtering and segmentation operations on the three-dimensional point cloud to obtain a workpiece point cloud for each workpiece; Determining an inclination angle of a point cloud plane equation of each workpiece according to the workpiece point cloud; Determining a minimum bounding box of the workpiece point cloud according to the point cloud plane equation, and calculating the volume of the workpiece point cloud according to the minimum bounding box; A size type detection result of each workpiece is determined according to the area of ​​the workpiece point cloud, the volume of the workpiece point cloud and the inclination angle of the point cloud plane equation.

8. The baking temperature adjustment method for a coating production line according to claim 1, characterized in that: The coating production line is provided with a workpiece suspension rack, a three-dimensional laser camera and a plurality of two-dimensional cameras. The workpiece suspension area is the area below the workpiece suspension rack. All the two-dimensional cameras are in the same plane. The shooting direction of the three-dimensional laser camera is perpendicular to the shooting direction of each of the two-dimensional cameras. Accordingly, a two-dimensional image of the workpiece hanging area of ​​the coating production line is obtained, including: Using all of the two-dimensional cameras to capture partial images of the workpiece hanging area of ​​the coating production line from multiple directions, and stitching the partial images to obtain the two-dimensional image; Accordingly, obtaining a three-dimensional point cloud of the workpiece suspension area includes: Controlling the three-dimensional laser camera to move to preset positions in sequence, and acquiring a local point cloud of the workpiece hanging area at the preset positions; The local point clouds are spliced ​​to obtain a three-dimensional point cloud of the workpiece hanging area.

9. The method for regulating the baking temperature of a coating production line according to claim 1, characterized in that: Before determining the corresponding target baking temperature according to the total mass of the workpieces in the batch, the method further includes: Determine a mapping relationship between the total mass of the workpiece and the baking temperature based on historical data, establish a temperature prediction model based on the mapping relationship, and optimize the temperature prediction model using a gradient descent algorithm; Accordingly, determining a corresponding target baking temperature according to the total mass of the workpieces in the batch, and controlling the coating production line to perform a corresponding baking operation according to the target baking temperature, includes: Inputting the total mass of the workpieces in this batch into the temperature prediction model to obtain the target baking temperature; The opening of the natural gas regulating valve is adjusted according to the target baking temperature, so as to control the hot air blower to perform a baking operation on the workpieces in the coating production line according to the target baking temperature.

10. A baking temperature control system for a coating production line, characterized in that: include: A two-dimensional image processing module is used to obtain a two-dimensional image of the workpiece hanging area of ​​the coating production line and generate a workpiece category detection result for each workpiece in the two-dimensional image using a target detection algorithm model; a three-dimensional point cloud processing module, configured to obtain a three-dimensional point cloud of the workpiece suspension area and generate a corresponding size type detection result for each workpiece in the three-dimensional point cloud; A mass calculation module, configured to determine the total mass of the batch of workpieces based on a workpiece category detection result and / or a size type detection result of each workpiece in the workpiece suspension area; The temperature control module is used to determine the corresponding target baking temperature according to the total mass of the workpieces in this batch, and control the coating production line to perform the corresponding baking operation according to the target baking temperature.

11. An electronic device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps of the baking temperature adjustment method for the coating production line according to any one of claims 1 to 9 are implemented.

12. A storage medium, characterized in that: The storage medium stores computer-executable instructions, which, when loaded and executed by the processor, implement the steps of the baking temperature adjustment method for a coating production line according to any one of claims 1 to 9.

Images