Data mining methods and deep learning systems for smart industrial services

Abstract

The present invention relates to the field of data processing technology, and specifically provides a data mining method and deep learning system applied to smart industrial services. The present invention innovatively extracts material flow state characteristics and equipment operation quantitative characteristics by deeply mining the production factors in the automated production line image data, providing a rich data basis for the optimization of production line efficiency, and by retrieving the initial screening production line optimization annotation images that are compatible with these characteristics, two initial screening sets are formed to ensure the pertinence and feasibility of the optimization plan. At the same time, the concept of common weights for efficiency optimization is introduced to comprehensively evaluate the value of each initial screening annotation image for the current production line efficiency optimization, so as to accurately mine the target production line optimization annotation image for direct guidance of the efficiency optimization and upgrade of the industrial production line, which not only improves the accuracy and efficiency of the optimization process, but also brings significant economic benefits and competitiveness improvement to smart manufacturing enterprises.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing technology, and more specifically, to a data mining method and a deep learning system applied to smart industrial services. Background Art

[0002] In the current industrial manufacturing field, the efficient operation of automated production lines is one of the core goals pursued by enterprises. However, traditional production line optimization methods often rely on manual experience and limited historical data, making it difficult to fully and accurately tap into the potential optimization space of production lines. In particular, when faced with complex material flow conditions and equipment operation conditions, there is a lack of an effective technical means to systematically analyze and guide optimization. This not only limits the further improvement of production line efficiency, but also increases the operating and time costs of enterprises. Therefore, there is an urgent need for a new technology that can deeply mine automated production line image data, accurately match optimization solutions, and guide the optimization and upgrading of production line efficiency. Summary of the invention

[0003] The embodiments of the present invention at least provide a data mining method and a deep learning system applied to smart industrial services.

[0004] The embodiment of the present invention provides a data mining method for smart industrial services, which is applied to a deep learning system. The method includes: performing production factor mining on automated production line image data to obtain material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data; retrieving a number of preliminary screening production line optimization annotation images whose corresponding material flow state characteristics have an adaptation relationship with the material flow state characteristics of the automated production line image data to obtain a first preliminary screening production line optimization annotation image set, and retrieving a number of preliminary screening production line optimization annotation images whose corresponding equipment operation quantitative characteristics have an adaptation relationship with the equipment operation quantitative characteristics of the automated production line image data. annotate an image to obtain a second preliminary screening production line optimized annotated image set; determine the common weight of performance optimization between each preliminary screening production line optimized annotated image in the first preliminary screening production line optimized annotated image set and the second preliminary screening production line optimized annotated image set and the automated production line image data; based on the determined common weight of performance optimization, mine a number of target production line optimized annotated images that have an adaptation relationship with the automated production line image data from the first preliminary screening production line optimized annotated image set and the second preliminary screening production line optimized annotated image set; the target production line optimized annotated image is used to guide the performance optimization and upgrade of the industrial production line corresponding to the automated production line image data.

[0005] An embodiment of the present invention also provides a deep learning system, including a processor and a memory; the processor and the memory are communicatively connected, and the processor is used to read a computer program from the memory and execute it to implement the above method.

[0006] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. The computer program implements the above method when running.

[0007] The technical solution provided by the embodiment of the embodiment of the present invention may include the following beneficial effects: the present invention innovatively extracts the material flow state characteristics and the quantitative characteristics of equipment operation by deeply mining the production factors in the automated production line image data, thereby providing a rich data basis for the optimization of production line efficiency. Furthermore, by retrieving the initial screening production line optimization annotation images that have an adaptation relationship with these features, two initial screening sets are formed, ensuring the pertinence and feasibility of the optimization scheme. On this basis, the present invention introduces the concept of common weights for efficiency optimization, comprehensively evaluates the value of each initial screening annotation image for the current production line efficiency optimization, and thus accurately mines the target production line optimization annotation images. These target images directly guide the efficiency optimization and upgrading of industrial production lines, which not only improves the accuracy and efficiency of the optimization process, but also brings significant economic benefits and competitiveness improvements to intelligent manufacturing companies.

[0008] For a description of the effects of the above-mentioned deep learning system and computer-readable storage medium, please refer to the description of the above-mentioned method.

[0009] In order to make the above-mentioned purposes, features and advantages of the embodiments of the present invention more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the embodiments of the present invention and are used together with the specification to illustrate the technical solutions of the embodiments of the present invention. It should be understood that the following drawings only illustrate certain embodiments of the embodiments of the present invention and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 It is a block diagram of a deep learning system shown in an embodiment of the present invention.

[0012] Figure 2 It is a flow chart of a data mining method applied to smart industrial services shown in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] Here, exemplary embodiments will be described in detail, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the embodiments of the present invention. Instead, they are only examples of devices and methods consistent with some aspects of the embodiments of the present invention.

[0014] Figure 1 The structural diagram of the deep learning system 10 provided in the embodiment of the present invention includes a processor 102, a memory 104, and a bus 106. Among them, the memory 104 is used to store execution instructions, including internal memory and external memory. The internal memory can also be understood as an internal memory, which is used to temporarily store the operation data in the processor 102 and the data exchanged with the external memory such as the hard disk. The processor 102 exchanges data with the external memory through the internal memory. When the deep learning system 10 is running, the processor 102 communicates with the memory 104 through the bus 106, so that the processor 102 executes the data mining method applied to the smart industrial service of the embodiment of the present invention.

[0015] Please combine Figure 2 , Figure 2 It is a flow chart of a data mining method applied to smart industrial services provided by an embodiment of the present invention, which is applied to a deep learning system. The method may exemplarily include the following steps 202 to 208.

[0016] Step 202: Perform production factor mining on the automated production line image data to obtain material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data.

[0017] In the application scenario of smart industrial IoT, the deep learning system first needs to mine the production factors of the automated production line image data to obtain the material flow state characteristics and equipment operation quantitative characteristics. Through the deep learning algorithm, the deep learning system can identify key information such as the material type, quantity, flow speed, and equipment operation status and load conditions in the image.

[0018] For example, in an automated painting production line at a certain automobile manufacturer, the deep learning system can monitor the flow status of the car bodies on the production line in real time by analyzing the image data captured by the high-definition cameras on the production line, including the spacing between the car bodies, the flow speed, and whether there is accumulation or stagnation. At the same time, the deep learning system can also identify the operating status of the painting equipment, such as the spraying speed and spraying quality of the spraying robot, and equipment failures.

[0019] Step 204, retrieve several preliminary screening production line optimized annotated images whose corresponding material flow state characteristics are adaptively matched with the material flow state characteristics of the automated production line image data, to obtain a first preliminary screening production line optimized annotated image set, and retrieve several preliminary screening production line optimized annotated images whose corresponding equipment operation quantitative characteristics are adaptively matched with the equipment operation quantitative characteristics of the automated production line image data, to obtain a second preliminary screening production line optimized annotated image set.

[0020] After obtaining the production factors (material flow state characteristics and equipment operation quantitative characteristics) of the automated production line image data, the deep learning system will further retrieve the initial screening production line optimized annotated images that are compatible with these production factors, namely: the first initial screening production line optimized annotated image set and the second initial screening production line optimized annotated image set.

[0021] For the first screening production line optimization annotated image set, the deep learning system will search for annotated images that match the material flow state characteristics in the current production line image data. These annotated images may come from historical production line optimization cases, showing the best production line layout, material distribution solutions, etc. under different material flow conditions.

[0022] For the optimized annotated image set of the second primary screening production line, the deep learning system will retrieve annotated images that match the quantitative characteristics of the current equipment operation. These images may show the best maintenance strategies and parameter adjustment plans under different equipment operating conditions.

[0023] Continuing with the example of the automated painting production line of an automobile manufacturer, the deep learning system can retrieve those production line layout images with smooth body flow and no accumulation from the historical database as part of the first preliminary screening production line optimization annotation image set. At the same time, the deep learning system will also retrieve those parameter setting images with stable operation of the painting equipment and low failure rate as part of the second preliminary screening production line optimization annotation image set.

[0024] Step 206: determine the performance optimization commonality weight between each primary screening production line optimized annotation image in the first primary screening production line optimized annotation image set and the second primary screening production line optimized annotation image set and the automated production line image data.

[0025] After obtaining two sets of annotated images for initial screening production line optimization, the deep learning system will further determine the common weights of performance optimization between these images and the current automated production line image data. The purpose of this step is to evaluate the potential value of each initial screening annotated image for the current production line performance optimization.

[0026] The commonality weight of performance optimization can be calculated based on multiple factors, including the similarity between the annotated image and the current production line image data, the historical success rate of the optimization solution represented by the annotated image, and the expected cost-effectiveness ratio of implementing the optimization solution.

[0027] In the example of an automobile manufacturing plant, the deep learning system can calculate the similarity score of each initial screening annotated image with the current painting production line image data, and evaluate the success rate and cost-effectiveness of these optimization solutions based on historical data. Ultimately, the deep learning system will assign a performance optimization commonality weight to each initial screening annotated image to guide subsequent production line optimization decisions.

[0028] Step 208: Based on the determined common weights for performance optimization, a number of target production line optimized annotated images that are compatible with the automated production line image data are mined from the first preliminary screening production line optimized annotated image set and the second preliminary screening production line optimized annotated image set; the target production line optimized annotated images are used to guide the performance optimization and upgrade of the industrial production line corresponding to the automated production line image data.

[0029] Finally, based on the determined common weights of performance optimization, the deep learning system will mine several target production line optimization annotated images that are compatible with the current automated production line image data from the two initial screening production line optimization annotated image sets. These target images represent the most promising solutions for optimizing the performance of the current production line.

[0030] Taking the coating production line of an automobile manufacturer as an example, the deep learning system can select those labeled images with high similarity scores, high success rates and reasonable cost-effectiveness ratios from the two initial screening sets as the target production line optimization labeled images. The target production line optimization labeled images may show specific production line layout adjustment plans, equipment parameter optimization strategies or material distribution improvement plans.

[0031] Ultimately, these target production line optimization annotation images will be used to guide the efficiency optimization and upgrade of the automated coating production line. Production line managers can make actual production line adjustments and improvements based on the optimization solutions shown in these images, in order to achieve improved production efficiency, reduced costs, and enhanced overall competitiveness.

[0032] In summary, through the application of deep learning systems in the efficiency optimization of automated production line image data, intelligent manufacturing companies can achieve accurate mining of production line production factors, intelligent recommendation of optimization plans, and effective guidance of efficiency optimization.

[0033] In an automated production line, the material flow state is one of the key factors affecting production efficiency. Through image recognition technology, the deep learning system can capture and analyze the material flow state on the production line in real time. Taking the automated painting production line of an automobile manufacturer as an example, the material flow state characteristics can include the following aspects:

[0034] (1) Vehicle spacing characteristics: The deep learning system can identify the average spacing between vehicles in the field of view of the high-definition camera that continuously passes through the production line. For example, the system may calculate that the average value of the vehicle spacing is 1.5 meters and the standard deviation is 0.2 meters, which means that the spacing between vehicles is relatively stable, but there is a certain fluctuation;

[0035] (2) Flow velocity characteristics: The system can also measure the flow velocity of the car body on the production line. For example, by tracking the position change of the same car body in two consecutive frames of images, the system can calculate that the average flow velocity of the car body is 0.8 m / s, and the velocity fluctuation range is ±0.1 m / s;

[0036] (3) Accumulation and stagnation characteristics: The deep learning system can identify whether there is accumulation or stagnation of car bodies on the production line. For example, the system may detect that the residence time of car bodies in a certain area exceeds a preset threshold (such as 5 seconds), thereby determining that there is a potential accumulation or stagnation problem in the area.

[0037] In addition to the material flow status, the operation status of the equipment is also an important factor affecting the efficiency of the production line. The deep learning system can also extract the quantitative operation characteristics of the equipment through image analysis technology. Taking coating equipment as an example, the quantitative operation characteristics of the equipment may include:

[0038] (1) Spraying speed characteristics: The system can identify the spraying speed of the spraying robot. For example, by analyzing the movement trajectory and spraying width of the spraying robot arm, the system can calculate that the average spraying speed is 0.6 square meters per second, and the speed fluctuation range is ±0.05 square meters per second;

[0039] (2) Spraying quality characteristics: The deep learning system can also evaluate the spraying quality of the spraying robot. For example, the system may give a spraying quality score by analyzing the uniformity and glossiness of the sprayed surface. The score ranges from 0 to 100, where 100 indicates the best spraying quality.

[0040] (3) Equipment failure characteristics: The system can identify the failure status of the equipment. For example, by analyzing the vibration and sound signals generated when the equipment is running, the system can determine whether the equipment has potential failures, such as bearing wear, motor overheating, etc.

[0041] It can be seen that the deep learning system can extract the material flow state characteristics and equipment operation quantitative characteristics in the automated production line through image recognition and analysis technology. These characteristics provide strong data support for subsequent production line optimization. In practical applications, these feature data will be used to guide the adjustment of production line layout, optimization of equipment parameters, and improvement of material distribution strategies, in order to achieve improved production efficiency and reduced costs.

[0042] Furthermore, an example of a material flow state feature vector may be:

[0043] Vehicle spacing feature vector: [1.5, 0.2] / / average value 1.5 meters, standard deviation 0.2 meters;

[0044] Flow velocity characteristic vector: [0.8, 0.1] / / average value 0.8 m / s, fluctuation range ±0.1 m / s;

[0045] Pile-up and stagnation feature vector: [0, 1] / / 0 means no pile-up, 1 means stagnation is detected (or vice versa, depending on the specific encoding).

[0046] An example of a device operation quantification feature vector may be:

[0047] Spraying speed characteristic vector: [0.6, 0.05] / / average value 0.6 m2 / s, fluctuation range ±0.05 m2 / s;

[0048] Spraying quality feature vector:

[90] / / Spraying quality score, ranging from 0 to 100, here is 90 points;

[0049] Equipment fault feature vector: [0, 1, 0] / / 0 indicates no bearing wear, 1 indicates motor overheating is detected, and 0 indicates no other faults (coded according to the specific fault type).

[0050] In practical applications, the above feature vectors will be used as inputs to machine learning or deep learning algorithms to make production line optimization decisions. The algorithm can predict production line performance, recommend optimization strategies, or identify potential problems based on these feature vectors. Through continuous learning and optimization, the algorithm will be able to more accurately guide the adjustment of production line layout, the optimization of equipment parameters, and the improvement of material distribution strategies, thereby improving production efficiency and reducing costs.

[0051] It is worth mentioning that the technical solutions recorded in step 204 and step 206 are the core invention points of this application, and they together constitute an automated production line image data efficiency optimization framework based on a deep learning system.

[0052] First, in step 204, the deep learning system will retrieve the optimized annotated images of the initial screening production line that are compatible with the production elements of the automated production line image data. The key to this step is the determination of the "adaptation relationship". For the material flow state characteristics, the deep learning system can look for historical annotated images whose body spacing, flow speed, and accumulation and stagnation states are similar to those of the current production line. For example, if the average body spacing of the current production line is 1.5 meters and the standard deviation is 0.2 meters, the deep learning system will look for historical annotated images with body spacing within this range. Similarly, for the quantitative characteristics of equipment operation, the deep learning system will also retrieve historical annotated images that match the current equipment spraying speed, spraying quality, and fault status.

[0053] Secondly, in step 206, the deep learning system will further determine the commonalities of performance optimization between these initial screening production line optimization annotated images and the current automated production line image data. The purpose of this step is to evaluate the potential value of each initial screening annotated image for the current production line performance optimization. The commonality weight of performance optimization is a numerical indicator that comprehensively considers multiple factors. It may include the similarity between the annotated image and the current production line image data, the historical success rate of the optimization solution represented by the annotated image, and the expected cost-effectiveness ratio of implementing the optimization solution. For example, a labeled image may have a similarity of up to 90% with the current production line image data, but the historical success rate of the optimization solution it represents is only 50%, and the expected cost-effectiveness ratio is not high, then its commonality weight of performance optimization will be relatively low.

[0054] For example, let the commonality weight of performance optimization be W, which can be determined by the following three factors:

[0055] 1. Similarity score (S): indicates the similarity between the annotated image and the current production line image data. The value range is [0, 1]. The higher the value, the higher the similarity.

[0056] 2. Historical success rate (H): indicates the historical success rate of the optimization solution represented by the annotated image. The value range is [0, 1]. The higher the value, the higher the success rate.

[0057] 3. Expected cost-effectiveness ratio (C): It indicates the expected cost-effectiveness of implementing the optimization plan. It can be measured by the ratio of cost savings to input costs. The theoretical value range is (0, +∞), but in practical applications it can be normalized to the range of [0, 1] for ease of calculation.

[0058] Taking these three factors into consideration, the following performance optimization commonality weight calculation formula can be designed:

[0059] W=Q1*S+Q2*H+Q3*C;

[0060] Among them, Q1, Q2 and Q3 are weight coefficients, indicating the proportion of each factor in the total weight, and Q1+Q2+Q3=1. Among them, the weight coefficient can be adjusted according to the actual application scenario.

[0061] For example, the common weights for performance optimization of the following three annotated images need to be calculated:

[0062] Image A: [image similarity score S, historical success rate H, expected cost-effectiveness ratio C (normalized)] = [0.85, 0.75, 0.80];

[0063] Image B: [image similarity score S, historical success rate H, expected cost-effectiveness ratio C (normalized)] = [0.90, 0.65, 0.90];

[0064] Image C: [image similarity score S, historical success rate H, expected cost-effectiveness ratio C (normalized)] = [0.75, 0.85, 0.70].

[0065] Assuming the weight coefficients are Q1=0.4, Q2=0.3, Q3=0.3, the performance optimization commonality weights of each annotated image can be calculated according to the formula:

[0066] The performance optimization commonality weight of image A is: 0.4×0.85+0.3×0.75+0.3×0.80=0.805;

[0067] The performance optimization commonality weight of image B is: 0.4×0.90+0.3×0.65+0.3×0.90=0.835;

[0068] The performance optimization commonality weight of image C is: 0.4×0.75+0.3×0.85+0.3×0.70=0.765.

[0069] It can be seen that according to the calculation results, the performance optimization commonality weight of image B is the highest, so it can be the most promising solution for optimizing the performance of the current production line.

[0070] In this way, the concept of common weights for performance optimization can be understood and applied more specifically, thereby more effectively guiding the performance optimization and upgrading of automated production lines.

[0071] Then, the deep learning system will comprehensively consider the common weights of performance optimization of all the pre-screened production line optimization annotated images, and mine several target production line optimization annotated images that have the highest adaptation relationship with the current automated production line image data. These target images represent the most promising solutions for optimizing the performance of the current production line. For example, the deep learning system can select annotated images with the top 10% of the common weights of performance optimization as target production line optimization annotated images.

[0072] Therefore, the technical solutions recorded in step 204 and step 206 realize accurate mining of automated production line image data and intelligent recommendation of optimization solutions through a deep learning system, which not only improves the accuracy and efficiency of production line performance optimization, but also provides strong data support and decision-making basis for intelligent manufacturing companies.

[0073] In addition, the technical solution recorded in step 208 is also the core invention of the present application, which realizes the effective use of the optimized annotation images of the target production line, thereby providing specific guidance for the efficiency optimization of the automated production line.

[0074] First, step 208 involves an in-depth analysis of the target production line optimized annotated images. These images are the optimized annotated images that have the highest fit with the current automated production line image data after being screened out in the previous steps. They represent the most promising solutions for optimizing the performance of the current production line. For example, if the current production line faces the problem of uneven body flow speed, then the target production line optimized annotated images may be those production line images that have successfully solved similar problems in the past.

[0075] Secondly, the deep learning system extracts the key optimization parameters in the target production line optimization annotation image. These parameters are specific values ​​or strategies to guide the optimization of production line performance. They may include the layout adjustment plan of the production line, the operating parameter setting of the equipment, the optimization of the material distribution strategy, etc. For example, a key optimization parameter may be "adjusting the spraying speed of the coating equipment to 0.7 square meters per second to reduce waste in the spraying process and improve production efficiency."

[0076] The deep learning system will then compare and analyze these key optimization parameters with the actual operating parameters of the current automated production line. The purpose of this step is to evaluate the feasibility and expected results of the optimization plan. For example, if the spraying speed of the current production line is 0.6 square meters per second, and the spraying speed in the target production line optimization annotation image is 0.7 square meters per second, then the system will analyze whether it is feasible to increase the spraying speed to 0.7 square meters per second, and the possible production efficiency improvement and cost savings brought about by this adjustment.

[0077] Finally, based on the results of comparison and analysis, the deep learning system will generate a performance optimization recommendation report for the current automated production line. This report will list in detail the specific content of the optimization plan, expected results, implementation difficulty and cost, and other key information to provide decision support for production line managers. For example, the report may point out: "By adjusting the spraying speed of the coating equipment to 0.7 square meters per second, it is expected that production efficiency can be increased by 5%, while reducing waste in the spraying process. It is recommended to try it in the next production cycle."

[0078] Therefore, the technical solution recorded in step 208 deeply analyzes the target production line optimization annotation image, extracts key optimization parameters, and compares and analyzes them with the actual operating parameters of the current production line, and finally generates a performance optimization recommendation report for the current automated production line. This solution provides specific and feasible production line optimization guidance for smart manufacturing companies, which helps to improve production efficiency and reduce costs.

[0079] In summary, the present invention deeply mines the production factors in the automated production line image data, innovatively extracts the material flow state characteristics and equipment operation quantitative characteristics, and provides a rich data foundation for the optimization of production line efficiency. Furthermore, by retrieving the initial screening production line optimization annotation images that are compatible with these characteristics, two initial screening sets are formed to ensure the pertinence and feasibility of the optimization scheme. On this basis, the present invention introduces the concept of common weights for efficiency optimization, comprehensively evaluates the value of each initial screening annotation image for the current production line efficiency optimization, and accurately mines the target production line optimization annotation images. These target images directly guide the efficiency optimization and upgrading of industrial production lines, which not only improves the accuracy and efficiency of the optimization process, but also brings significant economic benefits and competitiveness improvements to intelligent manufacturing companies.

[0080] In actual application, the technical solutions described in the above steps 202 to 208 can also be applied to the following scenarios.

[0081] (1) In the smart manufacturing workshop, the smart IoT technology uses sensors and cameras throughout the workshop to capture real-time information such as the operating status of production equipment, material flow, and environmental parameters. In step 202, the system will conduct in-depth analysis of these image data to extract key production factor features. Subsequently, the system will retrieve historical optimized annotated images based on these features and evaluate their common weights of performance optimization to discover optimization solutions suitable for the current workshop. Finally, by generating an optimization recommendation report, the workshop is guided to upgrade to intelligentization and improve production efficiency and quality.

[0082] (2) In smart warehousing and logistics centers, IoT technology is widely used in cargo tracking, inventory management, and intelligent scheduling. The system obtains the location, quantity, and status information of goods in real time through devices such as RFID tags, barcodes, and cameras. The system then analyzes these image data, identifies bottlenecks and optimization points in warehousing and logistics, and generates targeted optimization solutions based on historical optimization annotation images. Finally, these solutions are converted into specific operation guidelines to help warehousing and logistics centers achieve automated and intelligent management and improve logistics efficiency and accuracy.

[0083] (3) The deep learning system uses the Internet of Things technology to monitor and manage energy facilities such as power systems, heating systems, and refrigeration systems in real time. The system collects equipment operation data and environmental parameters such as temperature, humidity, and power consumption through sensors. The system then analyzes this data to identify energy waste and potential safety hazards, and based on historical optimization and annotation images, it proposes suggestions for energy saving and consumption reduction and optimizing energy distribution. Finally, these suggestions are transformed into specific implementation plans to help companies achieve efficient energy utilization and sustainable development.

[0084] (4) A smart industrial park is a comprehensive park that integrates multiple functions such as production, office, R&D, and logistics. IoT technology is widely used in the park, including environmental monitoring, safety management, equipment operation and maintenance, and other aspects. Through various sensors and cameras deployed in the park, the system can obtain real-time information such as park environmental data, equipment operating status, and personnel flow. Then, the system will analyze this data, identify problems and optimization points in park management, and propose comprehensive optimization solutions based on historical optimization annotation images. Finally, these solutions are transformed into specific implementation plans to help park managers achieve intelligent management and improve park operation efficiency and safety.

[0085] In other optional embodiments, the material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data are obtained by performing production factor mining on the automated production line image data through a target production factor mining algorithm; the method also includes: obtaining a deep learning network to be debugged, and generating a production factor mining algorithm based on the deep learning network to be debugged; debugging the production factor mining algorithm based on the learning examples corresponding to the production factor mining algorithm to obtain the target production factor mining algorithm.

[0086] Furthermore, the method also includes: performing a noise addition operation on selected image blocks in the historical production line image data to obtain historical noise image data, and obtaining a target image block corresponding to the noise distribution of the historical noise image data from the historical production line image data; parsing the image block corresponding to the noise distribution of the historical noise image data through an image block analysis network to obtain a discrimination score of the target image block corresponding to the noise distribution generated by the image block analysis network; improving the network configuration weights of the image block analysis network based on the discrimination score to obtain a target image block analysis network, and using the target image block analysis network as the deep learning network to be debugged.

[0087] In the present invention, in order to solve the problem of optimizing the efficiency of the automated coating production line, the target production factor mining algorithm can be used to deeply mine the automated production line image data, thereby extracting key material flow state characteristics and equipment operation quantitative characteristics.

[0088] First, the deep learning system needs to obtain the deep learning network to be debugged, and generate an algorithm specifically for production factor mining based on the network. This algorithm needs to be debugged through the learning samples corresponding to the production factor mining algorithm to obtain the target production factor mining algorithm that can accurately mine the material flow state characteristics and equipment operation quantitative characteristics.

[0089] In order to further improve the accuracy and robustness of the algorithm, the deep learning system adds noise to the historical production line image data to obtain historical noise image data, and obtains the target image blocks corresponding to the noise distribution. Then, the image blocks corresponding to these noise distributions are analyzed through the image block analysis network to obtain the discrimination score that the image blocks corresponding to the noise distribution are the target image blocks. Based on this discrimination score, the deep learning system improves the network configuration weights of the image block analysis network, obtains the target image block analysis network, and uses it as part of the deep learning network to be debugged.

[0090] In the specific implementation process, the deep learning system first conducts an in-depth analysis of the automated coating production line to identify the types and goals of production factors that need to be mined. Then, based on these goals and needs, it selects appropriate deep learning networks and learning examples for debugging and optimization. The deep learning system uses a large amount of historical production line image data as training samples, and through continuous training and adjustment, it gradually improves the accuracy and efficiency of the production factor mining algorithm.

[0091] At the same time, in order to improve the robustness and practical application value of the algorithm, the deep learning system adds noise to the historical production line image data. This step enables the algorithm to accurately mine production factors even in a noisy environment, thereby improving the practicality and reliability of the algorithm. In the noise addition operation, the deep learning system uses a variety of different noise models to simulate various interference factors in the actual production line environment. Then, the deep learning system inputs the noisy image data into the production factor mining algorithm for testing and optimization until the algorithm can stably mine production factors in a noisy environment.

[0092] After obtaining the target production factor mining algorithm, the deep learning system applies it to the actual automated coating production line image data to mine production factors and optimize production line efficiency. Through this series of steps and methods, the deep learning system realizes the deep mining and utilization of automated coating production line image data, providing strong support for the optimization of production line efficiency of intelligent manufacturing enterprises.

[0093] In some alternative embodiments, the material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data are obtained by performing production factor mining on the automated production line image data using a target production factor mining algorithm; the method also includes: obtaining a priori automated production line image data, and a priori target production line optimized annotation image corresponding to the priori automated production line image data; performing production factor mining on the priori automated production line image data and the priori target production line optimized annotation image respectively using a production factor mining algorithm to obtain the material flow state characteristics and equipment operation quantitative characteristics of the priori automated production line image data, and the material flow state characteristics and equipment operation quantitative characteristics of the priori target production line optimized annotation image; determining the material flow of the priori target production line optimized annotation image. The efficiency optimization commonality weight between the state characteristics and the material flow state characteristics of the prior automated production line image data is obtained to obtain the material processing efficiency optimization commonality weight corresponding to the prior target production line optimization annotation image, and the efficiency optimization commonality weight between the equipment operation quantitative characteristics of the prior target production line optimization annotation image and the equipment operation quantitative characteristics of the prior automated production line image data is determined to obtain the equipment operation efficiency optimization commonality weight corresponding to the prior target production line optimization annotation image; based on the material processing efficiency optimization commonality weight and the equipment operation efficiency optimization commonality weight corresponding to the prior target production line optimization annotation image, a factor mining error is constructed to obtain the target factor mining offset coefficient, and based on the target factor mining offset coefficient, the production factor mining algorithm is improved to obtain the target production factor mining algorithm.

[0094] It can be understood that this embodiment involves deep mining of automated production line image data through a target production factor mining algorithm, extracting key material flow state characteristics and equipment operation quantitative characteristics, thereby providing strong support for production line efficiency optimization.

[0095] First, the deep learning system needs to obtain the prior automated production line image data and the prior target production line optimization annotation images corresponding to these image data. These prior data are the basis for the deep learning system to mine production factors and train algorithms. Through the production factor mining algorithm, the deep learning system mines production factors on the prior automated production line image data and the prior target production line optimization annotation images respectively. The purpose of this step is to extract the material flow state characteristics and equipment operation quantitative characteristics from the image data, and provide basic data for subsequent analysis and optimization.

[0096] After obtaining the material flow state characteristics and equipment operation quantitative characteristics of the prior automated production line image data and the prior target production line optimization annotated image, the deep learning system further analyzes the common weights of efficiency optimization between these characteristics. Specifically, the deep learning system determines the common weights of efficiency optimization between the material flow state characteristics of the prior target production line optimization annotated image and the material flow state characteristics of the prior automated production line image data, and obtains the common weights of material handling efficiency optimization corresponding to the prior target production line optimization annotated image. Similarly, the deep learning system also determines the common weights of efficiency optimization between the equipment operation quantitative characteristics of the prior target production line optimization annotated image and the equipment operation quantitative characteristics of the prior automated production line image data, and obtains the common weights of equipment operation efficiency optimization corresponding to the prior target production line optimization annotated image.

[0097] The determination of these two commonality weights is a key step in the technical solution of the deep learning system. They reflect the degree of correlation between the production factor features in the prior target production line optimization annotation image and the production factor features in the prior automated production line image data. Through these two commonality weights, the deep learning system can evaluate the importance of the production factor features in the production line optimization annotation image for the optimization of production line efficiency and provide guidance for subsequent algorithm improvements.

[0098] Next, the deep learning system constructs the factor mining error based on the common weights of material handling efficiency optimization and equipment operation efficiency optimization corresponding to the prior target production line optimization annotation image. This error reflects the gap between the production factor mining algorithm and the ideal optimization state when extracting features. In order to reduce this error, the deep learning system needs to improve the production factor mining algorithm.

[0099] Therefore, the deep learning system obtains the target factor mining offset coefficient based on the factor mining error. This offset coefficient is a key parameter for algorithm improvement. It guides the deep learning system on how to adjust the production factor mining algorithm to more accurately extract features that have an important impact on the optimization of production line efficiency. Through continuous adjustment and optimization, the deep learning system finally obtains the target production factor mining algorithm. This algorithm can more accurately extract material flow state features and equipment operation quantitative features from the automated production line image data, providing stronger support for the optimization of production line efficiency.

[0100] In the specific implementation process, the deep learning system first conducts in-depth analysis and processing of the prior automated production line image data and the prior target production line optimization annotation images. The deep learning system uses image preprocessing technology to enhance the quality of the image and improve the accuracy of feature extraction. Then, the deep learning system selects a suitable deep learning network as the basis of the production factor mining algorithm, and optimizes the network's configuration weights through a large amount of training and adjustment.

[0101] During the training process, the deep learning system used a large amount of historical production line image data as training samples, and improved the accuracy and efficiency of the production factor mining algorithm through continuous iteration and optimization. At the same time, the deep learning system also introduced regularization technology and dropout strategy to prevent overfitting and improve the generalization ability of the algorithm.

[0102] In order to verify the effectiveness of the target production factor mining algorithm, the deep learning system applied it to the actual automated coating production line image data to mine production factors and optimize production line efficiency. By comparing the production line efficiency indicators before and after optimization, the deep learning system found that the target production factor mining algorithm can significantly improve the operating efficiency and coating utilization of the production line, and reduce energy consumption and scrap rate.

[0103] In addition, the deep learning system further analyzes and processes the mined production factors. For example, the deep learning system presents the flow state of the paint through data visualization technology, helping engineers better understand the operating status and existing problems of the production line. At the same time, the deep learning system also conducts time series analysis on the quantitative characteristics of equipment operation, discovers the periodic laws and potential failure modes of equipment operation, and provides strong support for preventive maintenance and fault diagnosis of equipment.

[0104] In this way, by introducing a priori automated production line image data and a priori target production line optimization annotation images, as well as production factor mining algorithms and common weight analysis methods, deep mining and utilization of automated painting production line image data are achieved. This embodiment not only improves the accuracy and efficiency of production line efficiency optimization, but also brings significant economic benefits and competitiveness improvements to enterprises. At the same time, by introducing target factor mining offset coefficients and algorithm improvement strategies, the accuracy and robustness of the production factor mining algorithm are further improved, providing new ideas and methods for production line upgrades and efficiency optimization of intelligent manufacturing enterprises.

[0105] In some preferred embodiments, the method of constructing a factor mining error based on the common weights of material processing efficiency optimization and the common weights of equipment operation efficiency optimization corresponding to the prior target production line optimization annotated image to obtain a target factor mining offset coefficient includes: constructing a factor mining error based on the common weights of material processing efficiency optimization corresponding to the prior target production line optimization annotated image to obtain a material factor mining offset coefficient, and constructing a factor mining error based on the common weights of equipment operation efficiency optimization corresponding to the prior target production line optimization annotated image to obtain an equipment factor mining offset coefficient; integrating the common weights of material processing efficiency optimization and the common weights of equipment operation efficiency optimization corresponding to the prior target production line optimization annotated image to obtain the multi-mode efficiency optimization common weights corresponding to the prior target production line optimization annotated image, and constructing a factor mining error based on the multi-mode efficiency optimization common weights corresponding to the prior target production line optimization annotated image to obtain a multi-mode factor mining offset coefficient; determining the target factor mining offset coefficient based on the material factor mining offset coefficient, the equipment factor mining offset coefficient and the multi-mode factor mining offset coefficient.

[0106] Based on this embodiment, the deep learning system deeply explores how to construct the element mining error based on the common weights of material processing efficiency optimization and equipment operation efficiency optimization corresponding to the prior target production line optimization annotation image, and then obtain the target element mining offset coefficient. This process involves multiple key steps and strategies to ensure that the deep learning system can accurately extract the production factor features that are critical to the optimization of production line efficiency from the image data of the automated coating production line.

[0107] First, the deep learning system constructs the factor mining error based on the common weights of material handling efficiency optimization and equipment operation efficiency optimization corresponding to the prior target production line optimization annotation image. Specifically, the deep learning system uses the common weights of material handling efficiency optimization to evaluate the accuracy of the production factor mining algorithm in extracting material flow state characteristics, and constructs the material factor mining error based on this. Similarly, the deep learning system also uses the common weights of equipment operation efficiency optimization to evaluate the accuracy of the algorithm in extracting quantitative characteristics of equipment operation, and constructs the equipment factor mining error. These two errors respectively reflect the gap between the algorithm and the ideal optimization state when extracting different types of production factor characteristics.

[0108] Next, the deep learning system integrates the common weights of material handling efficiency optimization and equipment operation efficiency optimization corresponding to the prior target production line optimization annotated images, and obtains the common weights of multi-modal efficiency optimization. The purpose of this step is to combine different types of production factor characteristics (i.e., material flow state characteristics and equipment operation quantitative characteristics) to more comprehensively evaluate the potential for production line efficiency optimization. Based on this multi-modal efficiency optimization common weight, the deep learning system further constructs a multi-modal factor mining error, which reflects the overall accuracy of the production factor mining algorithm when considering both material and equipment characteristics.

[0109] After obtaining the material factor mining error, equipment factor mining error, and multi-mode factor mining error, the deep learning system faces the challenge of how to determine the target factor mining offset coefficient. This offset coefficient is the key to algorithm improvement. It will guide the deep learning system on how to adjust the production factor mining algorithm to more accurately extract features that have an important impact on the optimization of production line efficiency. To this end, the deep learning system comprehensively considers the above three errors, and through certain mathematical operations and strategy adjustments, it finally determines the target factor mining offset coefficient.

[0110] In the specific implementation process, the deep learning system uses a variety of advanced technologies and methods to ensure the accuracy and effectiveness of the target factor mining offset coefficient. For example, the deep learning system uses a deep learning network to extract the production factor features in the image data, and optimizes the performance of the network through a lot of training and adjustment. At the same time, the deep learning system also introduces regularization technology and dropout strategy to prevent overfitting and improve the generalization ability of the algorithm. In addition, the deep learning system also uses strategies such as cross-validation and grid search to determine the optimal network structure and parameter configuration.

[0111] In order to verify the effectiveness of the target factor mining offset coefficient, the deep learning system applied it to the actual automated coating production line image data to mine production factors and optimize production line efficiency. By comparing the production line efficiency indicators before and after optimization (such as paint utilization, energy consumption, scrap rate, etc.), the deep learning system found that the target factor mining offset coefficient can significantly improve the accuracy and efficiency of the production factor mining algorithm, thereby improving the overall efficiency of the production line. Specifically, the optimized production line can more accurately control the flow state of the paint and the operating state of the equipment during operation, reduce unnecessary waste and loss, and improve product quality and output.

[0112] In addition, the deep learning system further analyzes and processes the mined production factors. For example, the deep learning system presents the flow state of the paint through data visualization technology, helping engineers better understand the operating status and existing problems of the production line. At the same time, the deep learning system also conducts time series analysis on the quantitative characteristics of equipment operation, discovers the periodic laws and potential failure modes of equipment operation, and provides strong support for preventive maintenance and fault diagnosis of equipment. These further analysis and processing work not only verifies the effectiveness of the target factor mining offset coefficient, but also provides enterprises with more ideas and methods for production line optimization and efficiency improvement.

[0113] It can be seen that by introducing the common weights of material handling efficiency optimization and equipment operation efficiency optimization corresponding to the prior target production line optimization annotation image, as well as constructing factor mining errors and determining target factor mining offset coefficients, the deep mining and utilization of automated coating production line image data is achieved. By introducing multi-mode efficiency optimization common weights and comprehensively considering the characteristics of different types of production factors, the comprehensiveness and robustness of the production factor mining algorithm are further improved, providing new ideas and methods for production line upgrades and efficiency optimization of intelligent manufacturing enterprises.

[0114] In other examples, the a priori target production line optimization and annotation image includes a first a priori production line optimization and annotation image that has an adaptation relationship with the a priori automated production line image data, and a second a priori production line optimization and annotation image that has no adaptation relationship with the a priori automated production line image data; the factor mining error constructed based on the material processing efficiency optimization common weights corresponding to the a priori target production line optimization and annotation image to obtain the material factor mining offset coefficient includes: constructing a factor mining error based on the material processing efficiency optimization common weights corresponding to the first a priori production line optimization and annotation image and the material processing efficiency optimization common weights corresponding to the second a priori production line optimization and annotation image to obtain the material factor mining offset coefficient, wherein the material factor mining offset coefficient has a first quantitative relationship with the material processing efficiency optimization common weights corresponding to the first a priori production line optimization and annotation image, and has a second quantitative relationship with the material processing efficiency optimization common weights corresponding to the second a priori production line optimization and annotation image.

[0115] It can be understood that this embodiment further explores the application of the prior target production line optimized annotation image in the production factor mining of the automated painting production line image data. In particular, the deep learning system subdivides the prior target production line optimized annotation image into two categories: one is the first prior production line optimized annotation image that has an adaptation relationship with the prior automated production line image data, and the other is the second prior production line optimized annotation image that has no adaptation relationship with the prior automated production line image data. This classification aims to gain a deeper understanding of the impact of different types of optimized annotation images on the production factor mining error and the material factor mining offset coefficient.

[0116] First, the deep learning system focuses on how to construct the factor mining error based on the common weights of material handling efficiency optimization corresponding to these two types of prior production line optimization annotated images, and then obtain the material factor mining offset coefficient. To achieve this goal, the deep learning system considers the common weights of material handling efficiency optimization corresponding to the first prior production line optimization annotated image and the second prior production line optimization annotated image, respectively. By incorporating these common weights into the construction process of the factor mining error, the deep learning system can more accurately evaluate the performance of the production factor mining algorithm in extracting material flow state characteristics.

[0117] When constructing the element mining error, the deep learning system pays special attention to the difference between the first prior production line optimization annotation image and the second prior production line optimization annotation image. Since these two types of images have different adaptation relationships with the prior automated production line image data, the commonality weights for material handling efficiency optimization they provide should also have different weights or influences when constructing the element mining error. Therefore, the deep learning system describes the relationship between the material element mining offset coefficient and these two types of commonality weights by introducing the first quantitative relationship and the second quantitative relationship.

[0118] Specifically, the first quantitative relationship describes the relationship between the material factor mining offset coefficient and the material processing efficiency optimization commonality weight corresponding to the first prior production line optimization annotation image. Since the first prior production line optimization annotation image has an adaptation relationship with the prior automated production line image data, the commonality weight it provides has a greater influence when constructing the factor mining error. Therefore, the first quantitative relationship reflects this influence and helps the deep learning system understand how to adjust the production factor mining algorithm according to the commonality weight of the first prior production line optimization annotation image to improve the accuracy of extracting material flow state features.

[0119] In contrast, the second quantitative relationship describes the relationship between the material factor mining offset coefficient and the material processing efficiency optimization commonality weight corresponding to the second prior production line optimization annotation image. Since the second prior production line optimization annotation image does not have an adaptation relationship with the prior automated production line image data, the commonality weight it provides may have little influence or even cause certain interference when constructing the factor mining error. Therefore, the second quantitative relationship helps the deep learning system understand how to deal with this interference, and when necessary, make corresponding adjustments to the production factor mining algorithm to reduce the error introduced by the second prior production line optimization annotation image.

[0120] After obtaining the material factor mining offset coefficient, the deep learning system further applies it to the actual automated coating production line image data to mine production factors and optimize production line efficiency. By comparing the production line efficiency indicators before and after optimization, the deep learning system found that the material factor mining offset coefficient can effectively guide the update and optimization of the production factor mining algorithm, thereby improving the production line efficiency. Specifically, the optimized production line can better control the flow state of the paint, reduce waste and loss, and improve paint utilization and product quality.

[0121] In addition, the deep learning system further analyzes and processes the mined production factors. For example, the deep learning system presents the flow state of the paint through data visualization technology, helping engineers to more intuitively understand the operating status and existing problems of the production line. At the same time, the deep learning system also conducts time series analysis on the flow state characteristics of the paint, discovers the periodic laws and potential problem patterns of the paint flow, and provides strong support for further optimization and preventive maintenance of the production line.

[0122] In addition to the application of material factor mining offset coefficients, the deep learning system also explored how to integrate and optimize it with other types of factor mining offset coefficients (such as equipment factor mining offset coefficients). By comprehensively considering the characteristics of different types of production factors and their corresponding offset coefficients, the deep learning system can more comprehensively evaluate and optimize the overall performance of the production line. This comprehensive evaluation and optimization method not only improves the accuracy and efficiency of the production factor mining algorithm, but also brings more significant economic benefits and competitiveness improvements to enterprises.

[0123] By introducing prior production line optimization annotation images that have different adaptation relationships with prior automated production line image data, and constructing factor mining errors and determining material factor mining offset coefficients based on the common weights of material processing efficiency optimization corresponding to these images, deep mining and utilization of automated painting production line image data are achieved. This embodiment not only improves the accuracy and robustness of the production factor mining algorithm, but also provides enterprises with more ideas and methods for production line optimization and efficiency improvement. By comprehensively considering the characteristics of different types of production factors and their corresponding offset coefficients, the deep learning system further improves the comprehensiveness and practicality of the production factor mining algorithm, and provides new solutions for production line upgrades and efficiency optimization of intelligent manufacturing enterprises.

[0124] For example, the material element mining offset coefficient can be calculated by the following formula:

[0125] DeltaM=alpha*W1+beta*(1-W2)

[0126] in:

[0127] DeltaM represents the material element mining offset coefficient;

[0128] alpha is a coefficient that indicates the degree of influence of W1 on DeltaM;

[0129] W1 is the commonality weight of material handling efficiency optimization corresponding to the first a priori production line optimization annotation image;

[0130] beta is a coefficient that indicates the degree of influence of W2 on DeltaM;

[0131] W2 is the commonality weight of material handling efficiency optimization corresponding to the second prior production line optimization annotation image;

[0132] (1-W2) is used to represent the potential negative impact on DeltaM when the second prior production line optimization annotation image has no matching relationship with the prior automated production line image data.

[0133] In some other alternative embodiments, the production factor mining of the automated production line image data to obtain the material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data includes: obtaining a pixel cluster sequence corresponding to the automated production line image data, and extracting pixel cluster wavelet descriptors corresponding to several pixel clusters contained in the pixel cluster sequence; wherein the pixel cluster wavelet descriptor of each pixel cluster contains the associated pixel semantics of each pixel cluster in the pixel cluster sequence; and extracting the equipment operation quantitative characteristics of the automated production line image data based on the pixel cluster wavelet descriptors corresponding to the several pixel clusters.

[0134] It should be noted that this embodiment proposes an innovative technical solution, which aims to extract material flow state characteristics and equipment operation quantitative characteristics from complex production line image data through sophisticated image analysis methods, providing strong data support for subsequent production optimization.

[0135] First, this embodiment focuses on the acquisition and processing of image data of automated painting production lines. At this stage, the core task is to capture image data during the operation of the production line and convert it into a format that can be used for subsequent analysis and processing. Specifically, the system uses high-precision image acquisition equipment to capture various operating scenes on the painting production line in real time, including the spraying of paint, the flow of materials, the operation of equipment, etc. These image data are then converted into a sequence of pixel clusters, that is, a series of interrelated pixel sets, each of which contains specific information about the operating status of the production line.

[0136] Secondly, in order to extract valuable information from the pixel cluster sequence, this embodiment introduces the concept of pixel cluster wavelet descriptor. This is an innovative image feature description method that can accurately characterize the associated pixel semantics of each pixel cluster in the pixel cluster sequence. In other words, the pixel cluster wavelet descriptor not only reflects the properties of the pixel cluster itself, but also reveals the spatial and temporal relationship between it and other pixel clusters. This description method greatly enriches the expressive power of image data and lays a solid foundation for subsequent feature extraction.

[0137] After extracting the pixel cluster wavelet descriptors, this embodiment further uses these descriptors to mine the quantitative characteristics of equipment operation in the image data of the automated painting production line. This step is a key link in the technical solution, which is directly related to the accuracy and effectiveness of the subsequent production line efficiency optimization. By deeply analyzing the information contained in the pixel cluster wavelet descriptors, the system can accurately identify the operating status of various equipment in the production line, including their operating speed, load conditions, failure modes, etc. These quantitative characteristics of equipment operation provide valuable reference information for subsequent production line optimization.

[0138] Specifically, in the process of extracting the quantitative characteristics of equipment operation, the system first determines the location and layout of each device in the production line based on the spatial relationship information in the pixel cluster wavelet descriptor. Then, by analyzing the temporal relationship information in the descriptor, the system can track the changes in the equipment's operating status and calculate the equipment's operating speed and load. In addition, the system can also use the abnormal pattern information in the descriptor to detect possible equipment failures or abnormal conditions, providing early warnings for timely maintenance and repairs.

[0139] It is worth mentioning that this embodiment is not limited to the extraction of quantitative characteristics of equipment operation. After deeply mining the production line image data, the system can further extract the material flow state characteristics. This feature is of great significance for optimizing material management of the coating production line, reducing waste and improving coating utilization. By analyzing the material flow pattern and information in the pixel cluster wavelet descriptor, the system can accurately depict the flow path, speed and distribution state of the coating on the production line. This information provides strong data support for subsequent material flow optimization.

[0140] After extracting the material flow state characteristics and equipment operation quantitative characteristics, this embodiment further applies these characteristics to the actual production line efficiency optimization. By comparing the production line efficiency indicators before and after optimization, the system can accurately evaluate the effect of feature extraction and optimization. After applying this embodiment, the material flow of the coating production line is smoother, the equipment operation state is more stable, and the overall production efficiency has been significantly improved.

[0141] In addition to direct performance improvement, this embodiment also brings more convenience and benefits to the production management of enterprises. For example, through data visualization technology, the system can present the complex production line operation status to engineers and managers in an intuitive way, helping them better understand the actual operation of the production line. At the same time, the system can also perform time series analysis on the extracted features to discover the periodic laws and potential problem patterns of production line operation, providing strong support for preventive maintenance and optimization.

[0142] In summary, this embodiment achieves in-depth mining and utilization of automated painting production line image data through innovative image analysis methods. It not only extracts rich material flow state characteristics and equipment operation quantitative characteristics, but also provides strong data support for subsequent production line efficiency optimization. The proposal and application of this technical solution not only improves the production efficiency and quality of automobile manufacturers, but also brings new ideas and directions for the development of the field of intelligent manufacturing.

[0143] Specifically, when extracting the material flow state characteristics, the system makes full use of the dynamic information and correlation analysis in the pixel cluster wavelet descriptor. By modeling the spatiotemporal evolution of the pixel cluster sequence, the system is able to capture the flow trajectory, speed changes, and distribution patterns of the paint on the production line. These characteristics are crucial to understanding the material flow state of the production line. They reveal the dynamic behavior of the paint during the spraying process and the key factors that may affect the paint utilization and product quality.

[0144] At the same time, the system also combines machine learning algorithms and pattern recognition technology when extracting quantitative characteristics of equipment operation. By extracting and classifying the characteristics of pixel cluster wavelet descriptors, the system can automatically identify different types of equipment in the production line and monitor their operating status. This automated equipment monitoring method not only improves production efficiency, but also reduces production problems caused by manual monitoring errors or omissions.

[0145] In addition, this embodiment also takes into account the complexity and variability of the production line environment. In practical applications, the image data of the coating production line may be affected by a variety of factors such as lighting conditions, paint color, equipment layout, etc. In order to meet these challenges, the system uses a robust algorithm when extracting features, and performs a lot of verification and tuning. This ensures that under different environmental conditions, the system can accurately extract the material flow state characteristics and equipment operation quantitative characteristics.

[0146] In some further optional embodiments, the step of obtaining a pixel cluster sequence corresponding to the automated production line image data comprises: performing pixel-level decomposition processing on the automated production line image data to obtain an original pixel cluster sequence, and associating a foreground image block at a key pixel label of the original pixel cluster sequence to obtain the pixel cluster sequence; the step of extracting quantitative equipment operation features of the automated production line image data based on pixel cluster wavelet descriptors corresponding to the plurality of pixel clusters comprises: using the pixel cluster wavelet descriptor corresponding to the foreground image block as the quantitative equipment operation features of the automated production line image data.

[0147] It can be understood that this embodiment aims to mine the production factors in the automated coating production line through in-depth image data analysis, thereby achieving quantitative evaluation of the equipment operating status and optimization of the production line efficiency.

[0148] First, this embodiment focuses on the refined processing of automated painting production line image data. In this process, the system performs pixel-level splitting on the production line image data, which is the basis for subsequent analysis. Through this process, the system can decompose complex production line images into a series of original pixel clusters, each of which contains specific information about the production line's operating status. This pixel-level splitting process not only retains detailed information in the image, but also facilitates subsequent feature extraction.

[0149] However, simply obtaining the original pixel cluster sequence is not enough to fully describe the operating status of the production line. In order to further enhance the expressiveness of the image data, this embodiment performs association processing of key pixel labels based on the original pixel cluster sequence. The core of this step is to locate those pixel clusters in the original pixel cluster sequence that are crucial to describing the operating status of the production line, and associate them with the foreground image blocks. The foreground image blocks are those areas in the production line image that can directly reflect the operating status of the equipment, such as the nozzles of the spraying equipment, the gears of the transmission device, etc. Through this association processing, the system can more accurately capture the key operating information in the production line.

[0150] After obtaining the pixel cluster sequence associated with the foreground image block, this embodiment further uses the pixel cluster wavelet descriptor to extract the quantitative characteristics of equipment operation in the production line image data. The pixel cluster wavelet descriptor is an innovative image feature description method that can accurately describe the spatiotemporal relationship of each pixel cluster in the pixel cluster sequence, including its shape, texture, motion trajectory, etc. This description method not only retains the properties of the pixel cluster itself, but also reveals the complex relationship between it and other pixel clusters.

[0151] When extracting the quantitative characteristics of equipment operation, this embodiment pays special attention to the pixel cluster wavelet descriptors corresponding to the foreground image blocks. This is because the foreground image blocks directly reflect the operating status of the equipment in the production line, so the pixel cluster wavelet descriptors corresponding to them contain rich quantitative information about the equipment operation. By deeply analyzing these descriptors, the system can accurately extract key quantitative characteristics such as the equipment's operating speed, load conditions, and failure modes.

[0152] Specifically, the system first uses the spatial relationship information in the pixel cluster wavelet descriptor to determine the location and layout of each device in the production line. This step is crucial to understanding the overall operating status of the production line. Then, by analyzing the temporal relationship information in the descriptor, the system can track the changes in the operating status of the equipment and calculate the operating speed and load of the equipment. This equipment operating status monitoring method based on image data is not only highly accurate, but also can achieve real-time monitoring and feedback.

[0153] In addition to the running speed and load conditions, this embodiment can also use the abnormal pattern information in the pixel cluster wavelet descriptor to detect possible faults or abnormal conditions of the equipment. This function is of great significance for preventive equipment maintenance and production line optimization. By timely discovering and handling potential fault problems, the system can effectively reduce the downtime of the production line and improve the overall production efficiency.

[0154] After extracting the quantitative characteristics of equipment operation, this embodiment further applies these characteristics to the actual production line efficiency optimization. By comparing the production line efficiency indicators before and after optimization, such as coating utilization, product quality, production line downtime, etc., the system can accurately evaluate the effect of feature extraction and optimization. In addition to direct performance improvement, this embodiment also brings more convenience and benefits to the production management of the enterprise. For example, through data visualization technology, the system can present complex production line operation status to engineers and managers in an intuitive way. This helps them better understand the actual operation of the production line, so as to formulate a more scientific and reasonable production optimization strategy. At the same time, the system can also perform time series analysis on the extracted quantitative characteristics of equipment operation to discover the periodic laws and potential problem patterns of production line operation. This provides strong support for preventive maintenance and optimization.

[0155] In addition, this embodiment also fully considers the complexity and variability of the production line environment. In practical applications, the image data of the coating production line may be affected by a variety of factors such as lighting conditions, paint color, equipment layout, etc. In order to meet these challenges, the system uses a robust algorithm when extracting features, and performs a lot of verification and tuning. This ensures that the system can accurately extract the quantitative characteristics of equipment operation under different environmental conditions.

[0156] In a preferred embodiment, the extraction of pixel cluster wavelet descriptors corresponding to the plurality of pixel clusters contained in the pixel cluster sequence comprises: obtaining a target production factor mining algorithm; wherein the target production factor mining algorithm comprises a multi-order production factor mining branch, wherein the multi-order production factor mining branch comprises a first-order production factor mining branch and remaining production factor mining branches except the first-order production factor mining branch; extracting the pixel cluster wavelet descriptors corresponding to the plurality of pixel clusters from the pixel cluster sequence through the first-order production factor mining branch, and obtaining a pixel cluster wavelet descriptor sequence generated by the first-order production factor mining branch; Through each order of residual production factor mining branch, the pixel cluster wavelet descriptors corresponding to the several pixel clusters are extracted from the pixel cluster wavelet descriptor sequence generated by the previous order production factor mining branch of the each order of residual production factor mining branch to obtain the pixel cluster wavelet descriptor sequence generated by the each order of residual production factor mining branch; then, based on the pixel cluster wavelet descriptors corresponding to the several pixel clusters, the quantitative characteristics of the equipment operation of the automated production line image data are extracted, including: based on the pixel cluster wavelet descriptor sequence generated by the last order production factor mining branch in the multi-order production factor mining branch, the quantitative characteristics of the equipment operation of the automated production line image data are extracted.

[0157] It can be understood that this embodiment uses a multi-order production factor mining algorithm to conduct in-depth analysis of the image data of the automated painting production line and extract rich quantitative characteristics of equipment operation.

[0158] First, this embodiment focuses on the in-depth analysis and feature extraction of pixel cluster sequences. In this process, the deep learning system introduces a target production factor mining algorithm, which has a multi-order production factor mining branch, including a first-order production factor mining branch and a remaining production factor mining branch other than the first order. The design of this multi-order structure aims to gradually extract key information from the image data in a layer-by-layer manner, thereby providing strong support for the subsequent quantitative feature extraction of device operation.

[0159] Specifically, through the first-order production factor mining branch, the deep learning system extracts pixel cluster wavelet descriptors corresponding to several pixel clusters from the pixel cluster sequence. This step is the basis of feature extraction. It can convert complex image data into a series of descriptors with clear physical meanings, which facilitates subsequent analysis. After obtaining the pixel cluster wavelet descriptor sequence generated by the first-order production factor mining branch, the deep learning system further extracts deeper feature information from the descriptor sequence generated by the previous-order production factor mining branch through each order of residual production factor mining branches. This step-by-step deepening process not only retains the key information in the image data, but also further explores the inherent connections and laws between the data.

[0160] In the multi-order production factor mining algorithm, each order production factor mining branch undertakes different tasks. The first-order production factor mining branch focuses on the basic features in the image data, such as edges, textures, etc. The subsequent remaining production factor mining branches pay more attention to the complex relationships and patterns between data, such as the operating status and failure mode of the equipment. Through this layer-by-layer approach, the deep learning system can gradually mine the deep information in the image data and provide strong support for the quantitative evaluation of the equipment operating status.

[0161] Then, based on the pixel cluster wavelet descriptor sequence generated by the last-order production factor mining branch in the multi-order production factor mining branch, the deep learning system extracts the quantitative characteristics of the equipment operation of the automated production line image data. This step is the core of the technical solution. It can convert the previously extracted feature information into specific quantitative indicators, thereby achieving an accurate assessment of the equipment's operating status. Through these quantitative characteristics, the deep learning system can clearly understand key information such as the equipment's operating speed, load conditions, and failure modes, providing strong support for subsequent production line optimization and decision-making.

[0162] In practical applications, this embodiment has shown significant advantages. First, through the multi-order production factor mining algorithm, the deep learning system can more deeply mine the key information in the image data, improving the accuracy and comprehensiveness of feature extraction. Secondly, through a layer-by-layer approach, the deep learning system can gradually understand the inherent laws and patterns of the data, providing more powerful support for the quantitative evaluation of the equipment operation status. Finally, through the extracted quantitative characteristics of equipment operation, the deep learning system can realize real-time monitoring and optimization of the production line operation status, improving the overall production efficiency and product quality.

[0163] By applying this embodiment, the deep learning system can accurately extract the quantitative characteristics of equipment operation in the production line image data and optimize the production line efficiency. Specifically, at the key equipment of the coating production line, the deep learning system installed a high-definition camera for real-time image acquisition. Then, the collected image data is processed and analyzed using this embodiment. Through verification, the deep learning system found that this technical solution can accurately identify the key equipment in the production line and extract key quantitative characteristics such as the equipment's operating status and failure mode.

[0164] During the verification process, the deep learning system also compared the effects of traditional image analysis methods and this embodiment in feature extraction and production line optimization. The results show that traditional image analysis methods have great limitations in feature extraction and cannot accurately identify key equipment and operating status in the production line. However, this embodiment can deeply mine key information in image data through a multi-order production factor mining algorithm to improve the accuracy and comprehensiveness of feature extraction. At the same time, in terms of production line optimization, this embodiment can also formulate a more scientific and reasonable optimization strategy based on the extracted quantitative characteristics of equipment operation, thereby improving overall production efficiency and product quality.

[0165] In the next step, the production factor mining of the automated production line image data is performed to obtain the material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data, and also includes: based on the pixel cluster wavelet descriptor of each pixel cluster, determining the spatiotemporal variation characteristic value of each pixel cluster in the pixel cluster sequence; based on the spatiotemporal variation characteristic values ​​corresponding to several pixel clusters contained in the pixel cluster sequence, generating the material flow state characteristics of the automated production line image data.

[0166] It can be understood that when this embodiment is applied, the image data of the automated coating production line can be mined for production factors to extract rich material flow state characteristics and equipment operation quantitative characteristics.

[0167] First, this embodiment focuses on the in-depth analysis and feature extraction of pixel cluster sequences. In this process, the deep learning system introduces a pixel cluster wavelet descriptor, which can effectively characterize the spatiotemporal variation characteristics of each pixel cluster in the image data. Based on the pixel cluster wavelet descriptor of each pixel cluster, the deep learning system can determine the spatiotemporal variation characteristic values ​​of the pixel cluster in the pixel cluster sequence. This step is the basis of feature extraction, which can convert complex image data into a series of characteristic values ​​with clear physical meaning, facilitating subsequent analysis.

[0168] Specifically, the pixel cluster wavelet descriptor is extracted through the first-order production factor mining branch in the multi-order production factor mining algorithm. This branch focuses on the basic features in the image data, such as edges, textures, etc., and transforms these features into descriptors with spatiotemporal characteristics through methods such as wavelet transform. After obtaining the pixel cluster wavelet descriptor of each pixel cluster, the deep learning system can further calculate its spatiotemporal variation eigenvalue in the pixel cluster sequence. This eigenvalue can reflect the variation trend of the pixel cluster in time and space, and provide strong support for the subsequent material flow state feature extraction.

[0169] Secondly, based on the spatiotemporal variation feature values ​​corresponding to the pixel clusters contained in the pixel cluster sequence, the deep learning system generates the material flow state features of the automated production line image data. This step is the core of the technical solution, which can convert the previously extracted feature values ​​into specific material flow state features, thereby achieving accurate evaluation of the material flow state of the production line.

[0170] In the process of generating material flow state characteristics, the deep learning system comprehensively considers the temporal and spatial variation characteristics of each pixel cluster in the pixel cluster sequence and the relationship between them. By analyzing these characteristic values, the deep learning system can identify key information such as material flow pattern, speed, direction, etc. in the production line, providing strong support for subsequent production line optimization and decision-making. For example, the deep learning system can determine whether the production line has problems such as blockage and backlog based on the material flow state characteristics, and take corresponding optimization measures in a timely manner.

[0171] Then, in addition to the material flow state characteristics, this embodiment also focuses on the extraction of quantitative characteristics of equipment operation. In the multi-order production factor mining algorithm, the subsequent remaining production factor mining branch undertakes this task. These branches pay more attention to the complex relationships and patterns between data, such as the operating status of equipment, failure mode, etc.

[0172] By going deeper layer by layer, the remaining production factor mining branch can gradually mine the deep information in the image data and convert it into specific quantitative characteristics of equipment operation. These characteristics can clearly reflect key information such as the equipment's operating speed, load conditions, and failure modes, providing strong support for subsequent production line optimization and decision-making. For example, the deep learning system can determine whether the equipment needs maintenance or replacement based on the quantitative characteristics of the equipment's operation, and formulate a corresponding maintenance plan.

[0173] In practical applications, this embodiment has shown significant advantages. First, by introducing the pixel cluster wavelet descriptor, the deep learning system can more accurately characterize the spatiotemporal variation characteristics in the image data, improving the accuracy and comprehensiveness of feature extraction. Secondly, by comprehensively considering the spatiotemporal variation characteristics of each pixel cluster in the pixel cluster sequence and the relationship between them, the deep learning system can generate material flow state characteristics with rich information, providing strong support for production line optimization and decision-making. Finally, by extracting quantitative characteristics of equipment operation, the deep learning system can realize real-time monitoring and optimization of the production line operation status, improving overall production efficiency and product quality.

[0174] In some alternative embodiments, the material flow state characteristics of the automated production line image data are generated based on the spatiotemporal variation characteristic values ​​corresponding to the several pixel clusters included in the pixel cluster sequence, including: determining the characteristic variables corresponding to the several pixel clusters included in the preset pixel feature space based on the preset pixel feature space and the spatiotemporal variation characteristic values ​​corresponding to the several pixel clusters included in the pixel cluster sequence, and generating the material flow state characteristics of the automated production line image data based on the determined characteristic variables; wherein, in the process of determining the characteristic variables of any pixel cluster included in the preset pixel feature space, if the pixel cluster sequence includes the any pixel cluster, the characteristic variables of the any pixel cluster are determined based on the spatiotemporal variation characteristic values ​​of the any pixel cluster in the pixel cluster sequence and the thermal value of the any pixel cluster in the pixel cluster sequence; if the pixel cluster sequence does not include the any pixel cluster, the pre-configured characteristic variables are used as the characteristic variables of the any pixel cluster.

[0175] Furthermore, this embodiment extracts rich material flow state characteristics by performing in-depth analysis on the image data of the automated coating production line.

[0176] First, this embodiment focuses on the in-depth analysis and feature extraction of pixel cluster sequences. In this process, the deep learning system introduces the spatiotemporal variation feature value of the pixel cluster, which can effectively characterize the spatiotemporal variation characteristics of each pixel cluster in the image data. Based on the spatiotemporal variation feature value of each pixel cluster, the deep learning system can further explore its performance in the preset pixel feature space and determine the feature variables corresponding to the several pixel clusters contained in the space.

[0177] The preset pixel feature space is a multidimensional space, in which each dimension represents a specific pixel cluster feature. By determining the characteristic variables of each pixel cluster in the preset pixel feature space, the deep learning system can fully describe the state of material flow in the production line image data. This step is the basis of feature extraction, which can convert complex image data into a series of characteristic variables with clear physical meanings, facilitating subsequent analysis.

[0178] Secondly, in the process of determining the characteristic variables of any pixel cluster contained in the preset pixel feature space, the deep learning system considers two cases. If any pixel cluster is included in the pixel cluster sequence, the deep learning system will determine its characteristic variables based on its spatiotemporal variation characteristic value and thermal value in the pixel cluster sequence. The thermal value is an indicator that reflects the importance and activity of the pixel cluster in the image data. By combining the spatiotemporal variation characteristic value and thermal value, the deep learning system can more accurately describe the characteristics of the pixel cluster in the material flow state.

[0179] Specifically, the deep learning system will fuse the temporal and spatial variation characteristic values ​​and thermal values, and transform them into characteristic variables through certain mathematical transformations and mapping relationships. This characteristic variable can comprehensively reflect the temporal and spatial variation trends of pixel clusters and their importance in the material flow state. In this way, the deep learning system can accurately extract key information related to the material flow state from the production line image data.

[0180] Exemplarily, in order to fuse the spatiotemporal variation characteristic values ​​and thermal values ​​and convert them into characteristic variables through mathematical transformation and mapping, it can be achieved through the following specific implementation details.

[0181] 1. Data preprocessing:

[0182] First, the image data of the automated painting production line is preprocessed, including denoising and contrast enhancement, to ensure the accuracy of subsequent feature extraction.

[0183] Next, the image data is divided into several pixel clusters, and the spatiotemporal variation characteristic value and thermal value of each pixel cluster are calculated.

[0184] 2. Eigenvalue normalization:

[0185] Since the spatiotemporal variation eigenvalues ​​and thermal values ​​may have different dimensions and magnitudes, they need to be normalized to ensure that they have the same weight when fused. For example, a linear normalization method can be used to map the eigenvalues ​​to the interval [0, 1].

[0186] 3. Feature Fusion:

[0187] The weighted summation method is used to fuse the spatiotemporal variation characteristic value and the thermal value. Assuming the spatiotemporal variation characteristic value is F_tc and the thermal value is F_h, the fused characteristic value F_fused can be expressed as:

[0188] F_fused=T1*F_tc+T2*F_h;

[0189] Among them, T1 and T2 are the weights of the spatiotemporal variation characteristic value and thermal value respectively, and T1+T2=1.

[0190] The weights T1 and T2 may be adjusted according to actual conditions, for example, determined through experiments or experience.

[0191] 4. Mathematical transformation and mapping:

[0192] In order to transform the fused eigenvalues ​​F_fused into feature variables, certain mathematical transformations and mapping relationships may be used, such as linear transformation, logarithmic transformation, or power transformation.

[0193] This embodiment can adopt linear transformation, and the characteristic variable F_var can be expressed as:

[0194] F_var=a*F_fused+b;

[0195] Among them, a and b are the coefficients and constant terms of linear transformation, which can be determined according to actual conditions. Specific embodiments

[0197] In combination with the above content, in a specific embodiment, there is image data of an automated coating production line, which has been divided into several pixel clusters, and the spatiotemporal variation characteristic value F_tc and thermal value F_h of each pixel cluster are calculated. The following is a specific embodiment:

[0198] 1. Data preprocessing:

[0199] Perform denoising and contrast enhancement on image data.

[0200] Divide the image into 100 pixel clusters.

[0201] 2. Eigenvalue normalization:

[0202] The linear normalization method is used to map the spatiotemporal variation characteristic values ​​and thermal values ​​into the interval [0, 1].

[0203] For example, for a certain pixel cluster, its spatiotemporal variation characteristic value F_tc=0.6 and thermal value F_h=0.8.

[0204] 3. Feature Fusion:

[0205] Set weights T1=0.5, T2=0.5.

[0206] The fused eigenvalue F_fused=0.5*0.6+0.5*0.8=0.7.

[0207] 4. Mathematical transformation and mapping:

[0208] Set the coefficient a of the linear transformation to 1 and the constant term b to 0.

[0209] Then the characteristic variable F_var=1*0.7+0=0.7.

[0210] Through the above steps, the spatiotemporal variation characteristic values ​​and thermal values ​​can be fused and converted into characteristic variables through mathematical transformation and mapping. In practical applications, parameters such as weights, mathematical transformations, and mapping relationships can be adjusted according to specific circumstances to obtain more accurate feature extraction results.

[0211] If the pixel cluster sequence does not contain any of the pixel clusters, the deep learning system will use the pre-configured feature variables as its feature variables. The pre-configured feature variables are a set of feature values ​​pre-set before the implementation of the technical solution, which represent the characteristics that the pixel cluster may have under normal circumstances. By using pre-configured feature variables, the deep learning system can ensure that even if some pixel clusters are missing in the pixel cluster sequence, it can still generate complete material flow state characteristics.

[0212] Then, based on the determined feature variables, the deep learning system will generate the material flow state features of the automated production line image data. This step is the core of the technical solution, which can convert the previously extracted feature variables into specific material flow state features, thereby achieving accurate evaluation of the material flow state of the production line.

[0213] In the process of generating material flow state characteristics, the deep learning system will comprehensively consider the characteristic variables of each pixel cluster in the preset pixel feature space and the relationship between them. By analyzing these characteristic variables, the deep learning system can identify key information such as material flow pattern, speed, direction, etc. in the production line, providing strong support for subsequent production line optimization and decision-making. For example, the deep learning system can determine whether there are problems such as blockage and backlog in the production line based on the material flow state characteristics, and take corresponding optimization measures in a timely manner.

[0214] In addition, this embodiment also focuses on real-time and accuracy. By collecting and processing the image data of the automated coating production line in real time, the deep learning system can obtain the latest material flow state characteristics in a timely manner and accurately analyze and evaluate them. This helps companies to discover and solve potential production problems in a timely manner and improve production efficiency and product quality.

[0215] In practical applications, this embodiment has shown significant advantages. First, by introducing the temporal and spatial variation characteristic values ​​and thermal values ​​of pixel clusters, the deep learning system can more accurately describe the material flow state characteristics in the image data, improving the comprehensiveness and accuracy of feature extraction. Secondly, by comprehensively considering the characteristic variables of each pixel cluster in the preset pixel feature space and the relationship between them, the deep learning system can generate material flow state characteristics with rich information, providing strong support for production line optimization and decision-making. Finally, by real-time acquisition and processing of image data of the automated coating production line, the deep learning system can realize real-time monitoring and optimization of the material flow state, improving the overall production efficiency and product quality.

[0216] In some independent embodiments, determining the efficiency optimization commonality weight between each primary screening production line optimized annotated image in the first primary screening production line optimized annotated image set and the second primary screening production line optimized annotated image set and the automated production line image data includes: determining the efficiency optimization commonality weight between the material flow state characteristics of each primary screening production line optimized annotated image and the material flow state characteristics of the automated production line image data, and obtaining the material processing efficiency optimization commonality weight corresponding to each primary screening production line optimized annotated image; determining the equipment operation quantitative characteristics of each primary screening production line optimized annotated image and the material flow state characteristics of the automated production line image data; The common weights of efficiency optimization between the quantitative characteristics of equipment operation of the automated production line image data are used to obtain the common weights of equipment operation efficiency optimization corresponding to each of the preliminary screening production line optimized annotated images; the common weights of material handling efficiency optimization and equipment operation efficiency optimization corresponding to each of the preliminary screening production line optimized annotated images are integrated to obtain the multi-mode efficiency optimization common weights corresponding to each of the preliminary screening production line optimized annotated images, and the multi-mode efficiency optimization common weights corresponding to each of the preliminary screening production line optimized annotated images are used as the common weights of efficiency optimization between each of the preliminary screening production line optimized annotated images and the automated production line image data.

[0217] It can be understood that by conducting in-depth analysis of the initial screening production line optimization annotation images and the automated production line image data, the common weights of performance optimization between them can be determined, thereby providing strong data support for subsequent production line optimization.

[0218] First, this embodiment focuses on determining the common weights of efficiency optimization between the optimized annotation images of the initial screening production line and the image data of the automated production line. In this process, the deep learning system introduces two key dimensions: material flow state characteristics and equipment operation quantitative characteristics. By calculating the common weights of efficiency optimization in these two dimensions, a comprehensive evaluation of the production line efficiency is achieved.

[0219] Specifically, the deep learning system first determines the common weights for efficiency optimization between the material flow state characteristics of each initial screening production line optimization annotated image and the material flow state characteristics of the automated production line image data. This step is the key to evaluating the potential for optimizing the material flow state of the production line. By deeply analyzing the material flow state characteristics in the initial screening production line optimization annotated images, the deep learning system can identify those characteristic variables that are of great significance for improving production line efficiency. At the same time, the deep learning system compares these characteristic variables with the material flow state characteristics in the automated production line image data, and by calculating the similarity and correlation between them, determines the common weights for material handling efficiency optimization corresponding to each initial screening production line optimization annotated image.

[0220] Secondly, the deep learning system determines the common weights for performance optimization between the quantitative characteristics of equipment operation in each initial screening production line optimization annotated image and the quantitative characteristics of equipment operation in the automated production line image data. This step is the key to evaluating the potential for optimizing the operating status of production line equipment. By conducting in-depth analysis of the quantitative characteristics of equipment operation in the initial screening production line optimization annotated images, the deep learning system can identify those characteristic variables that are of great significance for improving equipment operation efficiency. At the same time, the deep learning system compares these characteristic variables with the quantitative characteristics of equipment operation in the automated production line image data, and determines the common weights for equipment operation performance optimization corresponding to each initial screening production line optimization annotated image by calculating the similarity and correlation between them.

[0221] After determining the common weights for optimizing material handling efficiency and optimizing equipment operation efficiency corresponding to each primary screening production line optimized annotated image, the deep learning system integrates these two weights to obtain the common weights for optimizing multi-mode efficiency corresponding to each primary screening production line optimized annotated image. This step integrates the optimization potential in two dimensions, material flow status and equipment operation status, to achieve a comprehensive evaluation of the overall efficiency of the production line. By integrating the common weights for optimizing material handling efficiency and the common weights for optimizing equipment operation efficiency, the deep learning system can obtain a more comprehensive and accurate evaluation result, providing strong data support for subsequent production line optimization.

[0222] In the specific implementation process, the deep learning system uses advanced image analysis technology and machine learning algorithms to achieve in-depth mining and analysis of the optimized annotation images of the initial screening production line and the image data of the automated production line. By using these technologies, the deep learning system can accurately extract the key feature variables in the image data and calculate the similarity and correlation between them. At the same time, the deep learning system also combines the actual data and empirical knowledge of the production line operation to further verify and adjust the calculated common weights of efficiency optimization to ensure its accuracy and reliability.

[0223] By applying this embodiment, the deep learning system can achieve accurate optimization of the efficiency of the automated painting production line. First, by determining the common weights of efficiency optimization between the initial screening production line optimization annotation image and the automated production line image data, the deep learning system can identify the optimization potential points in the production line. These optimization potential points may include bottlenecks in material flow, inefficient links in equipment operation, etc. By improving and optimizing these optimization potential points, the deep learning system can improve the overall efficiency of the production line, improve production efficiency and product quality.

[0224] Secondly, this embodiment can also provide strong support for the intelligent transformation of enterprises. Through in-depth mining and analysis of the optimized annotation images of the initial screening production line and the image data of the automated production line, the deep learning system can obtain richer and more comprehensive production line operation data. These data can not only help enterprises better understand the operating status and potential problems of the production line, but also provide enterprises with more scientific and accurate decision-making support. Data-based decision-making methods can improve the production efficiency and quality of enterprises, and enhance the market competitiveness and sustainable development capabilities of enterprises.

[0225] In addition, this embodiment also has certain versatility and scalability. Although this embodiment is described with the automated painting production line of an automobile manufacturer as the application scenario, the technical solution is also applicable to other types of manufacturing production lines and production scenarios. By deeply mining and analyzing image data in different production lines and production scenarios, the deep learning system can achieve accurate optimization of the performance of different production lines, providing support and assistance for more companies in their intelligent transformation.

[0226] It should be noted that the novelty of this embodiment is first reflected in its unique perspective and methodology. Traditionally, production line optimization often relies on manual experience or a single quantitative indicator, but this embodiment takes a different approach and digs out the common weights of efficiency optimization by deeply analyzing the association between the initial screening production line optimization annotation image and the automated production line image data. This perspective is not only novel, but also greatly enriches the data sources and analysis dimensions of production line optimization. Secondly, this embodiment also shows significant innovation in technical means. It not only focuses on single-aspect features such as material flow or equipment operation, but also refines and quantifies these features, and calculates the common weights of efficiency optimization between them and the automated production line image data. This multi-dimensional feature extraction and weight calculation method makes the analysis of production line optimization more comprehensive and in-depth, and can more accurately identify the optimization potential points of the production line. Finally, this embodiment also obtains the multi-mode efficiency optimization common weights corresponding to each initial screening production line optimization annotation image by integrating the common weights of efficiency optimization in different dimensions. This integration step not only improves the accuracy and comprehensiveness of the analysis, but also provides more scientific and systematic data support for subsequent production line optimization. This innovative integration method makes this embodiment have significant advantages and broad application prospects in the field of production line optimization. In summary, this embodiment provides a new idea and methodology for production line optimization through a unique perspective, innovative technical means and multi-dimensional integration method. It not only improves the accuracy and efficiency of production line optimization, but also provides strong support for the intelligent transformation and sustainable development of enterprises.

[0227] Furthermore, a readable storage medium is provided, on which a program is stored, and the program implements the above method when executed by a processor.

[0228] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described in the above exemplary embodiment can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

Claims

1. A data mining method applied to smart industrial services, characterized in that: The method is implemented by a deep learning system, and the method includes: Performing production factor mining on the automated production line image data to obtain material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data; Retrieving a number of preliminary screening production line optimized annotated images whose corresponding material flow state characteristics are in an adaptive relationship with the material flow state characteristics of the automated production line image data, to obtain a first preliminary screening production line optimized annotated image set, and retrieving a number of preliminary screening production line optimized annotated images whose corresponding equipment operation quantitative characteristics are in an adaptive relationship with the equipment operation quantitative characteristics of the automated production line image data, to obtain a second preliminary screening production line optimized annotated image set; Determine a performance optimization commonality weight between each primary screening production line optimized annotation image in the first primary screening production line optimized annotation image set and the second primary screening production line optimized annotation image set and the automated production line image data; Based on the determined common weights for performance optimization, a number of target production line optimization annotation images that are compatible with the automated production line image data are mined from the first preliminary screening production line optimization annotation image set and the second preliminary screening production line optimization annotation image set; the target production line optimization annotation images are used to guide the performance optimization and upgrade of the industrial production line corresponding to the automated production line image data; The production factor mining of the automated production line image data to obtain the material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data includes: Acquire a pixel cluster sequence corresponding to the automated production line image data, and extract pixel cluster wavelet descriptors corresponding to a plurality of pixel clusters contained in the pixel cluster sequence; wherein the pixel cluster wavelet descriptor of each pixel cluster contains the associated pixel semantics of each pixel cluster in the pixel cluster sequence; Based on the pixel cluster wavelet descriptors corresponding to the plurality of pixel clusters, respectively, the quantitative characteristics of equipment operation of the automated production line image data are extracted.

2. The method according to claim 1, characterized in that The material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data are obtained by performing production factor mining on the automated production line image data using a target production factor mining algorithm; the method further includes: Acquire a deep learning network to be debugged, and generate a production factor mining algorithm based on the deep learning network to be debugged; Based on the learning examples corresponding to the production factor mining algorithm, the production factor mining algorithm is debugged to obtain the target production factor mining algorithm.

3. The method according to claim 2, characterized in that The method further comprises: Performing a noise addition operation on a selected image block in the historical production line image data to obtain historical noise image data, and obtaining a target image block corresponding to the noise distribution of the historical noise image data from the historical production line image data; Analyzing the image block corresponding to the noise distribution of the historical noise image data through an image block analysis network, and obtaining the discriminant score of the image block corresponding to the noise distribution generated by the image block analysis network as the target image block; The network configuration weights of the image block analysis network are improved based on the discriminant score to obtain a target image block analysis network, and the target image block analysis network is used as the deep learning network to be debugged.

4. The method according to claim 1, characterized in that The material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data are obtained by performing production factor mining on the automated production line image data using a target production factor mining algorithm; the method further includes: Acquire a priori automated production line image data and a priori target production line optimized annotated image corresponding to the a priori automated production line image data; The production factor mining algorithm is used to respectively mine the a priori automated production line image data and the a priori target production line optimized annotated image to obtain material flow state characteristics and equipment operation quantitative characteristics of the a priori automated production line image data, and material flow state characteristics and equipment operation quantitative characteristics of the a priori target production line optimized annotated image; Determine the efficiency optimization commonality weight between the material flow state characteristics of the a priori target production line optimization annotation image and the material flow state characteristics of the a priori automated production line image data, and obtain the material processing efficiency optimization commonality weight corresponding to the a priori target production line optimization annotation image; and determine the efficiency optimization commonality weight between the equipment operation quantitative characteristics of the a priori target production line optimization annotation image and the equipment operation quantitative characteristics of the a priori automated production line image data, and obtain the equipment operation efficiency optimization commonality weight corresponding to the a priori target production line optimization annotation image; Based on the common weights of material processing efficiency optimization and equipment operation efficiency optimization corresponding to the prior target production line optimization annotation image, construct a factor mining error to obtain a target factor mining offset coefficient, and improve the production factor mining algorithm based on the target factor mining offset coefficient to obtain the target production factor mining algorithm; The method of constructing the element mining error based on the common weights of material processing efficiency optimization and equipment operation efficiency optimization corresponding to the prior target production line optimization annotation image to obtain the target element mining offset coefficient includes: Based on the material processing efficiency optimization commonality weight corresponding to the prior target production line optimization annotation image, the element mining error is constructed to obtain the material element mining offset coefficient, and based on the equipment operation efficiency optimization commonality weight corresponding to the prior target production line optimization annotation image, the element mining error is constructed to obtain the equipment element mining offset coefficient; Integrate the common weights of material handling efficiency optimization and equipment operation efficiency optimization corresponding to the a priori target production line optimization annotation image to obtain the multi-mode efficiency optimization common weights corresponding to the a priori target production line optimization annotation image, and construct a factor mining error based on the multi-mode efficiency optimization common weights corresponding to the a priori target production line optimization annotation image to obtain the multi-mode factor mining offset coefficient; Determine the target element mining offset coefficient based on the material element mining offset coefficient, the equipment element mining offset coefficient and the multi-mode element mining offset coefficient; The a priori target production line optimization annotation image includes a first a priori production line optimization annotation image that has an adaptation relationship with the a priori automated production line image data, and a second a priori production line optimization annotation image that has no adaptation relationship with the a priori automated production line image data; the material processing efficiency optimization commonality weight based on the a priori target production line optimization annotation image corresponding to construct the element mining error, and obtain the material element mining offset coefficient, including; Based on the common weights of material processing efficiency optimization corresponding to the first prior production line optimization and annotation image and the common weights of material processing efficiency optimization corresponding to the second prior production line optimization and annotation image, an element mining error is constructed to obtain the material element mining offset coefficient, wherein the material element mining offset coefficient has a first quantitative relationship with the common weights of material processing efficiency optimization corresponding to the first prior production line optimization and annotation image, and has a second quantitative relationship with the common weights of material processing efficiency optimization corresponding to the second prior production line optimization and annotation image.

5. The method according to claim 1, characterized in that The step of obtaining the pixel cluster sequence corresponding to the automated production line image data includes: performing pixel-level splitting processing on the automated production line image data to obtain an original pixel cluster sequence, and associating a foreground image block at a key pixel label of the original pixel cluster sequence to obtain the pixel cluster sequence; The extracting of the equipment operation quantitative features of the automated production line image data based on the pixel cluster wavelet descriptors corresponding to the plurality of pixel clusters respectively includes: using the pixel cluster wavelet descriptors corresponding to the foreground image blocks as the equipment operation quantitative features of the automated production line image data.

6. The method according to claim 1, characterized in that The step of extracting pixel cluster wavelet descriptors corresponding to the plurality of pixel clusters contained in the pixel cluster sequence respectively comprises: Obtaining a target production factor mining algorithm; wherein the target production factor mining algorithm includes a multi-order production factor mining branch, wherein the multi-order production factor mining branch includes a first-order production factor mining branch and remaining production factor mining branches except the first-order production factor mining branch; Extracting pixel cluster wavelet descriptors corresponding to the plurality of pixel clusters respectively from the pixel cluster sequence through the first-order production factor mining branch, and obtaining a pixel cluster wavelet descriptor sequence generated by the first-order production factor mining branch; By using each order of residual production factor mining branch, pixel cluster wavelet descriptors corresponding to the plurality of pixel clusters are extracted from a pixel cluster wavelet descriptor sequence generated by a previous order production factor mining branch of the each order of residual production factor mining branch, to obtain a pixel cluster wavelet descriptor sequence generated by the each order of residual production factor mining branch; The extracting of the equipment operation quantitative features of the automated production line image data based on the pixel cluster wavelet descriptors corresponding to the plurality of pixel clusters respectively includes: Based on the pixel cluster wavelet descriptor sequence generated by the last-order production factor mining branch in the multi-order production factor mining branch, the quantitative characteristics of equipment operation of the automated production line image data are extracted.

7. The method according to claim 1, characterized in that The performing of production factor mining on the automated production line image data to obtain material flow state characteristics and equipment operation quantitative characteristics of the automated production line image data further includes: Determine the spatiotemporal variation characteristic value of each pixel cluster in the pixel cluster sequence based on the pixel cluster wavelet descriptor of each pixel cluster; Generating material flow state characteristics of the automated production line image data based on the spatiotemporal variation characteristic values ​​respectively corresponding to the plurality of pixel clusters included in the pixel cluster sequence; The generating of the material flow state characteristics of the automated production line image data based on the spatiotemporal variation characteristic values ​​respectively corresponding to the plurality of pixel clusters included in the pixel cluster sequence comprises: Based on a preset pixel feature space and the spatiotemporal variation feature values ​​respectively corresponding to the several pixel clusters included in the pixel cluster sequence, feature variables respectively corresponding to the several pixel clusters included in the preset pixel feature space are determined, and based on the determined feature variables, material flow state features of the automated production line image data are generated; Wherein, in the process of determining the characteristic variables of any pixel cluster contained in the preset pixel characteristic space, if the pixel cluster sequence contains the any pixel cluster, the characteristic variables of the any pixel cluster are determined based on the spatiotemporal variation characteristic value of the any pixel cluster in the pixel cluster sequence and the thermal value of the any pixel cluster in the pixel cluster sequence; If the pixel cluster sequence does not include any pixel cluster, the pre-configured feature variable is used as the feature variable of any pixel cluster.

8. A deep learning system, characterized in that: It comprises a processor and a memory; the processor and the memory are communicatively connected, and the processor is used to read a computer program from the memory and execute it to implement the method described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program implements the method described in any one of claims 1 to 7 when running.

Images