Industrial process remote supervision method and system based on computer vision

Abstract

The invention relates to the technical field of industrial supervision, in particular to an industrial process remote supervision method and system based on computer vision, and the method comprises the following steps: collecting video data representing an industrial equipment process practical operation process; synchronously combining the industrial process standard identification model and the industrial process efficiency identification model to obtain an industrial process supervision model; and synchronously obtaining the real-time standard and the real-time efficiency of the industrial practical operation process of the industrial equipment according to the real-time data video by utilizing the industrial process supervision model so as to realize the real-time supervision of the industrial process. According to the method, the industrial process supervision model is constructed, and the real-time standard and real-time efficiency of the industrial practical operation process of the industrial equipment are synchronously obtained according to the real-time data video, so that a standard and efficient process flow supervision process is generated for real-time supervision of the industrial process, and the supervision effect is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial supervision, and in particular to a computer vision-based industrial process remote supervision method and system. Background Art

[0002] Data collection, transmission, and calculation in the Industrial Internet of Things mainly involve transmitting the data information collected by industrial gateway devices to the cloud platform, presenting the data in a visual way through edge computing, and monitoring the status of online devices through a unified Industrial Internet of Things platform, so as to facilitate timely understanding of equipment abnormality information and improve the supervision effect of managers on the process flow of industrial equipment.

[0003] At present, the supervision of the process flow of industrial equipment can only be done by manually judging whether the process flow is standard based on the collected process flow video. There is no standard and efficient process flow supervision process, and the supervision effect is poor. Summary of the invention

[0004] The purpose of the present invention is to provide a computer vision-based industrial process remote supervision method to solve the technical problems in the prior art that whether the process is standard can only be determined manually based on the collected process video, there is no standard and efficient process supervision process, and the supervision effect is poor.

[0005] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0006] A computer vision-based industrial process remote supervision method comprises the following steps:

[0007] Collect video data that represents the actual process flow of industrial equipment;

[0008] Using a first neural network, performing standardization recognition on the video data to obtain an industrial process standardization recognition model;

[0009] Using a second neural network, performing efficiency recognition on the video data to obtain an industrial process efficiency recognition model;

[0010] Synchronously combining the industrial process standardization identification model and the industrial process efficiency identification model to obtain an industrial process supervision model;

[0011] By utilizing the industrial process supervision model, the real-time standardization and real-time efficiency of the industrial equipment industrial operation process are synchronously obtained according to the real-time data video, so as to realize the real-time supervision of the industrial process.

[0012] As a preferred solution of the present invention, the construction of the industrial process standardization identification model includes:

[0013] Compare the similarity between the actual process of industrial equipment and the standard process, and use the similarity as the standard of video data;

[0014] Using the video data as an input item of the first neural network and using the standardization of the video data as an output item of the first neural network;

[0015] Using the first neural network to perform mapping training on the input items of the first neural network and the output items of the first neural network, so as to obtain the industrial process standardization identification model;

[0016] The industrial process standardization identification model is:

[0017] X = CNN1(S);

[0018] In the formula, X is the standard value, S is the video data, and CNN1 is the first neural network;

[0019] Among them, the loss function of the mapping training of the industrial process standardization identification model is the Euclidean distance between the standardization prediction value output by the industrial process standardization identification model and the standardization true value.

[0020] As a preferred solution of the present invention, the construction of the industrial process efficiency identification model includes:

[0021] Calculate the difference between the actual process speed of industrial equipment and the process standard speed, and use the difference as the efficiency of video data;

[0022] Using the video data as an input item of the second neural network, and using the efficiency of the video data as an output item of the second neural network;

[0023] Using the second neural network to perform mapping training on the input items of the second neural network and the output items of the second neural network to obtain the industrial process efficiency identification model;

[0024] The industrial process efficiency identification model is:

[0025] Y = CNN2(S);

[0026] Where Y is the efficiency value, S is the video data, and CNN2 is the second neural network;

[0027] Among them, the loss function of the mapping training of the industrial process efficiency identification model is the Euclidean distance between the standard prediction value and the standard true value output by the industrial process efficiency identification model.

[0028] As a preferred solution of the present invention, the construction of the industrial process supervision model includes:

[0029] The loss functions of the mapping training of the industrial process standardization identification model and the industrial process efficiency identification model are combined to obtain a combined target, which is:

[0030] Goal=[E1*(1-goal1) -1 log(goal1)]+[E2*(1-goal2) -1 log(goal2)];

[0031] Where Goal is the combined goal, E1 is the control parameter of the industrial process standardization identification model, E2 is the control parameter of the industrial process efficiency identification model, goal1 is the loss function of the mapping training of the industrial process standardization identification model, and goal2 is the loss function of the mapping training of the industrial process efficiency identification model;

[0032] Taking the combined target as the loss function of the industrial process supervision model, synchronously mapping training is performed on the industrial process standardization identification model and the industrial process efficiency identification model to obtain the industrial process supervision model;

[0033] The industrial process supervision model is:

[0034] [X,Y]=[CNN1(S),CNN2(S)];

[0035] Loss = Goal;

[0036] Where X is the standard value, Y is the efficiency value, S is the video data, CNN1 is the first neural network, CNN2 is the second neural network, and Loss is the loss function of the industrial process supervision model.

[0037] As a preferred solution of the present invention, when the industrial process standardization identification model and the industrial process efficiency identification model are combined, the output items of the industrial process standardization identification model and the industrial process efficiency identification model are normalized.

[0038] As a preferred solution of the present invention, the present invention provides a remote supervision system using the industrial process remote supervision method based on computer vision, comprising:

[0039] Industrial cameras are used to collect video data that represents the actual process flow of industrial equipment;

[0040] The central controller is used to utilize the industrial process supervision model to synchronously obtain the real-time standardization and real-time efficiency of the industrial equipment industrial operation process according to the real-time data video, so as to realize the real-time supervision of the industrial process;

[0041] A cloud server, configured to use a first neural network to perform standardization recognition on the video data to obtain an industrial process standardization recognition model;

[0042] Used to use a second neural network to perform efficiency recognition on the video data to obtain an industrial process efficiency recognition model;

[0043] for synchronously combining the industrial process standardization identification model and the industrial process efficiency identification model to obtain an industrial process supervision model; and

[0044] Store industrial process supervision models for central controller to call.

[0045] As a preferred solution of the present invention, the cloud server constructs the industrial process standardization identification model, including:

[0046] Compare the similarity between the actual process of industrial equipment and the standard process, and use the similarity as the standard of video data;

[0047] Using the video data as an input item of the first neural network and using the standardization of the video data as an output item of the first neural network;

[0048] Using the first neural network to perform mapping training on the input items of the first neural network and the output items of the first neural network, so as to obtain the industrial process standardization identification model;

[0049] The industrial process standardization identification model is:

[0050] X = CNN1(S);

[0051] In the formula, X is the standard value, S is the video data, and CNN1 is the first neural network;

[0052] Among them, the loss function of the mapping training of the industrial process standardization identification model is the Euclidean distance between the standardization prediction value output by the industrial process standardization identification model and the standardization true value.

[0053] As a preferred solution of the present invention, the cloud server constructs an industrial process efficiency identification model, including:

[0054] Calculate the difference between the actual process speed of industrial equipment and the process standard speed, and use the difference as the efficiency of video data;

[0055] Using the video data as an input item of the second neural network, and using the efficiency of the video data as an output item of the second neural network;

[0056] Using the second neural network to perform mapping training on the input items of the second neural network and the output items of the second neural network to obtain the industrial process efficiency identification model;

[0057] The industrial process efficiency identification model is:

[0058] Y = CNN2(S);

[0059] Where Y is the efficiency value, S is the video data, and CNN2 is the second neural network;

[0060] Among them, the loss function of the mapping training of the industrial process efficiency identification model is the Euclidean distance between the standard prediction value and the standard true value output by the industrial process efficiency identification model.

[0061] As a preferred solution of the present invention, the cloud server constructs an industrial process supervision model, including:

[0062] The loss functions of the mapping training of the industrial process standardization identification model and the industrial process efficiency identification model are combined to obtain a combined target, which is:

[0063] Goal=[E1*(1-goal1) -1 log(goal1)]+[E2*(1-goal2) -1 log(goal2)];

[0064] Where Goal is the combined goal, E1 is the control parameter of the industrial process standardization identification model, E2 is the control parameter of the industrial process efficiency identification model, goal1 is the loss function of the mapping training of the industrial process standardization identification model, and goal2 is the loss function of the mapping training of the industrial process efficiency identification model;

[0065] Taking the combined target as the loss function of the industrial process supervision model, synchronously mapping training is performed on the industrial process standardization identification model and the industrial process efficiency identification model to obtain the industrial process supervision model;

[0066] The industrial process supervision model is:

[0067] [X,Y]=[CNN1(S),CNN2(S)];

[0068] Loss = Goal;

[0069] Where X is the standard value, Y is the efficiency value, S is the video data, CNN1 is the first neural network, CNN2 is the second neural network, and Loss is the loss function of the industrial process supervision model.

[0070] As a preferred solution of the present invention, the industrial camera is communicatively connected to a cloud server, and the cloud server is communicatively connected to a central controller.

[0071] Compared with the prior art, the present invention has the following beneficial effects:

[0072] The present invention constructs an industrial process supervision model to achieve real-time standardization and real-time efficiency of industrial equipment industrial operation processes based on real-time data video, so as to enable real-time supervision of industrial processes, generate a standard and efficient process supervision process, and improve the supervision effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0074] Figure 1 A flowchart of a method for remote supervision of industrial processes based on computer vision provided by an embodiment of the present invention;

[0075] Figure 2 A block diagram of a remote monitoring system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0076] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] like Figure 1 As shown, the present invention provides a method for remote supervision of industrial processes based on computer vision, comprising the following steps:

[0078] Collect video data that represents the actual process flow of industrial equipment;

[0079] Using the first neural network, standardization recognition is performed on the video data to obtain an industrial process standardization recognition model;

[0080] The second neural network is used to identify the efficiency of the video data and obtain an industrial process efficiency identification model;

[0081] The industrial process standardization identification model and the industrial process efficiency identification model are synchronously combined to obtain an industrial process supervision model;

[0082] By utilizing the industrial process supervision model and based on real-time data video, the real-time standardization and real-time efficiency of the industrial equipment’s industrial operation process can be simultaneously obtained to achieve real-time supervision of the industrial process.

[0083] In order to realize remote supervision of the practical operation flow of industrial equipment processes, the present invention collects videos of the practical operation flow of the industrial equipment processes through industrial cameras, analyzes and identifies the collected video data, determines the standardization and efficiency of the practical operation flow, and completes the supervision of the practical operation flow of the industrial equipment processes. That is, it realizes remote supervision of the practical operation flow of industrial equipment processes by using computer vision, without the need for on-site supervision.

[0084] Specifically, in order to realize remote supervision of the process practice flow by using computer vision, the present invention constructs a neural network model that uses computer vision to identify the standardization degree of the process practice flow (i.e., an industrial process standardization recognition model), and constructs a neural network model that uses computer vision to identify the efficiency degree of the process practice flow (i.e., an industrial process efficiency recognition model), thereby respectively realizing computer vision analysis using neural networks to determine whether the captured process practice flow complies with standardized operations and whether the efficiency meets the requirements, that is, obtaining supervision indicators for supervising the process practice flow.

[0085] Furthermore, the industrial process standardization identification model and the industrial process efficiency identification model are combined to construct an industrial process supervision model, which enables the supervision indicators for the actual process flow to be obtained directly based on the captured video data, with high synchronization and improved supervision timeliness.

[0086] Among them, when combining the industrial process standardization identification model and the industrial process efficiency identification model, a combined training target is combined, that is, the loss function of the industrial process supervision model. In the loss function of the industrial process supervision model, the training priority of the industrial process standardization identification model and the industrial process efficiency identification model will be adaptively adjusted according to the model training situation. That is, when the loss function of the industrial process standardization identification model is larger, it indicates that the recognition accuracy of the industrial process standardization identification model is lower, and the proportion of the loss function of the industrial process standardization identification model in the loss function of the industrial process supervision model will be increased. Therefore, the training priority of the industrial process standardization identification model is increased, that is, the industrial process standardization identification model is trained first to enhance the model recognition effect of the industrial process standardization identification model. Similarly, the larger the loss function of the industrial process efficiency identification model is, the lower the recognition accuracy of the industrial process efficiency identification model is, and the proportion of the loss function of the industrial process efficiency identification model in the loss function of the industrial process supervision model will be increased. Therefore, the training priority of the industrial process efficiency identification model is increased, that is, the industrial process efficiency identification model is trained first to enhance the model recognition effect of the industrial process efficiency identification model.

[0087] Therefore, the construction of the combined objective can achieve adaptive configuration of resources between the two models, that is, the model with better training effect releases training resources to the model with less good training effect, and finally achieves synchronous optimal training of the two models, thereby achieving the industrial process supervision model in both standard recognition and efficiency recognition. High visual recognition effect.

[0088] In order to realize remote supervision of the process operation flow by computer vision, the present invention constructs a neural network model for identifying the standard degree of the process operation flow by computer vision (i.e., an industrial process standardization recognition model), and constructs a neural network model for identifying the efficiency degree of the process operation flow by computer vision (i.e., an industrial process efficiency recognition model), thereby respectively realizing computer vision analysis by neural network, determining whether the photographed process operation flow conforms to the standardized operation and whether the efficiency reaches the specified level, that is, obtaining the supervision indicators for supervising the process operation flow, which are as follows:

[0089] The construction of industrial process standardization identification model includes:

[0090] Compare the similarity between the actual process of industrial equipment and the standard process, and use the similarity as the standard of video data;

[0091] Using the video data as an input item of the first neural network and using the standardization of the video data as an output item of the first neural network;

[0092] Using the first neural network to perform mapping training on the input items of the first neural network and the output items of the first neural network, so as to obtain an industrial process standardization identification model;

[0093] The standard identification model of industrial processes is:

[0094] X = CNN1(S);

[0095] In the formula, X is the standard value, S is the video data, and CNN1 is the first neural network;

[0096] Among them, the loss function of the mapping training of the industrial process standardization identification model is the Euclidean distance between the standardization prediction value output by the industrial process standardization identification model and the standardization true value.

[0097] The construction of industrial process efficiency identification model includes:

[0098] Calculate the difference between the actual process speed of industrial equipment and the process standard speed, and use the difference as the efficiency of video data;

[0099] Using the video data as an input item of the second neural network, and using the efficiency of the video data as an output item of the second neural network;

[0100] Using the second neural network to perform mapping training on the input items of the second neural network and the output items of the second neural network, so as to obtain an industrial process efficiency recognition model;

[0101] The industrial process efficiency identification model is:

[0102] Y = CNN2(S);

[0103] Where Y is the efficiency value, S is the video data, and CNN2 is the second neural network;

[0104] Among them, the loss function of the mapping training of the industrial process efficiency identification model is the Euclidean distance between the standard predicted value and the standard true value output by the industrial process efficiency identification model.

[0105] The present invention combines the industrial process standardization identification model and the industrial process efficiency identification model to construct an industrial process supervision model, so as to obtain the supervision index for supervising the process operation directly according to the captured video data, with high synchronization and improved supervision timeliness, as follows:

[0106] Construction of industrial process supervision models, including:

[0107] The loss function of the combined industrial process standardization identification model and the industrial process efficiency identification model mapping training is used to obtain the combined target, which is:

[0108] Goal=[E1*(1-goal1) -1 log(goal1)]+[E2*(1-goal2) -1 log(goal2)];

[0109] Where Goal is the combined goal, E1 is the control parameter of the industrial process standardization identification model, E2 is the control parameter of the industrial process efficiency identification model, goal1 is the loss function of the mapping training of the industrial process standardization identification model, and goal2 is the loss function of the mapping training of the industrial process efficiency identification model;

[0110] The larger the goal1 in the combined goal, the greater the [E1*(1-goal1) -1 The larger the proportion of log(goal1)] in Goal, the larger the goal2 in the combined goal, then [E2*(1-goal2) -1 The larger the proportion of log(goal2)] in Goal, the larger the loss function of the industrial process standardization identification model, indicating that the recognition accuracy of the industrial process standardization identification model is, the proportion of the loss function of the industrial process standardization identification model will be increased in the loss function of the industrial process supervision model. Therefore, the training priority of the industrial process standardization identification model is increased, that is, the industrial process standardization identification model is trained first to enhance the model recognition effect of the industrial process standardization identification model. Similarly, the larger the loss function of the industrial process efficiency identification model, the lower the recognition accuracy of the industrial process efficiency identification model is, the proportion of the loss function of the industrial process efficiency identification model will be increased in the loss function of the industrial process supervision model. Therefore, the training priority of the industrial process efficiency identification model is increased, that is, the industrial process efficiency identification model is trained first to enhance the model recognition effect of the industrial process efficiency identification model.

[0111] Moreover, E1 and E2 are control parameters that can be set manually according to the actual application scenarios to autonomously control the training priority of the industrial process standardization identification model and the industrial process efficiency identification model. Therefore, in the scenarios where the emphasis is on supervising standardization or efficiency, E1 or E2 can be adjusted separately to achieve the control of training priority, that is, to adjust the sensitivity of standardization identification and efficiency identification in the industrial process supervision model. For example, if the emphasis is on supervising standardization, the value of E1 is increased, so that the sensitivity of standardization identification in the industrial process supervision model is higher, and it is easier to detect process operations with insufficient standardization, so that E1 and E2 as control parameters can adapt to more supervision scenarios.

[0112] Taking the combined target as the loss function of the industrial process supervision model, the industrial process standardization identification model and the industrial process efficiency identification model are synchronously mapped and trained to obtain the industrial process supervision model.

[0113] The industrial process supervision model is:

[0114] [X,Y]=[CNN1(S),CNN2(S)];

[0115] Loss = Goal;

[0116] Where X is the standard value, Y is the efficiency value, S is the video data, CNN1 is the first neural network, CNN2 is the second neural network, and Loss is the loss function of the industrial process supervision model.

[0117] When combining the industrial process standardization identification model and the industrial process efficiency identification model, a combined training objective is combined, that is, the loss function of the industrial process supervision model. In the loss function of the industrial process supervision model, the training priority of the industrial process standardization identification model and the industrial process efficiency identification model will be adaptively adjusted according to the model training situation. That is, the larger the loss function of the industrial process standardization identification model, the lower the recognition accuracy of the industrial process standardization identification model. The proportion of the loss function of the industrial process standardization identification model in the loss function of the industrial process supervision model will be increased, thereby increasing the training priority of the industrial process standardization identification model, that is, the industrial process standardization identification model is trained first to enhance the model recognition effect of the industrial process standardization identification model. Similarly, the larger the loss function of the industrial process efficiency identification model, the lower the recognition accuracy of the industrial process efficiency identification model. The proportion of the loss function of the industrial process efficiency identification model in the loss function of the industrial process supervision model will be increased, thereby increasing the training priority of the industrial process efficiency identification model, that is, the industrial process efficiency identification model is trained first to enhance the model recognition effect of the industrial process efficiency identification model.

[0118] Therefore, the construction of the combined objective can achieve adaptive configuration of resources between the two models, that is, the model with better training effect releases training resources to the model with less good training effect, and finally achieves synchronous optimal training of the two models, thereby achieving the industrial process supervision model in both standard recognition and efficiency recognition. High visual recognition effect.

[0119] When the industrial process standardization identification model and the industrial process efficiency identification model are combined, the output items of the industrial process standardization identification model and the industrial process efficiency identification model are normalized.

[0120] like Figure 2 As shown, the present invention provides a remote supervision system for an industrial process remote supervision method based on computer vision, comprising:

[0121] Industrial cameras are used to collect video data that represents the actual process flow of industrial equipment;

[0122] The central controller is used to use the industrial process supervision model to synchronously obtain the real-time standardization and real-time efficiency of the industrial equipment industrial operation process based on real-time data video, so as to realize real-time supervision of the industrial process;

[0123] The cloud server is used to use the first neural network to perform standard recognition on the video data to obtain an industrial process standard recognition model;

[0124] Used to use the second neural network to identify the efficiency of video data and obtain an industrial process efficiency identification model;

[0125] for synchronously combining the industrial process standardization identification model and the industrial process efficiency identification model to obtain an industrial process supervision model; and

[0126] Store industrial process supervision models for central controller to call.

[0127] In order to realize remote supervision of the practical operation flow of industrial equipment processes, the present invention collects videos of the practical operation flow of the industrial equipment processes through industrial cameras, analyzes and identifies the collected video data, determines the standardization and efficiency of the practical operation flow, and completes the supervision of the practical operation flow of the industrial equipment processes. That is, it realizes remote supervision of the practical operation flow of industrial equipment processes by using computer vision, without the need for on-site supervision.

[0128] Specifically, in order to realize remote supervision of the process practice flow by using computer vision, the present invention constructs a neural network model that uses computer vision to identify the standardization degree of the process practice flow (i.e., an industrial process standardization recognition model), and constructs a neural network model that uses computer vision to identify the efficiency degree of the process practice flow (i.e., an industrial process efficiency recognition model), thereby respectively realizing computer vision analysis using neural networks to determine whether the captured process practice flow complies with standardized operations and whether the efficiency meets the requirements, that is, obtaining supervision indicators for supervising the process practice flow.

[0129] Furthermore, the industrial process standardization identification model and the industrial process efficiency identification model are combined to construct an industrial process supervision model, which enables the supervision indicators for the actual process flow to be obtained directly based on the captured video data, with high synchronization and improved supervision timeliness.

[0130] Among them, when combining the industrial process standardization identification model and the industrial process efficiency identification model, a combined training target is combined, that is, the loss function of the industrial process supervision model. In the loss function of the industrial process supervision model, the training priority of the industrial process standardization identification model and the industrial process efficiency identification model will be adaptively adjusted according to the model training situation. That is, when the loss function of the industrial process standardization identification model is larger, it indicates that the recognition accuracy of the industrial process standardization identification model is lower, and the proportion of the loss function of the industrial process standardization identification model in the loss function of the industrial process supervision model will be increased. Therefore, the training priority of the industrial process standardization identification model is increased, that is, the industrial process standardization identification model is trained first to enhance the model recognition effect of the industrial process standardization identification model. Similarly, the larger the loss function of the industrial process efficiency identification model is, the lower the recognition accuracy of the industrial process efficiency identification model is, and the proportion of the loss function of the industrial process efficiency identification model in the loss function of the industrial process supervision model will be increased. Therefore, the training priority of the industrial process efficiency identification model is increased, that is, the industrial process efficiency identification model is trained first to enhance the model recognition effect of the industrial process efficiency identification model.

[0131] Therefore, the construction of the combined objective can achieve adaptive configuration of resources between the two models, that is, the model with better training effect releases training resources to the model with less good training effect, and finally achieves synchronous optimal training of the two models, thereby achieving the industrial process supervision model in both standard recognition and efficiency recognition. High visual recognition effect.

[0132] In order to realize remote supervision of the process operation flow by computer vision, the present invention constructs a neural network model for identifying the standard degree of the process operation flow by computer vision (i.e., an industrial process standardization recognition model), and constructs a neural network model for identifying the efficiency degree of the process operation flow by computer vision (i.e., an industrial process efficiency recognition model), thereby respectively realizing computer vision analysis by neural network, determining whether the photographed process operation flow conforms to the standardized operation and whether the efficiency reaches the specified level, that is, obtaining the supervision indicators for supervising the process operation flow, which are as follows:

[0133] The cloud server builds a standard identification model for industrial processes, including:

[0134] Compare the similarity between the actual process of industrial equipment and the standard process, and use the similarity as the standard of video data;

[0135] Using the video data as an input item of the first neural network and using the standardization of the video data as an output item of the first neural network;

[0136] Using the first neural network to perform mapping training on the input items of the first neural network and the output items of the first neural network, so as to obtain an industrial process standardization identification model;

[0137] The standard identification model of industrial processes is:

[0138] X = CNN1(S);

[0139] In the formula, X is the standard value, S is the video data, and CNN1 is the first neural network;

[0140] Among them, the loss function of the mapping training of the industrial process standardization identification model is the Euclidean distance between the standardization prediction value output by the industrial process standardization identification model and the standardization true value.

[0141] The cloud server builds an industrial process efficiency identification model, including:

[0142] Calculate the difference between the actual process speed of industrial equipment and the process standard speed, and use the difference as the efficiency of video data;

[0143] Using the video data as an input item of the second neural network and using the efficiency of the video data as an output item of the second neural network;

[0144] Using the second neural network to perform mapping training on the input items of the second neural network and the output items of the second neural network, so as to obtain an industrial process efficiency recognition model;

[0145] The industrial process efficiency identification model is:

[0146] Y = CNN2(S);

[0147] Where Y is the efficiency value, S is the video data, and CNN2 is the second neural network;

[0148] Among them, the loss function of the mapping training of the industrial process efficiency identification model is the Euclidean distance between the standard predicted value and the standard true value output by the industrial process efficiency identification model.

[0149] The cloud server builds an industrial process supervision model, including:

[0150] The loss function of the combined industrial process standardization identification model and the industrial process efficiency identification model mapping training is used to obtain the combined target, which is:

[0151] Goal=[E1*(1-goal1) -1 log(goal1)]+[E2*(1-goal2) -1 log(goal2)];

[0152] Where Goal is the combined goal, E1 is the control parameter of the industrial process standardization identification model, E2 is the control parameter of the industrial process efficiency identification model, goal1 is the loss function of the mapping training of the industrial process standardization identification model, and goal2 is the loss function of the mapping training of the industrial process efficiency identification model;

[0153] Taking the combined target as the loss function of the industrial process supervision model, the industrial process standardization identification model and the industrial process efficiency identification model are synchronously mapped and trained to obtain the industrial process supervision model.

[0154] The industrial process supervision model is:

[0155] [X,Y]=[CNN1(S),CNN2(S)];

[0156] Loss = Goal;

[0157] Where X is the standard value, Y is the efficiency value, S is the video data, CNN1 is the first neural network, CNN2 is the second neural network, and Loss is the loss function of the industrial process supervision model.

[0158] The industrial camera is connected to the cloud server for communication, and the cloud server is connected to the central controller for communication.

[0159] The industrial process standardization identification model and the industrial process efficiency identification model are combined to construct an industrial process supervision model, which can directly obtain supervision indicators for the actual process of the process based on the captured video data, with high synchronization and improved supervision timeliness.

[0160] Among them, when combining the industrial process standardization identification model and the industrial process efficiency identification model, a combined training target is combined, that is, the loss function of the industrial process supervision model. In the loss function of the industrial process supervision model, the training priority of the industrial process standardization identification model and the industrial process efficiency identification model will be adaptively adjusted according to the model training situation. That is, when the loss function of the industrial process standardization identification model is larger, it indicates that the recognition accuracy of the industrial process standardization identification model is lower, and the proportion of the loss function of the industrial process standardization identification model in the loss function of the industrial process supervision model will be increased. Therefore, the training priority of the industrial process standardization identification model is increased, that is, the industrial process standardization identification model is trained first to enhance the model recognition effect of the industrial process standardization identification model. Similarly, the larger the loss function of the industrial process efficiency identification model is, the lower the recognition accuracy of the industrial process efficiency identification model is, and the proportion of the loss function of the industrial process efficiency identification model in the loss function of the industrial process supervision model will be increased. Therefore, the training priority of the industrial process efficiency identification model is increased, that is, the industrial process efficiency identification model is trained first to enhance the model recognition effect of the industrial process efficiency identification model.

[0161] Therefore, the construction of the combined objective can achieve adaptive configuration of resources between the two models, that is, the model with better training effect releases training resources to the model with less good training effect, and finally achieves synchronous optimal training of the two models, thereby achieving the industrial process supervision model in both standard recognition and efficiency recognition. High visual recognition effect.

[0162] The present invention constructs an industrial process supervision model to achieve real-time standardization and real-time efficiency of industrial equipment industrial operation processes based on real-time data video, so as to enable real-time supervision of industrial processes, generate a standard and efficient process supervision process, and improve the supervision effect.

[0163] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A method for remote supervision of industrial processes based on computer vision, characterized in that: The following steps are involved: Collect video data that represents the actual process flow of industrial equipment; Using a first neural network, performing standardization recognition on the video data to obtain an industrial process standardization recognition model; Using a second neural network, performing efficiency recognition on the video data to obtain an industrial process efficiency recognition model; Synchronously combining the industrial process standardization identification model and the industrial process efficiency identification model to obtain an industrial process supervision model; By utilizing the industrial process supervision model, the real-time standardization and real-time efficiency of the industrial equipment industrial operation process are synchronously obtained according to the real-time data video, so as to realize the real-time supervision of the industrial process.

2. The method for remote supervision of industrial processes based on computer vision according to claim 1, characterized in that: The construction of the industrial process standardization identification model includes: Compare the similarity between the actual process of industrial equipment and the standard process, and use the similarity as the standard of video data; Using the video data as an input item of the first neural network and using the standardization of the video data as an output item of the first neural network; Using the first neural network to perform mapping training on the input items of the first neural network and the output items of the first neural network to obtain the industrial process standardization identification model; The industrial process standardization identification model is: X = CNN1(S); In the formula, X is the standard value, S is the video data, and CNN1 is the first neural network; Among them, the loss function of the mapping training of the industrial process standardization identification model is the Euclidean distance between the standardization prediction value output by the industrial process standardization identification model and the standardization true value.

3. The method for remote supervision of industrial processes based on computer vision according to claim 2, characterized in that: The construction of the industrial process efficiency identification model includes: Calculate the difference between the actual process speed of industrial equipment and the process standard speed, and use the difference as the efficiency of video data; Using the video data as an input item of the second neural network, and using the efficiency of the video data as an output item of the second neural network; Using the second neural network to perform mapping training on the input items of the second neural network and the output items of the second neural network to obtain the industrial process efficiency identification model; The industrial process efficiency identification model is: Y = CNN2(S); In the formula, Y is the efficiency value, S is the video data, and CNN2 is the second neural network; Among them, the loss function of the mapping training of the industrial process efficiency identification model is the Euclidean distance between the standard prediction value and the standard true value output by the industrial process efficiency identification model.

4. The method for remote supervision of industrial processes based on computer vision according to claim 3, characterized in that: The construction of the industrial process supervision model includes: The loss functions of the mapping training of the industrial process standardization identification model and the industrial process efficiency identification model are combined to obtain a combined target, which is: Goal=[E1*(1-goal1) -1 log(goal1)]+[E2*(1-goal2) -1 log(goal2)]; Where Goal is the combined goal, E1 is the control parameter of the industrial process standardization identification model, E2 is the control parameter of the industrial process efficiency identification model, goal1 is the loss function of the mapping training of the industrial process standardization identification model, and goal2 is the loss function of the mapping training of the industrial process efficiency identification model; Taking the combined target as the loss function of the industrial process supervision model, synchronously mapping training is performed on the industrial process standardization identification model and the industrial process efficiency identification model to obtain the industrial process supervision model; The industrial process supervision model is: [X,Y]=[CNN1(S),CNN2(S)]; Loss = Goal; Where X is the standard value, Y is the efficiency value, S is the video data, CNN1 is the first neural network, CNN2 is the second neural network, and Loss is the loss function of the industrial process supervision model.

5. The method for remote supervision of industrial processes based on computer vision according to claim 4, characterized in that: When the industrial process standardization identification model and the industrial process efficiency identification model are combined, the output items of the industrial process standardization identification model and the industrial process efficiency identification model are normalized.

6. A remote supervision system using the industrial process remote supervision method based on computer vision according to any one of claims 1 to 5, characterized in that: include: Industrial cameras are used to collect video data that represents the actual process flow of industrial equipment; The central controller is used to utilize the industrial process supervision model to synchronously obtain the real-time standardization and real-time efficiency of the industrial equipment industrial operation process according to the real-time data video, so as to realize the real-time supervision of the industrial process; A cloud server, configured to use a first neural network to perform standardization recognition on the video data to obtain an industrial process standardization recognition model; Used to use a second neural network to perform efficiency recognition on the video data to obtain an industrial process efficiency recognition model; for synchronously combining the industrial process standardization identification model and the industrial process efficiency identification model to obtain an industrial process supervision model; and Store industrial process supervision models for central controller to call.

7. A remote monitoring system according to claim 6, characterized in that: The cloud server constructs the industrial process standardization identification model, including: Compare the similarity between the actual process of industrial equipment and the standard process, and use the similarity as the standard of video data; Using the video data as an input item of the first neural network and using the standardization of the video data as an output item of the first neural network; Using the first neural network to perform mapping training on the input items of the first neural network and the output items of the first neural network to obtain the industrial process standardization identification model; The industrial process standardization identification model is: X = CNN1(S); In the formula, X is the standard value, S is the video data, and CNN1 is the first neural network; Among them, the loss function of the mapping training of the industrial process standardization identification model is the Euclidean distance between the standardization prediction value output by the industrial process standardization identification model and the standardization true value.

8. A remote monitoring system according to claim 7, characterized in that: The cloud server constructs an industrial process efficiency identification model, including: Calculate the difference between the actual process speed of industrial equipment and the process standard speed, and use the difference as the efficiency of video data; Using the video data as an input item of the second neural network, and using the efficiency of the video data as an output item of the second neural network; Using the second neural network to perform mapping training on the input items of the second neural network and the output items of the second neural network to obtain the industrial process efficiency identification model; The industrial process efficiency identification model is: Y = CNN2(S); In the formula, Y is the efficiency value, S is the video data, and CNN2 is the second neural network; Among them, the loss function of the mapping training of the industrial process efficiency identification model is the Euclidean distance between the standard prediction value and the standard true value output by the industrial process efficiency identification model.

9. A remote monitoring system according to claim 8, characterized in that: The cloud server constructs an industrial process supervision model, including: The loss functions of the mapping training of the industrial process standardization identification model and the industrial process efficiency identification model are combined to obtain a combined target, which is: Goal=[E1*(1-goal1) -1 log(goal1)]+[E2*(1-goal2) -1 log(goal2)]; Where Goal is the combined goal, E1 is the control parameter of the industrial process standardization identification model, E2 is the control parameter of the industrial process efficiency identification model, goal1 is the loss function of the mapping training of the industrial process standardization identification model, and goal2 is the loss function of the mapping training of the industrial process efficiency identification model; Taking the combined target as the loss function of the industrial process supervision model, synchronously mapping training is performed on the industrial process standardization identification model and the industrial process efficiency identification model to obtain the industrial process supervision model; The industrial process supervision model is: [X,Y]=[CNN1(S),CNN2(S)]; Loss = Goal; Where X is the standard value, Y is the efficiency value, S is the video data, CNN1 is the first neural network, CNN2 is the second neural network, and Loss is the loss function of the industrial process supervision model.

10. A remote monitoring system according to claim 6, characterized in that: The industrial camera is communicatively connected to a cloud server, and the cloud server is communicatively connected to a central controller.

Images