CN115797862A - A Method for Monitoring the Environment of Coating Room by Image Recognition Self-learning - Google Patents

Abstract

The invention discloses a method for monitoring the environment of a coating room through image recognition self-learning, which is characterized in that it comprises the following steps: S1: installing an infrared explosion-proof camera at a key position of the coating room system; S2: collecting process and environmental parameters; S3 : The collected parameter data and the constructed visual 3D model are collaboratively correlated with each other, and functions such as display prompts and alarms can be performed in the model space, and the collected data is included in the "big database"; S4: Establish image self-learning algorithm system. Beneficial effects of the present invention: This method adds an infrared explosion-proof camera to the coating room, integrates the equipment operation data transmitted by the center and the real-time data of environmental monitoring into the visualized three-dimensional model, and builds a collaborative space, which can prevent fires/machine abnormalities and other failures , Build monitoring, temperature sensor, smoke sensor and other multi-data monitoring, through big data self-learning, provide decision support in judging early equipment failures, fire evaluation, and quick fault detection.

Description

A Method for Monitoring the Environment of Coating Room by Image Recognition Self-learning

technical field

The invention relates to the technical field of monitoring the environment of a coating room in the metallurgical industry, in particular to a method for monitoring the environment of a coating room through image recognition self-learning.

Background technique

Metallurgy is the process and process of extracting metals or metal compounds from minerals, using various processing methods to make metals into metal materials with certain properties, color coating is to spray paint on the workpiece, so the environment of the coating room is bad, and the paint smells strong Infrared explosion-proof cameras are installed in the coating room of the cold rolling color coating unit to collect real-time data of equipment operation and environmental monitoring in the coating room. The collected data can be correlated with the visualized three-dimensional space for synergy and synchronization; data analysis can also be constructed It can judge and evaluate behaviors such as faults and abnormalities through self-learning, and provide decision support and solutions. Ensure the stable operation of maintenance equipment through fault and diagnosis, and ensure the safety of equipment personnel through early warning and judgment of dangerous scenes.

For example, the authorized announcement number is CN112446232U, and the authorized announcement date is 2021.03.05, a continuous self-learning image recognition method and system, including a continuous self-learning image recognition image server, and the continuous self-learning image recognition image server includes an image acquisition module, The first image marking module, the image recognition model training module, the new image marking module, the automatic/manual image verification module and the optimal model generation module, preferably, the image acquisition module is a special high-power surveillance camera for traffic. Preferably, the following steps are included: S1, at first through the image acquisition module, the image of the personnel is collected by the special high-magnification surveillance camera for traffic; S2, the image collected by S1 is sent to the initial marking module of the picture, and the initial marking module completes the process from Manually marked pictures are obtained from different picture marking personnel, and organized according to the structure required for image recognition model training, as the initial training set of the image recognition model.

For example, the authorized announcement number is CN05825235B, and the authorized announcement date is 2018.12.25, an image recognition method based on multi-feature map deep learning, including the training process of multi-feature map deep learning and the process of using the trained deep learning system for image recognition. , wherein, the training process of the multi-feature map deep learning includes the following steps: step a: obtain the grayscale image of the training sample set picture; obtain the feature of each pixel in the grayscale image for the grayscale image The feature map of composition, described feature map comprises: LBP feature map, gradient amplitude feature map and gradient direction feature map; Set the initial parameters of each network layer of deep convolutional network and classifier, the grayscale obtained in step a The feature map obtained in the figure and step b is input into the deep convolutional network to extract high-level features, that is, deep convolutional features, and the deep convolutional features are input into the classifier, and the classifier obtains the forward direction of the system The prediction output, wherein the parameters of the deep convolutional network and the classifier are the results of the previous study; the forward prediction output obtained in step c is compared with the label of the training sample set picture, and the error of the two Backpropagation, updating parameters of the deep convolutional network and parameters of the classifier according to the error.

The environment of the coating room of the color coating unit is harsh and the paint smells strong, but there is no video surveillance in this area, and the consequences caused by some failures are very serious. Therefore, it is urgent to design a method for monitoring the environment of the coating room through image recognition self-learning to solve the above problems. question.

Contents of the invention

The purpose of the present invention is to provide a method for monitoring the coating room environment through image recognition self-learning, so as to solve the above-mentioned deficiencies in the prior art.

In order to achieve the above object, the present invention provides the following technical solutions:

A method for monitoring the environment of a coating room through image recognition self-learning, comprising the following steps:

S1: Install infrared explosion-proof cameras at key positions of the coating room system;

S2: collecting process and environmental parameters;

S3: Collaboratively correlate the collected parameter data with the constructed visual 3D model, display prompts, alarm and other functions in the model space, and put the collected data into the "big database";

S4: Establish an image self-learning algorithm system;

S5: After the collected data reaches a certain scale, start learning and analyzing, and gradually optimize the judgment criteria;

S6: start image recognition and analysis of the implementation data;

S7: Comparison and inspection; (1) Compare the deviation between the implementation data and the data analysis conclusion in the "library" and optimize again; (2) Check and judge the synergistic stability between the analysis alarm and the three-dimensional space;

S8: Verify implementation.

As a preferred embodiment of the present invention, the collection parameters in step 2 include: (1) collection of process parameters in the coating chamber when key equipment such as roller coaters in the coating chamber are in operation, (2) collection of the environment in the coating chamber parameter.

As a preferred embodiment of the present invention, the process parameters include production data parameters such as top coating/back coating speed, roller pressure ratio, roller coating load, steel strip specification and speed, etc.

As a preferred embodiment of the present invention, the environmental parameters include ambient temperature, humidity, steel strip surface temperature, bearing temperature, smoke parameters and the like.

As a preferred embodiment of the present invention, the process parameters and environmental parameters are both detected and uploaded to acquire data by sensors.

As a preferred embodiment of the present invention, the specific process of establishing the learning algorithm in the step S4 is training image collection—preprocessing—feature extraction—model testing—outputting results.

As a preferred embodiment of the present invention, the specific process of self-learning and gradually optimizing the judgment criteria in the step S5 is: image input—layer 1—layer 2—layer 3—image label.

，其中，基于人类视觉的对比敏感度的感知曲线数字代码值集合与颜色分量水平集合之间的映射是基于所述函数模型的；并且从所述归一化颜色分量值Y确定颜色分量绝对值L，以便用于图形显示。As a preferred embodiment of the present invention, the image processing method is: (1) receiving the perceptual curve digital code value D of the image; (2) determining the normalized perceptual curve signal value V from the perceptual curve digital code value D ; (3) Determine the normalized color component value Y based on the following function model:

, wherein the mapping between the set of perceptual curve digital code values and the set of color component levels based on the contrast sensitivity of human vision is based on the functional model; and the absolute value of the color component is determined from the normalized color component value Y L for graphical display.

As a preferred embodiment of the present invention, the parameter n represents the number, m represents the time period, c1, c2 and c3 are predetermined values within a selected range.

As a preferred embodiment of the present invention, according to the different colors of the image, different scenes can be dealt with.

In the above technical solution, the present invention provides a method for monitoring the environment of the coating room through image recognition self-learning, which has beneficial effects: this method adds an infrared explosion-proof camera to the coating room, and the equipment operation data and environment monitoring transmitted by the center The real-time data is integrated in the visualized 3D model to build a collaborative space. For faults such as fire/machine abnormality, multiple data monitoring such as monitoring, temperature sensor, and smoke sensor are built. Through big data self-learning, it can judge early equipment faults, fire Provide decision support in terms of evaluation and quick fault detection, ensure the stable operation of maintenance equipment through fault and diagnosis, and ensure the safety of equipment personnel through early warning and judgment of dangerous scenes.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings that are required in the embodiments. Obviously, the accompanying drawings in the following description are only described in the present invention For some embodiments of the present invention, those skilled in the art can also obtain other drawings according to these drawings.

FIG. 1 is a schematic structural diagram of the specific process flow of the self-learning algorithm system provided by the embodiment of the method for monitoring the environment of the coating room through image recognition self-learning according to the present invention.

Fig. 2 is a schematic diagram of the specific process flow structure of the step-by-step optimization of the judgment standard provided by the embodiment of the method for monitoring the environment of the coating room through image recognition self-learning in the present invention.

Fig. 3 is a schematic diagram of a flow chart provided by an embodiment of a method for monitoring the environment of a coating room through image recognition self-learning in the present invention.

Detailed ways

In order to enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings.

As shown in Figures 1-3, a method for monitoring the environment of a coating room through image recognition self-learning provided by an embodiment of the present invention includes the following steps:

S1: Install infrared explosion-proof cameras at key positions of the coating room system;

S2: collecting process and environmental parameters;

S3: Collaboratively correlate the collected parameter data with the constructed visual 3D model, display prompts, alarm and other functions in the model space, and put the collected data into the "big database";

S4: Establish an image self-learning algorithm system;

S5: After the collected data reaches a certain scale, start learning and analyzing, and gradually optimize the judgment criteria;

S6: start image recognition and analysis of the implementation data;

S7: Comparison and inspection; (1) Compare the deviation between the implementation data and the data analysis conclusion in the "library" and optimize again; (2) Check and judge the synergistic stability between the analysis alarm and the three-dimensional space;

S8: Verify implementation.

Specifically, in this embodiment, a method for monitoring the coating room environment through image recognition self-learning includes the following steps: S1: installing an infrared explosion-proof camera at a key position of the coating room system; S2: collecting process and environmental parameters; S3 : The collected parameter data and the constructed visual 3D model are collaboratively correlated with each other, and functions such as display prompts and alarms can be performed in the model space, and the collected data is included in the "big database"; S4: Establish image self-learning algorithm System; S5: After the collected data reaches a certain scale, start learning and analyzing, and gradually optimize the judgment standard; S6: Start image recognition and analysis of the implementation data; S7: Comparison and inspection; (1) Compare the implementation data with the data in the "library" Analyze the deviation of the conclusion and optimize it again; (2) Check and judge the synergistic stability between the alarm and the three-dimensional space; S8: verify the implementation, this method adds an infrared explosion-proof camera in the coating room, and the equipment operation data transmitted by the center and the environment monitoring The real-time data is integrated in the visualized 3D model to build a collaborative space. For faults such as fire/machine abnormality, multiple data monitoring such as monitoring, temperature sensor, and smoke sensor are built. Through big data self-learning, it can judge early equipment faults, fire Provide decision support in terms of evaluation, quick fault detection, etc.

The invention provides a method for monitoring the environment of the coating room through image recognition self-learning. This method adds an infrared explosion-proof camera to the coating room, and integrates the equipment operation data transmitted by the center and the real-time data of environmental monitoring into the visualized three-dimensional model. In the middle, build a collaborative space. For faults such as fire/machine abnormality, build monitoring, temperature sensor, smoke sensor and other multi-data monitoring. Through big data self-learning, provide decision-making in judging early equipment failures, fire evaluation, and quick fault detection. Support, ensure the stable operation of maintenance equipment through fault and diagnosis, and ensure the safety of equipment personnel through early warning and judgment of dangerous scenes.

In another embodiment provided by the present invention, as shown in Figure 1, the collection parameters in step 2 include: (1) collection of process parameters during the operation of key equipment such as the roller coater in the coating room and the coating room, (2 ) to collect the environmental parameters in the coating chamber.

In another embodiment provided by the present invention, as shown in FIG. 1 , the process parameters include production data parameters such as top coating/back coating speed, roller pressure ratio, roller coating load, strip specification and speed, etc.

As a preferred embodiment of the present invention, the environmental parameters include ambient temperature, humidity, steel strip surface temperature, bearing temperature, smoke parameters and the like.

In another embodiment provided by the present invention, both process parameters and environmental parameters are detected and uploaded to acquire data through sensors.

In another embodiment provided by the present invention, as shown in FIG. 2 , the specific process of establishing a learning algorithm in step S4 is training image collection—preprocessing—feature extraction—model testing—outputting results.

In another embodiment provided by the present invention, as shown in FIG. 3 , the specific process of self-learning and gradually optimizing the judgment criteria in step S5 is: image input—layer 1—layer 2—layer 3—image label.

，其中，基于人类视觉的对比敏感度的感知曲线数字代码值集合与颜色分量水平集合之间的映射是基于函数模型的；并且从归一化颜色分量值Y确定颜色分量绝对值L，以便用于图形显示。In another embodiment provided by the present invention, the image processing method is: (1) receiving the perceptual curve digital code value D of the image; (2) determining the normalized perceptual curve signal value V from the perceptual curve digital code value D; ( 3) Determine the normalized color component value Y based on the following function model:

, where the mapping between the set of perceptual curve digital code values and the set of color component levels based on the contrast sensitivity of human vision is based on a functional model; and the absolute value of the color component L is determined from the normalized color component value Y in order to use in the graphic display.

In another embodiment provided by the present invention, the parameter n represents the number, m represents the time period, and c1, c2 and c3 are predetermined values within a selected range.

In another embodiment provided by the present invention, different scenes can be handled according to different colors of images.

Embodiment one

A method for monitoring the environment of a coating room through image recognition self-learning, comprising the following steps: S1: installing an infrared explosion-proof camera at a key position of the coating room system; S2: collecting process and environmental parameters; S3: combining the collected parameter data with the The constructed visualized 3D models are coordinated and correlated with each other, and can perform functions such as display prompts and alarms in the model space, and the collected data are included in the "big database"; S4: Establish an image self-learning algorithm system; S5: The collected data reaches After a certain scale, start to learn and analyze, and gradually optimize the judgment standard; S6: start image recognition and analysis of the implementation data; S7: comparison and inspection; (1) compare the deviation between the implementation data and the data analysis conclusion in the "library" and optimize again; (2) Check, judge and analyze the synergistic stability between the alarm and the three-dimensional space; S8: Verify the implementation

In another embodiment provided by the present invention, as shown in Figure 1, the collection parameters in step 2 include: (1) collection of process parameters (process parameters Including production data parameters such as top/back coating speed, roller pressure ratio, roller coating load, strip steel specification and speed, etc.), (2) Collect environmental parameters in the coating room (environmental parameters include ambient temperature, humidity, strip surface temperature, bearing temperature, smoke parameters, etc.).

In another embodiment provided by the present invention, both process parameters and environmental parameters are detected and uploaded to acquire data through sensors.

In another embodiment provided by the present invention, as shown in FIG. 2 , the specific process of establishing a learning algorithm in step S5 is training image collection—preprocessing—feature extraction—model testing—outputting results.

In another embodiment provided by the present invention, as shown in FIG. 3 , the specific process of self-learning and gradually optimizing the judgment criteria in step S5 is: image input—layer 1—layer 2—layer 3—image label.

，其中，基于人类视觉的对比敏感度的感知曲线数字代码值集合与颜色分量水平集合之间的映射是基于函数模型的；并且从归一化颜色分量值Y确定颜色分量绝对值L，以便用于图形显示，参数n代表数量、m代表时间段、c1、c2和c3是选定范围内预定值，根据图像不同的颜色，可应对不同的场景。In another embodiment provided by the present invention, the image processing method is: (1) receiving the perceptual curve digital code value D of the image; (2) determining the normalized perceptual curve signal value V from the perceptual curve digital code value D; ( 3) Determine the normalized color component value Y based on the following function model:

, where the mapping between the set of perceptual curve digital code values and the set of color component levels based on the contrast sensitivity of human vision is based on a functional model; and the absolute value of the color component L is determined from the normalized color component value Y in order to use For graphic display, the parameter n represents the number, m represents the time period, c1, c2 and c3 are predetermined values within the selected range, and can deal with different scenes according to different colors of the image.

，其中，基于人类视觉的对比敏感度的感知曲线数字代码值集合与颜色分量水平集合之间的映射是基于所述函数模型的；并且从所述归一化颜色分量值Y确定颜色分量绝对值L，以便用于图形显示，参数n代表数量、m代表时间段、c1、c2和c3是选定范围内预定值，根据图像不同的颜色，可应对不同的场景。Working principle: The steps of the method of monitoring the environment of the coating room through image recognition self-learning are S1: installing an infrared explosion-proof camera at a key position of the coating room system; S2: collecting process and environmental parameters, (1) collecting the collected coating in the coating room Process parameters of key equipment such as roller coaters in the layer chamber during operation. The process parameters include: production data parameters including top coating/back coating speed, roller pressure ratio, roller coating load, strip steel specification and speed, etc.; (2) collecting coating Environmental parameters in the layer room, including environmental temperature, humidity, steel surface temperature, bearing temperature, smoke parameters, etc.; S3: The collected parameter data and the constructed visual 3D model are collaboratively correlated, which can be carried out in the model space Display prompts, alarms and other functions, and put the collected data into the "big database"; S4: Establish an image self-learning algorithm system, and the specific process of establishing a learning algorithm is training image collection—preprocessing—feature extraction—model testing—output Results; S5: After the collected data reaches a certain scale, start to learn and analyze, and gradually optimize the judgment criteria. The specific process of self-learning and gradual optimization of the judgment criteria is: image input—layer 1—layer 2—layer 3—image label; S6: Start the image recognition and analysis of the implementation data; S7: Comparison and inspection; (1) Compare the deviation between the implementation data and the data analysis conclusion in the "library" and optimize again; (2) Check and judge the collaborative stability of the analysis alarm and the three-dimensional space; S8: Verify the implementation, the image processing method is to receive the digital code value D of the perceptual curve of the image; (2) determine the normalized perceptual curve signal value V from the digital code value D of the perceptual curve; (3) based on the following function model Determine the normalized color component value Y:

, wherein the mapping between the set of perceptual curve digital code values and the set of color component levels based on the contrast sensitivity of human vision is based on the functional model; and the absolute value of the color component is determined from the normalized color component value Y L, for graphic display, the parameter n represents the number, m represents the time period, c1, c2 and c3 are predetermined values in the selected range, and can deal with different scenes according to different colors of the image.

Certain exemplary embodiments of the present invention have been described above only by way of illustration, and it goes without saying that those skilled in the art can use various methods without departing from the spirit and scope of the present invention. The described embodiments are modified. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

Claims (10)

1. A method for monitoring the coating room environment through image recognition self-learning, characterized in that, comprising the following steps: S1: Install infrared explosion-proof cameras at key positions of the coating room system; S2: collecting process and environmental parameters; S3: Collaboratively correlate the collected parameter data with the constructed visual 3D model, display prompts, alarm and other functions in the model space, and put the collected data into the "big database"; S4: Establish an image self-learning algorithm system; S5: After the collected data reaches a certain scale, start learning and analyzing, and gradually optimize the judgment criteria; S6: start image recognition and analysis of the implementation data; S7: Comparison and inspection; (1) Compare the deviation between the implementation data and the data analysis conclusion in the "library" and optimize again; (2) Check and judge the synergistic stability between the analysis alarm and the three-dimensional space; S8: Verify implementation. 2. A method for monitoring the coating room environment through image recognition self-study according to claim 1, characterized in that, the collection parameters in the step 2 include: (1) Collecting the coating room and collecting the rollers in the coating room (2) Collect the environmental parameters in the coating room. 3. a kind of method by image recognition self-study monitoring coating room environment according to claim 2, is characterized in that, described process parameter comprises surface coating/back coating speed, roller pressure ratio, roller coating load, strip steel Production data parameters such as specifications and speeds, etc. 4. A method for monitoring the coating room environment through image recognition self-learning according to claim 2, wherein the environmental parameters include ambient temperature, humidity, steel strip surface temperature, bearing temperature, smoke parameters and the like. 5. A method for monitoring the coating room environment through image recognition self-learning according to claim 2, characterized in that, the process parameters and environmental parameters are all uploaded and acquired through sensor detection. 6. a kind of method by image recognition self-study monitoring coating room environment according to claim 1, is characterized in that, the specific process of setting up learning algorithm in the described step S4 is training image acquisition-preprocessing-feature extraction — Model testing — Output results. 7. a kind of method by image recognition self-study monitoring coating chamber environment according to claim 1, is characterized in that, the concrete process of self-study in the described step S5, progressively optimizing judgment standard is: image input-layer 1—layer 2—layer 3—image label. 8. A method for monitoring the coating room environment through image recognition self-study according to claim 1, characterized in that the image processing method is: (1) receive the digital code value D of the perception curve of the image; ( 2) Determine the normalized perceptual curve signal value V from the perceptual curve digital code value D; (3) Determine the normalized color component value Y based on the following function model:

, wherein the mapping between the set of perceptual curve digital code values and the set of color component levels based on the contrast sensitivity of human vision is based on the functional model; and the absolute value of the color component is determined from the normalized color component value Y L for graphical display. 9. A method for monitoring the coating room environment through image recognition self-learning according to claim 8, characterized in that, said parameter n represents the number, m represents the time period, c1, c2 and c3 are within the selected range predetermined value. 10. A method for monitoring the environment of a coating room through image recognition self-learning according to claim 8, characterized in that, according to the different colors of the images, different scenes can be dealt with.

Images