Updating method, device and equipment of coal mine underground AI visual model and medium

Abstract

The invention relates to an underground coal mine AI visual model updating method and device, equipment and a medium, and the method comprises the steps: starting an initial AI visual model, and carrying out the video monitoring of an underground coal mine operation scene; analyzing the monitoring video by using the initial AI visual model, and extracting at least one video frame containing an unsafe behavior as a target picture when identifying that the unsafe behavior exists in the monitoring video; each target picture is marked; combining the plurality of marked pictures with an original training picture set, and constructing a new training picture set; and using the new training picture set to re-train the initial AI vision model, and replacing the initial AI vision model with the AI vision model obtained after re-training. According to the method, the model is trained and updated in time according to new scenes and new conditions occurring in operation of the underground coal mine production environment, the adaptability and accuracy of the AI visual model are improved, and the maintenance cost is reduced.

Description

Technical Field

[0001] The present invention relates to the field of coal mine safety technology, and in particular to a method, device, equipment and medium for updating an AI visual model in an underground coal mine. Background Art

[0002] Coal mine safety has always been the focus of widespread social attention. According to statistics, the vast majority (over 90%) of safety accidents in coal mine production are caused by improper operation and weak safety awareness. In order to fundamentally prevent the occurrence of these accidents, coal mine safety management urgently needs to build an efficient and real-time monitoring system. This system should be able to accurately capture unsafe behaviors and abnormal conditions of objects in coal mines. Once potential dangers or signs of accidents are found, corresponding measures should be taken immediately to eliminate safety accidents in the bud.

[0003] Against this background, AI (Artificial Intelligence) visual intelligent recognition technology for coal mines came into being. Its unique all-weather monitoring, high-accuracy recognition, and efficient processing capabilities have brought revolutionary changes to coal mine safety production. This technology has not only significantly improved the intelligence level of coal mines, but also played an immeasurable role in ensuring production safety, reducing labor costs, and improving management efficiency. However, the extreme complexity and particularity of the underground environment of coal mines poses severe challenges to AI visual recognition technology. Summary of the invention

[0004] The present invention aims to solve one of the technical problems in the related art at least to a certain extent.

[0005] To this end, the first purpose of the present invention is to propose a method for updating the AI ​​vision model in underground coal mines. This method can timely train and update the model according to new scenes and new situations that arise in the operation of the underground production environment of coal mines, thereby improving the adaptability and accuracy of the AI ​​vision model and reducing maintenance costs.

[0006] The second purpose of the present invention is to provide a device for updating the AI ​​visual model in underground coal mines.

[0007] A third objective of the present invention is to provide an electronic device.

[0008] A fourth object of the present invention is to provide a computer-readable storage medium.

[0009] A fifth object of the present invention is to provide a computer program product.

[0010] To achieve the above-mentioned purpose, the first aspect of the embodiment of the present invention proposes a method for updating an AI visual model in an underground coal mine, comprising: starting an initial AI visual model to perform video monitoring of an underground coal mine operation scene; using the initial AI visual model to analyze the monitoring video, and when unsafe behavior is identified in the monitoring video, extracting at least one video frame containing the unsafe behavior as a target image; annotating each of the target images, wherein the annotation content includes the type of unsafe behavior and the location of occurrence in the target image; merging multiple annotated images with the original training image set to construct a new training image set; using the new training image set to retrain the initial AI visual model, and replacing the initial AI visual model with the retrained AI visual model.

[0011] According to the method for updating the AI ​​visual model in the coal mine underground of the embodiment of the present invention, the initial AI visual model is started, the working scene in the coal mine is monitored by video, and the monitoring video is analyzed by using the initial AI visual model. When unsafe behavior is identified in the monitoring video, at least one video frame containing the unsafe behavior is extracted as the target image, and each target image is annotated. The annotation content includes the type of unsafe behavior and the location of occurrence in the target image. After that, multiple annotated images are merged with the original training image set to construct a new training image set, and the new training image set is used to retrain the initial AI visual model, and the retrained AI visual model replaces the initial AI visual model. Therefore, the method can timely train and update the model according to new scenes and new situations that appear in the operation of the underground production environment of the coal mine, improve the adaptability and accuracy of the AI ​​visual model, and reduce maintenance costs.

[0012] In addition, the method for updating the AI ​​visual model in coal mines proposed in the first aspect of the embodiment of the present invention may also have the following additional technical features:

[0013] According to one embodiment of the present invention, the step of acquiring the initial AI visual model includes:

[0014] Acquire multiple sample images related to coal mine operation scenes, and perform preprocessing on each of the sample images including denoising, edge enhancement, and brightness and contrast adjustment;

[0015] Annotate the preprocessed sample images to obtain annotated image sets;

[0016] The initial visual model is trained using the annotated picture set under a deep learning framework, and parameters in the initial visual model are adjusted.

[0017] According to an embodiment of the present invention, before merging the plurality of annotated images with the original training image set to construct a new training image set, the method further includes:

[0018] Verify each annotated image to confirm the correctness of the initial AI visual model's recognition result of the annotated image;

[0019] If the verification result shows that the recognition result of the initial AI visual model for the annotated picture is wrong, the annotated picture is classified into an abnormal picture library; wherein the abnormal picture library also includes annotated pictures that the initial AI visual model failed to recognize and in which unsafe behaviors actually occurred;

[0020] Correspondingly, the step of merging the multiple annotated images with the original training image set to construct a new training image set includes:

[0021] A plurality of annotated pictures are obtained from the abnormal picture library, and are merged with the original training picture set to construct the new training picture set.

[0022] According to one embodiment of the present invention, before replacing the initial AI vision model with the retrained AI vision model, the method further includes:

[0023] Analyzing the retrained AI vision model and the original AI vision model using the same test picture set, and calculating their respective accuracies;

[0024] If the accuracy of the retrained AI vision model is higher than that of the original AI vision model, replacing the initial AI vision model with the retrained AI vision model;

[0025] The retrained AI visual model and its corresponding training image set are updated to the cloud database for deployment and use in the underground coal mine environment.

[0026] To achieve the above-mentioned purpose, the second aspect of the embodiment of the present invention proposes an updating device for the AI ​​visual model in an underground coal mine, comprising: a starting module, used to start the initial AI visual model and perform video monitoring of the working scene in the underground coal mine; an analysis module, used to use the initial AI visual model to analyze the monitoring video, and when unsafe behavior is identified in the monitoring video, extract at least one video frame containing the unsafe behavior as the target image; a labeling module, used to label each of the target images, and the labeling content includes the type of unsafe behavior and the location of occurrence in the target image; a construction module, used to merge multiple labeled images with the original training picture set to construct a new training picture set; an update module, used to use the new training picture set to retrain the initial AI visual model, and replace the initial AI visual model with the retrained AI visual model.

[0027] According to the updating device of the AI ​​visual model in the coal mine underground of the embodiment of the present invention, the initial AI visual model is started by the starting module to conduct video monitoring of the working scene in the coal mine underground, and the monitoring video is analyzed by the initial AI visual model through the analysis module. When unsafe behavior is identified in the monitoring video, at least one video frame containing the unsafe behavior is extracted as the target image, and each target image is annotated by the annotation module, and the annotation content includes the type of unsafe behavior and the location of occurrence in the target image. The multiple annotated images are merged with the original training image set by the construction module to construct a new training image set, and the initial AI visual model is retrained by using the new training image set through the update module, and the retrained AI visual model replaces the initial AI visual model. Therefore, the device can timely train and update the model according to new scenes and new situations in the operation of the underground production environment of the coal mine, improve the adaptability and accuracy of the AI ​​visual model, and reduce the maintenance cost.

[0028] In addition, the updating device for the AI ​​visual model in coal mines provided in the second aspect of the present invention may also have the following additional technical features:

[0029] According to one embodiment of the present invention, the device further includes:

[0030] An acquisition module is used to acquire multiple sample images related to coal mine operation scenes, and preprocess each of the sample images including denoising, edge enhancement, and brightness and contrast adjustment, and annotate the preprocessed sample images to obtain an annotated image set, and use the annotated image set to train the initial visual model under a deep learning framework, and adjust the parameters in the initial visual model.

[0031] According to one embodiment of the present invention, the device further includes:

[0032] A determination module is used to verify each annotated picture to confirm the correctness of the recognition result of the initial AI visual model for the annotated image. If the verification result shows that the recognition result of the initial AI visual model for the annotated picture is wrong, the annotated picture is classified into an abnormal picture library; wherein the abnormal picture library also includes annotated pictures that the initial AI visual model fails to recognize and in which unsafe behaviors actually occur;

[0033] Correspondingly, the construction module is used to merge the multiple annotated pictures with the original training picture set to construct a new training picture set, including:

[0034] A plurality of annotated pictures are obtained from the abnormal picture library, and are merged with the original training picture set to construct the new training picture set.

[0035] According to one embodiment of the present invention, before the updating module is used to replace the initial AI vision model with the retrained AI vision model, it is also used to:

[0036] Analyzing the retrained AI vision model and the original AI vision model using the same test picture set, and calculating their respective accuracies;

[0037] If the accuracy of the retrained AI vision model is higher than that of the original AI vision model, replacing the initial AI vision model with the retrained AI vision model;

[0038] The retrained AI visual model and its corresponding training image set are updated to the cloud database for deployment and use in the underground coal mine environment.

[0039] To achieve the above object, the third aspect of the present invention further provides an electronic device, including:

[0040] at least one processor; and

[0041] a memory communicatively connected to the at least one processor; wherein,

[0042] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned method for updating the AI ​​visual model in underground coal mines.

[0043] The battery equipment of the embodiment of the present invention, by executing the above-mentioned method for updating the AI ​​vision model in underground coal mines, timely trains and updates the model according to new scenes and new situations that arise in the operation of the underground production environment of coal mines, thereby improving the adaptability and accuracy of the AI ​​vision model and reducing maintenance costs.

[0044] In order to achieve the above-mentioned purpose, the fourth aspect of the present invention also proposes a computer-readable storage medium, and the computer instructions are used to enable the computer to execute the above-mentioned method for updating the AI ​​visual model in underground coal mines.

[0045] The computer-readable storage medium of an embodiment of the present invention, by executing the above-mentioned method for updating the AI ​​vision model in an underground coal mine, timely trains and updates the model according to new scenes and new situations that arise in the operation of the underground production environment of the coal mine, thereby improving the adaptability and accuracy of the AI ​​vision model and reducing maintenance costs.

[0046] To achieve the above objectives, the fifth aspect of the present invention further proposes a computer program product, which, when the instruction processor in the computer program product is executed, executes the above-mentioned method for updating the AI ​​visual model of underground coal mines.

[0047] The computer program product of the embodiment of the present invention, by executing the above-mentioned method for updating the AI ​​vision model in underground coal mines, timely trains and updates the model according to new scenes and new situations that arise in the operation of the underground production environment of coal mines, thereby improving the adaptability and accuracy of the AI ​​vision model and reducing maintenance costs.

[0048] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0050] Figure 1 is a flow chart of a method for updating an AI visual model in an underground coal mine according to an embodiment of the present invention;

[0051] Figure 2 4 is a block diagram of a device for updating an AI visual model in an underground coal mine according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0053] The following describes the updating method, device, equipment and medium of the AI ​​visual model in an underground coal mine according to an embodiment of the present invention with reference to the accompanying drawings.

[0054] Figure 1 It is a flowchart of a method for updating an AI vision model in an underground coal mine according to an embodiment of the present invention.

[0055] like Figure 1 As shown, the method for updating the AI ​​visual model in an underground coal mine according to an embodiment of the present invention includes:

[0056] S1, start the initial AI vision model to conduct video monitoring of the working scenes underground in the coal mine.

[0057] In an embodiment of the present invention, the steps of acquiring the initial AI visual model include: acquiring multiple sample images related to coal mine operation scenes, and preprocessing each sample image including denoising, enhancing edges, and adjusting brightness and contrast; annotating the preprocessed sample images to obtain an annotated image set; using the annotated image set, training the initial visual model under a deep learning framework, and adjusting the parameters in the initial visual model.

[0058] First, a large number of sample pictures closely related to the coal mine operating environment are collected from multiple channels (such as public picture databases, on-site photography in coal mines, etc.) to ensure the diversity and comprehensiveness of the picture collection.

[0059] Due to the insufficient light and high noise levels in the underground environment, a series of refined preprocessing measures are implemented on the collected images, including but not limited to image denoising, edge sharpening, adaptive adjustment of brightness and contrast, and dehazing for the unique fogging phenomenon underground, thereby significantly improving image quality.

[0060] According to the scene requirements, determine the targets that need to be identified, such as whether the miners are wearing helmets correctly, whether they have illegally crossed the safety area, etc. Then, use advanced image annotation tools to draw a rectangular box on the picture to mark the location of the target object, ensuring that the annotation range is neither too large nor too small, and just covers the outer contour of the target object. For pictures containing multiple targets, annotate them one by one, and record the serial number, name and location information of each target in detail in the annotation file. After completing the annotation of all pictures, organize the image data and annotation data in a unified standard format to form a complete set of annotated pictures for subsequent model training.

[0061] According to the special needs and data characteristics of coal mines, select appropriate deep learning frameworks such as TensorFlow or PyTorch to support an efficient and stable model training process. Design a customized model training algorithm strategy for coal mine scenarios, and set reasonable training parameters (such as learning rate, batch size, and number of iterations) to maximize model performance. Package the labeled image set, training algorithm, and parameters into a training image and deploy it on the training platform for model training. During this period, according to the quality inspection results during the training process, continuously iterate and adjust the algorithm and parameters until the initial AI vision model achieves the ideal effect.

[0062] Afterwards, the initial AI vision model and related resources (including training images, optimal hyperparameters, and training image sets) can be uploaded to the algorithm mall to facilitate rapid deployment and use by coal mining companies. Coal mining companies can directly obtain models from the algorithm mall and integrate them into the video surveillance system to achieve real-time and accurate monitoring of key factors such as miner behavior and equipment status, effectively improving the safety and efficiency of coal mining operations.

[0063] S2 uses the initial AI visual model to analyze the surveillance video, and when unsafe behavior is identified in the surveillance video, extracts at least one video frame containing the unsafe behavior as a target image.

[0064] First, all resources such as models, programs, and image collections provided in the algorithm mall are deployed to the computing environment in the coal mine. Then, the AI ​​visual model intelligent analysis service automatically pulls the latest version of the model file from the algorithm mall and starts the service process. After the service is started, it immediately establishes a connection with the video acquisition equipment such as cameras in the coal mine, accesses the high-definition video data stream in real time, and prepares for subsequent intelligent analysis.

[0065] After the AI ​​visual model analysis service is started, it immediately enters a continuous monitoring state. It uses the deployed initial AI visual model to analyze the incoming video stream frame by frame. During the analysis process, the initial AI visual model uses image recognition and processing capabilities to automatically detect unsafe behaviors in the video, such as miners not wearing safety helmets and illegally crossing safety areas. Once an unsafe behavior is identified, the system will immediately trigger the screenshot mechanism to accurately extract at least one video frame containing the unsafe behavior from the video stream as the target image.

[0066] The extracted target images will be automatically saved by the system to the specified location, and key information such as timestamp and unsafe behavior type will be attached to facilitate subsequent safety review and accident tracing. At the same time, the system will also send alarm information to relevant personnel according to the preset alarm rules, reminding them to pay attention to and deal with potential safety hazards in a timely manner.

[0067] S3, annotate each target image, and the annotation content includes the type of unsafe behavior and the location where it occurs in the target image.

[0068] The marking steps are as follows:

[0069] First, the unsafe behaviors in each target image are clearly classified, such as "not wearing a safety helmet", "illegally crossing the safety area", "improper operation causing sparks", etc., and the type of behavior is clearly marked in the image or the record associated with it.

[0070] Use professional image annotation tools to directly mark the specific location of the unsafe behavior on the target image. This is usually achieved by drawing graphics such as rectangles, ellipses or lines to ensure that the annotation range accurately covers the main part of the unsafe behavior. At the same time, text descriptions can be added to the annotations to further clarify the specific details or characteristics of the unsafe behavior.

[0071] Integrate the annotated information of each target image (including the type of unsafe behavior, location, timestamp, etc.) into a unified database or management system for subsequent query, analysis and processing. Ensuring the accuracy and completeness of the annotated information is a key link in ensuring the effectiveness of the entire safety monitoring system.

[0072] After the marking is completed, a quality check should be carried out to ensure that all markings meet the specifications and requirements. Any doubtful or uncertain markings should be reviewed and corrected in a timely manner to ensure the accuracy and reliability of the marking results.

[0073] S4, merge multiple annotated images with the original training image set to construct a new training image set.

[0074] Before constructing a new training picture set, each annotated picture is verified to confirm the correctness of the initial AI visual model's recognition results for the annotated images. If the verification results show that the initial AI visual model's recognition results for the annotated pictures are incorrect, the annotated pictures are classified into the abnormal picture library; the abnormal picture library also includes annotated pictures that the initial AI visual model failed to recognize and in which unsafe behaviors actually occurred. Then, multiple annotated pictures are obtained from the abnormal picture library and merged with the original training picture set to construct a new training picture set.

[0075] S5, using a new training picture set, retrains the initial AI vision model, and replaces the initial AI vision model with the retrained AI vision model.

[0076] Before replacing the initial AI vision model with the retrained AI vision model, the retrained AI vision model and the original AI vision model are analyzed using the same test image set, and their respective accuracies are calculated. If the accuracy of the retrained AI vision model is higher than that of the original AI vision model, the retrained AI vision model replaces the initial AI vision model, and the retrained AI vision model and its corresponding training image set are updated to the cloud database for deployment and use in the underground coal mine environment.

[0077] The present invention first clarifies the training goal, pulls the latest training image from the algorithm mall, and verifies its integrity. Confirm that the image set data annotated in the previous step is ready, and perform data cleaning and preprocessing to ensure data quality. Split the data set according to a reasonable ratio (such as 70% training set, 15% validation set, 15% test set) to ensure that the data is evenly distributed. According to the model training hyperparameter guide published in the mall, perform detailed parameter configuration, including learning rate, batch size, number of iterations, etc.

[0078] Send the configured training task information to the server to start the model training. Monitor the training progress in real time by viewing the log or training interface, including changes in indicators such as loss function value and accuracy. Once an abnormality is found during training (such as gradient disappearance, overfitting, etc.), immediately troubleshoot the problem and adjust the training strategy according to the log prompts. After the training is completed, generate the model file and perform preliminary verification to ensure that the model is not damaged.

[0079] Load the model file generated by this training and the original model file in the mall into the test service respectively. Use a set of identical and representative test data sets to perform inference analysis on the two models, record and compare their inference results. Calculate the accuracy of the two models on the test data set and perform comparative analysis. If the newly trained model has a higher accuracy, proceed to the next step of the release process; otherwise, return to check possible reasons such as data annotation quality and training parameter settings, and retrain.

[0080] Update the model file with a new version number to ensure that the old and new versions can be clearly distinguished. At the same time, keep the old version model in the algorithm mall so that it can be rolled back when needed. Update the model description and performance indicators in the mall, and publish notifications of new model launches so that users can understand and use the latest models. The intelligent service system pulls the latest model files from the mall regularly or on demand, updates the model version in the local service, and ensures that the service always uses the optimal AI vision model.

[0081] Collect user feedback on the new model, including usage effects, existing problems, etc. Based on user feedback and actual application, continuously optimize the model structure and training parameters to improve the accuracy and generalization ability of the model. Regularly update model training and usage documents to ensure that users can easily get started and make full use of the model functions.

[0082] In summary, according to the method for updating the AI ​​visual model in coal mine underground according to the embodiment of the present invention, the initial AI visual model is started, the working scene in the coal mine is monitored by video, and the monitoring video is analyzed using the initial AI visual model. When unsafe behavior is identified in the monitoring video, at least one video frame containing the unsafe behavior is extracted as the target image, and each target image is annotated. The annotation content includes the type of unsafe behavior and the location of occurrence in the target image. After that, multiple annotated images are merged with the original training image set to construct a new training image set, and the new training image set is used to retrain the initial AI visual model, and the retrained AI visual model replaces the initial AI visual model. Therefore, the method can timely train and update the model according to new scenes and new situations that appear in the operation of the underground production environment of the coal mine, improve the adaptability and accuracy of the AI ​​visual model, and reduce maintenance costs.

[0083] Figure 2 It is a flow chart of an updating device for an AI visual model in an underground coal mine according to an embodiment of the present invention.

[0084] like Figure 2 As shown, the updating device 200 for the AI ​​visual model of an underground coal mine according to an embodiment of the present invention includes: a starting module 210, an analysis module 220, a labeling module 230, a construction module 240 and an updating module 250.

[0085] Among them, the startup module 210 is used to start the initial AI visual model and perform video monitoring of the working scene underground in the coal mine. The analysis module 220 is used to analyze the monitoring video using the initial AI visual model, and when unsafe behavior is identified in the monitoring video, extract at least one video frame containing unsafe behavior as the target image. The annotation module 230 is used to annotate each of the target images, and the annotation content includes the type of unsafe behavior and the location of occurrence in the target image. The construction module 240 is used to merge multiple annotated images with the original training picture set to construct a new training picture set. The update module 250 is used to use the new training picture set to retrain the initial AI visual model, and replace the initial AI visual model with the retrained AI visual model.

[0086] According to one embodiment of the present invention, the device further includes:

[0087] An acquisition module is used to acquire multiple sample images related to coal mine operation scenes, and preprocess each of the sample images including denoising, edge enhancement, and brightness and contrast adjustment, and annotate the preprocessed sample images to obtain an annotated image set, and use the annotated image set to train the initial visual model under a deep learning framework, and adjust the parameters in the initial visual model.

[0088] According to one embodiment of the present invention, the device further includes:

[0089] A determination module is used to verify each annotated picture to confirm the correctness of the recognition result of the initial AI visual model for the annotated image. If the verification result shows that the recognition result of the initial AI visual model for the annotated picture is wrong, the annotated picture is classified into an abnormal picture library; wherein the abnormal picture library also includes annotated pictures that the initial AI visual model fails to recognize and in which unsafe behaviors actually occur;

[0090] Correspondingly, the construction module 240 is used to merge the multiple annotated pictures with the original training picture set to construct a new training picture set, including:

[0091] A plurality of annotated pictures are obtained from the abnormal picture library, and are merged with the original training picture set to construct the new training picture set.

[0092] According to one embodiment of the present invention, before the updating module 250 is used to replace the initial AI vision model with the retrained AI vision model, it is also used to:

[0093] Analyzing the retrained AI vision model and the original AI vision model using the same test picture set, and calculating their respective accuracies;

[0094] If the accuracy of the retrained AI vision model is higher than that of the original AI vision model, replacing the initial AI vision model with the retrained AI vision model;

[0095] The retrained AI visual model and its corresponding training image set are updated to the cloud database for deployment and use in the underground coal mine environment.

[0096] It should be noted that for details not disclosed in the updating device for the AI ​​visual model in an underground coal mine in an embodiment of the present invention, please refer to the details disclosed in the updating method for the AI ​​visual model in an underground coal mine in an embodiment of the present invention, and the details will not be repeated here.

[0097] According to the updating device of the AI ​​visual model in the coal mine underground of the embodiment of the present invention, the initial AI visual model is started by the starting module to conduct video monitoring of the working scene in the coal mine underground, and the monitoring video is analyzed by the initial AI visual model through the analysis module. When unsafe behavior is identified in the monitoring video, at least one video frame containing the unsafe behavior is extracted as the target image, and each target image is annotated by the annotation module, and the annotation content includes the type of unsafe behavior and the location of occurrence in the target image. The multiple annotated images are merged with the original training image set by the construction module to construct a new training image set, and the initial AI visual model is retrained by using the new training image set through the update module, and the retrained AI visual model replaces the initial AI visual model. Therefore, the device can timely train and update the model according to new scenes and new situations in the operation of the underground production environment of the coal mine, improve the adaptability and accuracy of the AI ​​visual model, and reduce the maintenance cost.

[0098] Based on the above embodiments, the present invention further provides an electronic device.

[0099] An electronic device according to an embodiment of the present invention comprises: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned method for updating the AI ​​visual model of underground coal mines.

[0100] It should be noted that the processor in the electronic device of the embodiment of the present invention may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or a general-purpose processor or any conventional processor, etc.

[0101] The battery equipment of the embodiment of the present invention, by executing the above-mentioned method for updating the AI ​​vision model in underground coal mines, timely trains and updates the model according to new scenes and new situations that arise in the operation of the underground production environment of coal mines, thereby improving the adaptability and accuracy of the AI ​​vision model and reducing maintenance costs.

[0102] Based on the above embodiments, the present invention also proposes a computer-readable storage medium.

[0103] The computer instructions in the computer-readable storage medium of the embodiment of the present invention are used to enable the computer to execute the above-mentioned method for updating the AI ​​visual model in an underground coal mine.

[0104] The computer-readable storage medium of an embodiment of the present invention, by executing the above-mentioned method for updating the AI ​​vision model in an underground coal mine, timely trains and updates the model according to new scenes and new situations that arise in the operation of the underground production environment of the coal mine, thereby improving the adaptability and accuracy of the AI ​​vision model and reducing maintenance costs.

[0105] Based on the above embodiments, the present invention also proposes a computer program product.

[0106] In an embodiment of the present invention, when the instruction processor in the computer program product is executed, the above-mentioned method for updating the AI ​​visual model in an underground coal mine is executed.

[0107] The computer program product of the embodiment of the present invention, by executing the above-mentioned method for updating the AI ​​vision model in underground coal mines, timely trains and updates the model according to new scenes and new situations that arise in the operation of the underground production environment of coal mines, thereby improving the adaptability and accuracy of the AI ​​vision model and reducing maintenance costs.

[0108] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0109] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0110] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0111] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0112] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0113] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0114] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0115] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present invention. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for updating an AI visual model in an underground coal mine, characterized in that: include: Start the initial AI vision model to conduct video monitoring of the working scenes in the coal mine; Analyzing the surveillance video using the initial AI visual model, and when unsafe behavior is identified in the surveillance video, extracting at least one video frame containing the unsafe behavior as a target image; Annotating each of the target images, wherein the annotation content includes the type of unsafe behavior and the location where the unsafe behavior occurs in the target image; Merge multiple annotated images with the original training image set to construct a new training image set; The initial AI vision model is retrained using the new training picture set, and the retrained AI vision model replaces the initial AI vision model.

2. The method according to claim 1, characterized in that The steps of obtaining the initial AI visual model include: Acquire multiple sample images related to coal mine operation scenes, and perform preprocessing on each of the sample images including denoising, edge enhancement, and brightness and contrast adjustment; Annotate the preprocessed sample images to obtain annotated image sets; The initial visual model is trained using the annotated picture set under a deep learning framework, and parameters in the initial visual model are adjusted.

3. The method according to claim 1, characterized in that Before merging the plurality of annotated images with the original training image set to construct a new training image set, the method further includes: Verify each annotated image to confirm the correctness of the initial AI visual model's recognition result of the annotated image; If the verification result shows that the recognition result of the initial AI visual model for the annotated picture is wrong, the annotated picture is classified into an abnormal picture library; wherein the abnormal picture library also includes annotated pictures that the initial AI visual model failed to recognize and in which unsafe behaviors actually occurred; Correspondingly, the step of merging the multiple annotated images with the original training image set to construct a new training image set includes: A plurality of annotated pictures are obtained from the abnormal picture library, and are merged with the original training picture set to construct the new training picture set.

4. The method according to claim 1, characterized in that: Before replacing the initial AI vision model with the retrained AI vision model, the method further includes: Analyzing the retrained AI vision model and the original AI vision model using the same test picture set, and calculating their respective accuracies; If the accuracy of the retrained AI vision model is higher than that of the original AI vision model, replacing the initial AI vision model with the retrained AI vision model; The retrained AI visual model and its corresponding training image set are updated to the cloud database for deployment and use in the underground coal mine environment.

5. A device for updating an AI visual model in a coal mine, characterized in that: include: The startup module is used to start the initial AI visual model and conduct video monitoring of the working scenes in the coal mine; An analysis module, configured to analyze the surveillance video using the initial AI visual model, and when unsafe behavior is identified in the surveillance video, extract at least one video frame containing the unsafe behavior as a target image; A labeling module, used to label each of the target images, wherein the labeling content includes the type of unsafe behavior and the location where the unsafe behavior occurs in the target image; A construction module is used to merge multiple annotated images with the original training image set to construct a new training image set; An updating module is used to retrain the initial AI vision model using the new training picture set, and replace the initial AI vision model with the retrained AI vision model.

6. The device according to claim 5, characterized in that The device also includes: An acquisition module is used to acquire multiple sample images related to coal mine operation scenes, and preprocess each of the sample images including denoising, edge enhancement, and brightness and contrast adjustment, and annotate the preprocessed sample images to obtain an annotated image set, and use the annotated image set to train the initial visual model under a deep learning framework, and adjust the parameters in the initial visual model.

7. The device according to claim 5, characterized in that The device also includes: A determination module is used to verify each annotated picture to confirm the correctness of the recognition result of the initial AI visual model for the annotated image. If the verification result shows that the recognition result of the initial AI visual model for the annotated picture is wrong, the annotated picture is classified into an abnormal picture library; wherein the abnormal picture library also includes annotated pictures that the initial AI visual model fails to recognize and in which unsafe behaviors actually occur; Correspondingly, the construction module is used to merge the multiple annotated pictures with the original training picture set to construct a new training picture set, including: A plurality of annotated pictures are obtained from the abnormal picture library, and are merged with the original training picture set to construct the new training picture set.

8. The device according to claim 5, characterized in that The updating module is used to replace the initial AI vision model with the retrained AI vision model, and is also used to: Analyzing the retrained AI vision model and the original AI vision model using the same test picture set, and calculating their respective accuracies; If the accuracy of the retrained AI vision model is higher than that of the original AI vision model, replacing the initial AI vision model with the retrained AI vision model; The retrained AI visual model and its corresponding training image set are updated to the cloud database for deployment and use in the underground coal mine environment.

9. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for updating the AI ​​visual model of an underground coal mine as described in any one of claims 1-4.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the program instructions are executed by the processor, the method for updating the AI ​​visual model of an underground coal mine described in any one of claims 1-4 is implemented.

Images