CN115995119B - Gas cylinder filling link illegal behavior identification method and system based on Internet of things - Google Patents
Gas cylinder filling link illegal behavior identification method and system based on Internet of things Download PDFInfo
- Publication number
- CN115995119B CN115995119B CN202310286287.8A CN202310286287A CN115995119B CN 115995119 B CN115995119 B CN 115995119B CN 202310286287 A CN202310286287 A CN 202310286287A CN 115995119 B CN115995119 B CN 115995119B
- Authority
- CN
- China
- Prior art keywords
- behavior
- image
- gas cylinder
- pipeline
- pipeline connection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a gas cylinder filling link illegal behavior recognition method and system based on the Internet of things, and relates to the technical field of behavior recognition. The method specifically comprises the following steps: editing a video of a gas cylinder filling site to obtain a gas cylinder image; labeling the gas cylinder image to obtain a behavior image showing behavior and a pipeline image showing pipeline connection condition; identifying the pipeline image according to the pipeline connection state detection model to obtain a pipeline connection state detection result; if the pipeline connection is normal, inputting the behavior image into a behavior detection model to conduct behavior classification, matching the behavior classification result with the actual gas cylinder label information input condition, and judging whether the rule is violated according to the matching result. According to the invention, the pipeline connection state, the label recording condition and the actions of the operators are comprehensively analyzed to obtain an accurate illegal behavior recognition result, so that subjective cheating behaviors of the operators are avoided, and the safety of the gas cylinder filling link can be ensured.
Description
Technical Field
The invention relates to the technical field of behavior recognition, in particular to a gas cylinder filling link illegal behavior recognition method and system based on the Internet of things.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The gas cylinder is used as a carrier of urban energy sources, and has the advantages of convenient transportation, strong sealing performance and the like. With the increasing popularity of gas cylinder use, various gas cylinders are spread throughout restaurants, homes, and other public places. And partial gas cylinder practitioners have weak safety consciousness, and illegal filling is performed in a gas cylinder filling link, so that a plurality of potential safety hazards are caused.
At present, various gas cylinders such as vehicle-mounted gas cylinders, industrial gas cylinders, liquefied petroleum gas cylinders, hydrogen cylinders and the like are additionally provided with or fixed with related gas cylinder electronic identification marks (two-dimensional codes, electronic RFID chips and other information carriers) according to related requirements in gas cylinder safety technical regulations in delivery links and filling links.
According to the filling requirement and specification of the gas cylinder, the electronic reading identification of the gas cylinder needs to be verified when the gas cylinder is filled, and after the gas cylinder is verified to be qualified, the filling operation can be performed; because part of filling personnel do not fill according to the flow required by the filling regulations at the filling site, cheating actions such as not scanning the electronic information marks of the gas cylinder or scanning the electronic information marks which are not additionally arranged on the gas cylinder body to carry out filling operation are adopted, supervision is avoided, and potential safety hazards are increased; meanwhile, the safety helmet is not worn according to requirements, tools are not worn, smoking, calling and filling are not carried out in a specified area, filling pipelines are not connected according to specified requirements, and other filling actions such as illegal operation and the like cause potential safety hazards in a filling link. At present, no effective means is available for supervising the illegal operation of the gas cylinder filling link and evidence collection and mark retention cannot be carried out.
The existing illegal behavior recognition technology mainly focuses on reading data in an RFID tag in a radio frequency recognition mode to judge whether the filling is in compliance or not, and cannot avoid the cheating behavior of scanning the RFID tag which is not additionally arranged on the gas cylinder body to carry out the filling operation. The existing gas cylinder filling environment violation identification process only judges whether the rule is violated or not according to whether the label information is input, and cannot be combined with the scanning action and the pipeline connection state to conduct the cheating behavior identification, so that great potential safety hazards are caused.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide a method and a system for identifying the illegal behavior of a gas cylinder filling link based on the Internet of things, which are used for identifying the illegal behavior through AI (advanced technology) based on the Internet of things, and specifically adopting a model to label and classify the gas cylinder state and the filling behavior in a target video, comprehensively analyze the pipeline connection state, the label recording condition and the action of an operator to obtain an accurate illegal behavior identification result, avoid the subjective cheating behavior of the operator and ensure the safety of the gas cylinder filling link.
In order to achieve the above object, the present invention is realized by the following technical scheme:
the invention provides a gas cylinder filling link illegal behavior identification method based on the Internet of things, which comprises the following steps:
acquiring a gas cylinder filling site video, and editing the gas cylinder filling site video to obtain a gas cylinder image;
labeling the gas cylinder image according to the image content to obtain a behavior image showing behavior actions and a pipeline image showing pipeline connection conditions;
identifying the pipeline image according to the pipeline connection state detection model to obtain a pipeline connection state detection result; performing model training by using a historical pipeline image dataset to obtain a pipeline connection state detection model;
if the pipeline connection is normal, inputting the behavior image into a behavior detection model to conduct behavior classification, and if the pipeline is identified to be not connected, not conducting the next processing operation, defaulting to be not in a filling link; performing model training by using a historical behavior image dataset to obtain a behavior detection model;
and matching the behavior classification result with the actual gas cylinder label information input condition, and judging whether the rule is violated according to the matching result.
Further, the specific steps of editing the gas cylinder filling site video include: and extracting key frames from the gas cylinder filling site video by adopting a frame difference maximum value inter-frame difference method to obtain a gas cylinder image.
Further, according to the image content, the image of the connection state of the pipeline is marked as a pipeline image, and the behavior image of the action of the target person is marked as a behavior image.
Further, the model training is performed by using a historical pipeline image dataset, and the specific steps for obtaining the pipeline connection state detection model are as follows:
acquiring a historical pipeline image to form a historical pipeline image data set, and performing image enhancement processing on pictures in the historical pipeline image data set;
the processed historical pipeline image data set is divided into a training set, a verification set and a test set, and is trained by using a YOLO V8 network to generate a pipeline connection state detection model.
Furthermore, the training is performed by using a YOLO V8 network, and the specific steps for generating the pipeline connection state detection model are as follows: configuring training parameters of a pipeline connection state detection model according to requirements, and calling a YOLO training command to start training; and verifying the model according to the training result of each time, and selecting the best fitting model as a pipeline connection state detection model.
Further, behavioral classification is divided into two categories: a body label scanning action and a valve opening action.
Further, the model training is performed by using the historical behavior image dataset to obtain a behavior detection model, and the specific steps are as follows:
and acquiring a historical behavior image to form a historical behavior image data set, and training by using a YOLOV8+SlowFast network after the data set is established to generate a behavior detection model.
Further, the specific steps of inputting the behavior image into the behavior detection model for behavior classification are as follows:
realizing target detection by adopting a target detection algorithm, and determining initial coordinates of the target;
tracking the target based on the initial target coordinates, and continuously labeling the target coordinates.
Performing action recognition on the target based on the target coordinate movement track;
and continuously framing the target by using the detection frame, and displaying the action recognition result and the confidence level on the frame.
Further, the behavior classification result is matched with the actual gas cylinder label information input condition, and the specific steps of judging whether the rule is violated according to the matching result are as follows:
on the premise that the connection of the identification pipeline is normal, confirming whether the repeater receives RFID tag information or not, and if the RFID tag information does not exist and the valve opening action is identified, alarming is carried out; if the valve opening action is not recognized, continuing to confirm whether the relay receives the RFID tag information or not until the relay receives the RFID tag information;
when the relay receives the RFID tag information, confirming whether the target has the action of scanning the bottle body tag, if so, judging the tag content by the relay, and determining whether to charge; if the valve opening action is not recognized, alarming is carried out; if the valve opening action is not recognized, the relay is continuously confirmed whether the RFID tag information is received.
The second aspect of the invention provides a gas cylinder filling link violation identification system based on the Internet of things, which comprises the following steps:
the image acquisition module is configured to acquire a gas cylinder filling site video, clip the gas cylinder filling site video and acquire a gas cylinder image;
the image labeling module is configured to label the gas cylinder image according to the image content to obtain a behavior image showing behavior actions and a pipeline image showing pipeline connection conditions;
the pipeline connection state identification module is configured to identify the pipeline image according to the pipeline connection state detection model to obtain a pipeline connection state detection result; performing model training by using a historical pipeline image dataset to obtain a pipeline connection state detection model;
the image classification module is configured to input the behavior image into the behavior detection model for behavior classification if the pipeline connection is normal, and defaults to not enter a filling link if the pipeline is identified to be not connected and the next processing operation is not performed; performing model training by using a historical behavior image dataset to obtain a behavior detection model;
and the illegal behavior judging module is configured to match the behavior classification result with the actual gas cylinder label information input condition and judge whether the behavior is illegal according to the matching result.
The one or more of the above technical solutions have the following beneficial effects:
the invention discloses a method and a system for identifying the illegal behavior of a gas cylinder filling link based on the Internet of things, which are used for identifying the illegal behavior of the gas cylinder filling link through AI and giving an alarm to the illegal behavior in real time, specifically, a model is respectively established for detecting the pipeline connection condition and the behavior illegal condition through clipping and labeling of a field video, and the gas cylinder filling link safety can be ensured by combining the read RFID information according to the gas cylinder filling pipe connection condition, the scanning gas cylinder label action and the valve opening action, so that an accurate illegal behavior identification result is obtained, the subjective cheating behavior of an operator is avoided, and the gas cylinder filling link safety is ensured.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
Fig. 1 is a flowchart of a method for identifying rule violations in a gas cylinder filling link based on the internet of things in a first embodiment of the invention.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It should be noted that, in the embodiments of the present invention, related data such as a gas cylinder filling video is related, when the above embodiments of the present invention are applied to specific products or technologies, user permission or consent is required to be obtained, and the collection, use and processing of related data is required to comply with related laws and regulations and standards of related countries and regions.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
Term interpretation:
RFID: radio Frequency Identification, radio frequency identification.
AI: artificial Intelligence, artificial intelligence.
YOLO: you only look once you need only look at once, AI proper nouns, is a network for target detection. YOLOV8: (version 8) YOLO version 8. YOLOV8s: the small version of V8.
Anchor Box: and an anchor frame.
EIOU: effective cross-correlation ratio.
CIOU: complete-IOU, complete cross-correlation.
mAP: mean Average Precision, average accuracy.
SlowFast: slow, proper nouns, a video recognition method.
IPC: IP Camera, network Camera.
pytorch: proper nouns, a deep learning framework.
pytorchvideo: the pytorch video behavior analysis framework.
detectron: the proper noun is a Facebook AI Institute (FAIR) open-source software system, and the most advanced target detection algorithm is realized.
detectron2: is a deep learning framework of facebook AI research (FAIR) reconstruction detecton.
Faster R-CNN: R-CNN is Regions with CNN features, CNN is Convolutional Neural Networks, convolutional neural network, and Faster R-CNN is a framework for realizing target detection task of an image by using an end-to-end deep learning model.
deepsort: depth tracking, proper noun, a target tracking algorithm.
Neck, a part of the Neck, proper noun, yolo framework.
Focal Loss: center loss, AI proper nouns.
BCELoss: binary Cross Entropy Loss, binary cross entropy loss.
Embodiment one:
the first embodiment of the invention provides a method for identifying the illegal behavior of a gas cylinder filling link based on the Internet of things, which utilizes AI to identify the illegal behavior. And (5) finding hidden danger and timely alarming and notifying. The method specifically comprises the following steps:
step 1, acquiring a gas cylinder filling site video, and editing the gas cylinder filling site video to obtain a gas cylinder image.
And 2, labeling the gas cylinder image according to the image content to obtain a behavior image showing behavior and a pipeline image showing pipeline connection condition.
And step 3, identifying the pipeline image according to the pipeline connection state detection model to obtain a pipeline connection state detection result.
And step 4, if the pipeline connection is normal, inputting the behavior image into a behavior detection model to conduct behavior classification, and if the pipeline is identified to be not connected, not conducting the next processing operation, defaulting to be not in a filling link.
And step 5, matching the behavior classification result with the actual gas cylinder label information input condition, and judging whether the rule is violated according to the matching result.
In the step 1, video is clipped, video containing the contents of a connecting gas cylinder filling pipe, a scanning gas cylinder label and valve opening behavior is selected, the length of a video clip can be set to 3-10 seconds, the integrity of recognition action is ensured, and meanwhile, the rapidity of recognition is ensured. The method for editing the video of the gas cylinder filling site comprises the following specific steps of: and extracting key frames from the gas cylinder filling site video by adopting a frame difference maximum value inter-frame difference method to obtain a gas cylinder image.
In step 2, labeling is performed on the gas cylinder image obtained by extracting the key frame in step 1, the image shot in the pipeline connection state is labeled as a pipeline image according to the image content, and the behavior image shot in the action of the target person is labeled as a behavior image. In addition to the line image and the behavior image, the remaining images are discarded.
The labeling process in this embodiment is manual labeling+automatic labeling. The process is as follows:
1. a small amount of data is manually noted.
2. Based on the existing labeling rules and data sets, iterative training is performed by using a YOLO V8 network. The data set is divided into a training set and a verification set, and the training set is utilized to train the YOLO V8 network to obtain a target detection model. In the training process, a verification set with known labeling data is adopted to verify the model obtained through training, the highest labeling accuracy of the verification set is recorded as the optimal accuracy, and training is stopped when the optimal accuracy is not reached in 10 continuous loop iterations, so that the model with the optimal accuracy is obtained as the optimal target detection model in the current stage.
3. And (5) automatically labeling by adopting a model reasoning method.
4. And correcting the labeling result by adopting a manual mode.
5. The corrected data is re-added to the training set to train a more optimal model.
6. Through loop iteration of 2-5 steps, an optimal model can be trained step by step for labeling.
In the step 3, model training is carried out by utilizing a historical pipeline image data set, and the specific steps for obtaining the pipeline connection state detection model are as follows:
(1) And acquiring a historical pipeline image to form a historical pipeline image data set, and performing image enhancement processing on pictures in the historical pipeline image data set. Mosaic enhancement (Mosaic), mix enhancement (mix up), spatial perturbation (random perspective), and color+perturbation (HSV) image enhancement.
In the embodiment, a mosaic enhancement method is adopted, 4 pictures are randomly extracted from a historical pipeline image dataset, and are spliced into a new picture according to a random scaling, random cutting and random arrangement mode, so that the new picture is supplemented into the original dataset, and the dataset is greatly enriched.
(2) The processed historical pipeline image data set is divided into a training set, a verification set and a test set, and is trained by using a YOLO V8 network to generate a pipeline connection state detection model.
According to the embodiment, a YOLO V8 network is adopted, so that an Anchor box is not required to be arranged, the position and size information of a gas bottle opening detection frame can be directly obtained from an image, and whether a bottle opening is connected with a pipeline or not is identified.
Specifically, the marked historical pipeline image data set is divided into a training set, a verification set and a test set. And configuring training parameters of the pipeline connection state detection model according to requirements, such as training batch size, training process iteration times, picture size, learning rate, optimizers and the like, and calling a YOLO training command to start training. In the training process, the models are verified according to each training result, and the best fitting model is selected and used as a pipeline connection state detection model.
The invention improves the traditional YOLO V8 network, properly simplifies the Neck region of the traditional YOLO V8s model, and deletes 19X 19 feature map branches suitable for detecting objects with larger sizes, thereby reducing the complexity of the model; the EIOU loss function is used for replacing the CIOU loss function used by the YOLO V8s algorithm to optimize the training model, so that the accuracy of the algorithm is improved, and the detection instantaneity is improved. Compared with the traditional YOLO V8s model, the improved YOLO V8s model can accurately detect whether pipelines are connected, the size of the improved YOLO V8s model is reduced by 2.15 and MB, the mAP of the improved model algorithm reaches 0.932, and the improved YOLO V8s model is improved by 6.3 percent compared with the traditional YOLO V8s model.
In step 4, the behavior image is input into a behavior detection model to perform behavior classification, and in this embodiment, the behavior classification is divided into two classes: a body label scanning action and a valve opening action. Model training is carried out by utilizing a historical behavior image dataset to obtain a behavior detection model, and the specific steps are as follows:
the method comprises the steps of collecting historical behavior images to form a historical behavior image data set, and training by using a YOLO V8+SlowFast network after the data set is established to generate a behavior detection model so as to identify body scanning body label actions and valve opening actions.
The embodiment adopts a pytorchvideo framework to combine target detection and behavior classification to realize behavior detection. The target detection framework under the pytorchvideo framework is the fast R-CNN under its own detectron2 tool, which is slower, and behavior detection is discontinuous, so this embodiment replaces with YOLO V8 network. The behavior classification adopts a behavior classification framework SlowFast under a pytorchvideo framework, so that a YOLOV8+SlowFast network is formed.
The specific steps of inputting the behavior image into the behavior detection model for behavior classification are as follows:
realizing target detection by adopting a target detection algorithm, and determining initial coordinates of the target;
tracking the target based on the initial target coordinates, and continuously labeling the target coordinates.
Performing action recognition on the target based on the target coordinate movement track;
and continuously framing the target by using the detection frame, and displaying the action recognition result and the confidence level on the frame.
More specifically, the behavior image is input into the YOLO V8 network, and when the YOLO V8 is utilized to carry out target detection by using a self-contained target detection algorithm and is executed frame by frame, the target detection frame can be seen to move along with the target, and the deepsort is adopted to carry out target tracking. And then inputting the video sequence and the detection frame information into a SlowFast network to conduct behavior classification, and outputting the behavior type of each detection frame to achieve the purpose of behavior detection. The dataset is derived from video taken by the filling site IPC. In this embodiment, the video is clipped into a section of 10 seconds, 30 frames are cut every second, and behavior labeling is performed on the behavior image after classification, and labeling information includes: video name, video frame number, coordinate value, behavior class number. The coordinate values are coordinate values of the target, and refer to XY-axis coordinates of the upper left corner and the lower right corner of a rectangular frame marked after the whole target person in the video is identified, namely (x 1, y1, x2, y 2).
The embodiment also improves the YOLO V8 network, appropriately simplifies the neg region of the YOLO V8 network, and deletes the 19×19 feature map branch suitable for detecting the object with larger size, thereby reducing the complexity of the model; the EIOU loss function is used for replacing the CIOU loss function used by the YOLOv8s algorithm to optimize the training model, so that the accuracy of the algorithm is improved, and the detection instantaneity is improved.
The SlowFast is a novel video recognition method, which can simulate the retina nerve operation principle in primate vision, and extract effective information in video at a slow frame frequency and a fast frame frequency, so as to improve action classification and action recognition effects. Compared with other methods, the overall computation complexity of the SlowFast is lower, and the accuracy is higher. In the embodiment, the Focal Loss function is used for replacing the BCELoss Loss function in the SlowFast, so that the problem of unbalance of positive and negative sample data is solved.
In the embodiment, the improved YOLO V8 network is used for realizing target detection, deepsort is used for realizing target tracking, and slow is used for realizing action recognition and giving out confidence coefficient, so that the identification of the body label scanning action and the valve opening action of the body is realized.
In particular, the label scanning operation here is an operation of scanning a label on a bottle body by a target person, and is not an operation of scanning a label at any position, and thus it is necessary to trace the action trace of the target person, in order to avoid the cheating act of scanning an RFID label not attached to a bottle body to perform a filling operation. In addition, the YOLOV8+SlowFast network can be used for identifying potential safety hazard behaviors such as wearing no safety helmet, wearing no tool, smoking, making a call and the like, historical violation data are trained by utilizing the YOLOV8+SlowFast network to obtain other violation recognition models, and once the behaviors are detected, an alarm is directly given, so that the safety behavior standardization degree of a gas cylinder filling link is further improved.
In step 5, the behavior classification result is matched with the actual gas cylinder label information input condition, and whether the rule is violated or not is judged according to the matching result, as shown in fig. 1:
if the pipeline is identified to be not connected, the next processing operation is not performed, and the filling link is not defaulted;
on the premise that the connection of the identification pipeline is normal, confirming whether the repeater receives RFID tag information or not, and if the RFID tag information does not exist and the valve opening action is identified, alarming is carried out; if the valve opening action is not recognized, continuing to confirm whether the relay receives the RFID tag information or not until the relay receives the RFID tag information;
when the relay receives the RFID tag information, confirming whether the target has the action of scanning the bottle body tag, if so, judging the tag content by the relay, and determining whether to charge; if the valve opening action is not recognized, alarming is carried out; if the valve opening action is not recognized, the relay is continuously confirmed whether the RFID tag information is received.
Therefore, the embodiment can carry out valve opening filling only when the label content is matched, the pipeline connection is normal, the RFID label information is received and the bottle body label scanning action is carried out, otherwise, any valve opening action can give an alarm, the illegal action judgment logic of the embodiment is clear and reasonable, the probability of misjudgment is reduced while the cheating action is avoided, the illegal action in the gas bottle filling process can be effectively identified, and the safety guarantee is provided for gas bottle filling.
Embodiment two:
the second embodiment of the invention provides a gas cylinder filling link violation identification system based on the Internet of things, which comprises the following steps:
the image acquisition module is configured to acquire a gas cylinder filling site video, clip the gas cylinder filling site video and acquire a gas cylinder image;
the image labeling module is configured to label the gas cylinder image according to the image content to obtain a behavior image showing behavior actions and a pipeline image showing pipeline connection conditions;
the pipeline connection state identification module is configured to identify the pipeline image according to the pipeline connection state detection model to obtain a pipeline connection state detection result; performing model training by using a historical pipeline image dataset to obtain a pipeline connection state detection model;
the image classification module is configured to input the behavior image into the behavior detection model for behavior classification if the pipeline connection is normal, and defaults to not enter a filling link if the pipeline is identified to be not connected and the next processing operation is not performed; performing model training by using a historical behavior image dataset to obtain a behavior detection model;
and the illegal behavior judging module is configured to match the behavior classification result with the actual gas cylinder label information input condition and judge whether the behavior is illegal according to the matching result.
The steps involved in the second embodiment correspond to those of the first embodiment of the method, and the detailed description of the second embodiment can be found in the related description section of the first embodiment. It will be appreciated by those skilled in the art that the modules or steps of the invention described above may be implemented by general-purpose computer means, alternatively they may be implemented by program code executable by computing means, whereby they may be stored in storage means for execution by computing means, or they may be made into individual integrated circuit modules separately, or a plurality of modules or steps in them may be made into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.
Claims (9)
1. The gas cylinder filling link illegal behavior identification method based on the Internet of things is characterized by comprising the following steps of:
acquiring a gas cylinder filling site video, and editing the gas cylinder filling site video to obtain a gas cylinder image;
labeling the gas cylinder image according to the image content to obtain a behavior image showing behavior actions and a pipeline image showing pipeline connection conditions;
identifying the pipeline image according to the pipeline connection state detection model to obtain a pipeline connection state detection result; performing model training by using a historical pipeline image dataset to obtain a pipeline connection state detection model;
if the pipeline connection is normal, inputting the behavior image into a behavior detection model to conduct behavior classification, and if the pipeline is identified to be not connected, not conducting the next processing operation, defaulting to be not in a filling link; performing model training by using a historical behavior image dataset to obtain a behavior detection model; behavior classification is divided into two categories: scanning a bottle label and opening a valve;
matching the behavior classification result with the actual gas cylinder label information input condition, and judging whether the rule is violated according to the matching result; on the premise that the connection of the identification pipeline is normal, confirming whether the relay receives RFID tag information, when the relay receives the RFID tag information, confirming whether a target has a scanning bottle body tag action, if so, judging the tag content by the relay, and determining whether to charge; if the valve opening action is not recognized, alarming is carried out; if the valve opening action is not recognized, the relay is continuously confirmed whether the RFID tag information is received.
2. The method for identifying the illegal behavior of the gas cylinder filling link based on the internet of things according to claim 1, wherein the specific step of editing the video of the gas cylinder filling site comprises the following steps: and extracting key frames from the gas cylinder filling site video by adopting a frame difference maximum value inter-frame difference method to obtain a gas cylinder image.
3. The method for identifying the illegal gas cylinder filling link based on the Internet of things according to claim 1 is characterized in that the image of the shot pipeline connection condition is marked as a pipeline image and the behavior image of the shot target person action is marked as a behavior image according to the image content.
4. The method for identifying the illegal behavior of the gas cylinder filling link based on the internet of things according to claim 1, wherein the specific steps of obtaining the pipeline connection state detection model by using the historical pipeline image dataset are as follows:
acquiring a historical pipeline image to form a historical pipeline image data set, and performing image enhancement processing on pictures in the historical pipeline image data set;
the processed historical pipeline image data set is divided into a training set, a verification set and a test set, and is trained by using a YOLO V8 network to generate a pipeline connection state detection model.
5. The method for identifying the illegal behavior of the filling link of the gas cylinder based on the Internet of things according to claim 4, wherein training is performed by using a YOLO V8 network, and the specific steps of generating a pipeline connection state detection model are as follows: configuring training parameters of a pipeline connection state detection model according to requirements, and calling a YOLO training command to start training; and verifying the model according to the training result of each time, and selecting the best fitting model as a pipeline connection state detection model.
6. The method for identifying the illegal behavior of the gas cylinder filling link based on the Internet of things according to claim 1, wherein the model training is performed by using a historical behavior image dataset to obtain a behavior detection model, and the specific steps are as follows:
and acquiring a historical behavior image to form a historical behavior image data set, and training by using a YOLOV8+SlowFast network after the data set is established to generate a behavior detection model.
7. The method for identifying the illegal behavior of the gas cylinder filling link based on the internet of things according to claim 6, wherein the specific steps of inputting the behavior image into the behavior detection model to conduct behavior classification are as follows:
realizing target detection by adopting a target detection algorithm, and determining initial coordinates of the target;
tracking the target based on the initial target coordinates, and continuously labeling the target coordinates;
performing action recognition on the target based on the target coordinate movement track;
and continuously framing the target by using the detection frame, and displaying the action recognition result and the confidence level on the frame.
8. The method for identifying the illegal behavior of the gas cylinder filling link based on the Internet of things according to claim 1, wherein the specific steps of matching the behavior classification result with the actual gas cylinder label information input condition and judging whether the gas cylinder filling link is illegal according to the matching result are as follows:
on the premise that the connection of the identification pipeline is normal, confirming whether the repeater receives RFID tag information or not, and if the RFID tag information does not exist and the valve opening action is identified, alarming is carried out; if the valve opening action is not recognized, continuing to confirm whether the relay receives the RFID tag information or not until the relay receives the RFID tag information.
9. Gas cylinder fills link illegal behavior identification system based on thing networking, its characterized in that includes:
the image acquisition module is configured to acquire a gas cylinder filling site video, clip the gas cylinder filling site video and acquire a gas cylinder image;
the image labeling module is configured to label the gas cylinder image according to the image content to obtain a behavior image showing behavior actions and a pipeline image showing pipeline connection conditions;
the pipeline connection state identification module is configured to identify the pipeline image according to the pipeline connection state detection model to obtain a pipeline connection state detection result; performing model training by using a historical pipeline image dataset to obtain a pipeline connection state detection model;
the image classification module is configured to input the behavior image into the behavior detection model for behavior classification if the pipeline connection is normal, and defaults to not enter a filling link if the pipeline is identified to be not connected and the next processing operation is not performed; performing model training by using a historical behavior image dataset to obtain a behavior detection model; behavior classification is divided into two categories: scanning a bottle label and opening a valve;
the illegal behavior judging module is configured to match the behavior classification result with the actual gas cylinder label information input condition and judge whether the behavior is illegal according to the matching result; on the premise that the connection of the identification pipeline is normal, confirming whether the relay receives RFID tag information, when the relay receives the RFID tag information, confirming whether a target has a scanning bottle body tag action, if so, judging the tag content by the relay, and determining whether to charge; if the valve opening action is not recognized, alarming is carried out; if the valve opening action is not recognized, the relay is continuously confirmed whether the RFID tag information is received.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310286287.8A CN115995119B (en) | 2023-03-23 | 2023-03-23 | Gas cylinder filling link illegal behavior identification method and system based on Internet of things |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310286287.8A CN115995119B (en) | 2023-03-23 | 2023-03-23 | Gas cylinder filling link illegal behavior identification method and system based on Internet of things |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115995119A CN115995119A (en) | 2023-04-21 |
| CN115995119B true CN115995119B (en) | 2023-07-28 |
Family
ID=85993820
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310286287.8A Active CN115995119B (en) | 2023-03-23 | 2023-03-23 | Gas cylinder filling link illegal behavior identification method and system based on Internet of things |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115995119B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116189099B (en) * | 2023-04-25 | 2023-10-10 | 南京华苏科技有限公司 | Method for detecting and stacking exposed garbage based on improved yolov8 |
| CN116229278B (en) * | 2023-05-10 | 2023-08-04 | 广东电网有限责任公司珠海供电局 | Method and system for detecting rust defect of vibration damper of power transmission line |
| CN117115640A (en) * | 2023-07-04 | 2023-11-24 | 北京市农林科学院 | Improved YOLOv 8-based pest and disease damage target detection method, device and equipment |
| CN117437224B (en) * | 2023-12-20 | 2024-03-29 | 山东特联信息科技有限公司 | Gas cylinder defect image data processing system and method based on artificial intelligence |
| CN118692015A (en) * | 2024-08-28 | 2024-09-24 | 山东特联信息科技有限公司 | AI video-based intelligent filling supervision system and method for liquefied petroleum gas cylinder |
| CN118705537A (en) * | 2024-08-30 | 2024-09-27 | 山东特联信息科技有限公司 | Industrial gas cylinder filling safety management system and method based on AI video |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488646A (en) * | 2015-11-25 | 2016-04-13 | 广东未来信息技术有限公司 | Intelligent supervision device and method of bottled liquefied gas supply based on Internet of Things |
| CN205507474U (en) * | 2015-11-25 | 2016-08-24 | 广东未来信息技术有限公司 | Bottled gas intelligence filling control equipment based on internet of things |
| WO2022022368A1 (en) * | 2020-07-28 | 2022-02-03 | 宁波环视信息科技有限公司 | Deep-learning-based apparatus and method for monitoring behavioral norms in jail |
Family Cites Families (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN203414785U (en) * | 2013-06-27 | 2014-01-29 | 成都弘巨立物联网科技有限公司 | Gas station and CNG (Compressed Natural Gas) refueling station safety supervision system based on Internet of Things |
| US11165797B2 (en) * | 2016-04-22 | 2021-11-02 | Sophos Limited | Detecting endpoint compromise based on network usage history |
| CN110456723A (en) * | 2019-08-15 | 2019-11-15 | 成都睿晓科技有限公司 | A kind of emptying area, gas station security management and control system based on deep learning |
| CN110490126B (en) * | 2019-08-15 | 2023-04-18 | 成都睿晓科技有限公司 | Safe deposit box safety control system based on artificial intelligence |
| CN211403204U (en) * | 2020-03-16 | 2020-09-01 | 厚普智慧物联科技有限公司 | Hydrogen energy fills dress management and control system based on AI is supplementary |
| CN111290325A (en) * | 2020-03-16 | 2020-06-16 | 厚普智慧物联科技有限公司 | AI (Artificial intelligence) -assistance-based hydrogen energy filling management and control method and system |
| CN111445705B (en) * | 2020-05-24 | 2024-07-19 | 威海伟丰智能工程有限公司 | Use method of fuel gas vehicle filling traceability system based on vehicle identification technology |
| CN111898934A (en) * | 2020-06-17 | 2020-11-06 | 东营市东电网络科技有限公司 | Hazardous chemical substance road transportation informatization safety supervision system |
| CN112004063B (en) * | 2020-09-03 | 2022-04-22 | 四川弘和通讯有限公司 | Method for monitoring connection correctness of oil discharge pipe based on multi-camera linkage |
| CN112270262B (en) * | 2020-10-28 | 2022-04-08 | 福州市君则信息技术有限公司 | Intelligent filling gas cylinder detection system and method |
| CN113076683B (en) * | 2020-12-08 | 2023-08-08 | 国网辽宁省电力有限公司锦州供电公司 | Modeling method of convolutional neural network model for transformer substation behavior monitoring |
| CN112686595A (en) * | 2020-12-15 | 2021-04-20 | 上海东普信息科技有限公司 | Method, device, equipment and storage medium for detecting illegal behavior of logistics operation |
| CN112712252A (en) * | 2020-12-28 | 2021-04-27 | 黄兴良 | Intelligent vehicle gas cylinder dynamic supervision method and system |
| CN113011300A (en) * | 2021-03-10 | 2021-06-22 | 中用科技有限公司 | Method, system and equipment for AI visual identification of violation behavior |
| CN113327420A (en) * | 2021-06-03 | 2021-08-31 | 安徽达尔智能控制系统股份有限公司 | Traffic monitoring system based on wisdom traffic thing networking |
| CN115635864A (en) * | 2021-07-19 | 2023-01-24 | 郑州闪象新能源科技有限公司 | Vehicle end detection method for live-line gun drawing and vehicle |
| CN113901906A (en) * | 2021-09-30 | 2022-01-07 | 正星科技股份有限公司 | Mobile risk early warning method and device for fuel truck nozzle, electronic equipment and medium |
| CN114727069A (en) * | 2022-04-22 | 2022-07-08 | 北京国旺盛源智能终端科技有限公司 | Charging station maintenance method, device, equipment and computer readable medium |
| CN115188077A (en) * | 2022-07-18 | 2022-10-14 | 东莞先知大数据有限公司 | Detection method, device and equipment for illegal charging of battery and storage medium |
| CN115527393A (en) * | 2022-09-27 | 2022-12-27 | 上汽通用五菱汽车股份有限公司 | Parking method for preventing illegal parking of charging pile parking space and computer readable storage medium |
-
2023
- 2023-03-23 CN CN202310286287.8A patent/CN115995119B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488646A (en) * | 2015-11-25 | 2016-04-13 | 广东未来信息技术有限公司 | Intelligent supervision device and method of bottled liquefied gas supply based on Internet of Things |
| CN205507474U (en) * | 2015-11-25 | 2016-08-24 | 广东未来信息技术有限公司 | Bottled gas intelligence filling control equipment based on internet of things |
| WO2022022368A1 (en) * | 2020-07-28 | 2022-02-03 | 宁波环视信息科技有限公司 | Deep-learning-based apparatus and method for monitoring behavioral norms in jail |
Non-Patent Citations (3)
| Title |
|---|
| Crowdsourced reliable labeling of safety-rule violations on images of complex construction scenes for advanced vision-based workplace safety;Yanyu Wang等;Advanced Engineering Informatics;第42卷;1-16 * |
| 基于ResNet-18-KNN的变电站人员电气违规行为识别;白景涛;周照宇;王中一;孙黎霞;;机械制造与自动化(02);174-177 * |
| 基于物联网的车用气瓶监管系统;马玉鹏等;《计算机与现代化》(第3期);96-100 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115995119A (en) | 2023-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115995119B (en) | Gas cylinder filling link illegal behavior identification method and system based on Internet of things | |
| CN111626123B (en) | Video data processing method, device, computer equipment and storage medium | |
| CN107808358B (en) | Automatic detection method for image watermark | |
| CN110543857A (en) | Contraband identification method, device and system based on image analysis and storage medium | |
| CN106557747B (en) | The method and device of identification insurance single numbers | |
| CN110009913A (en) | A kind of non-at-scene law enforcement picture intelligent checks method and system of vehicles running red light | |
| CN108170813A (en) | A kind of method and its system of full media content intelligent checks | |
| CN111914778B (en) | Video behavior positioning method based on weak supervision learning | |
| CN110969166A (en) | Small target identification method and system in inspection scene | |
| CN108491821A (en) | Vehicle insurance accident discrimination method, system and storage medium based on image procossing and deep learning | |
| CN113642474A (en) | Hazardous area personnel monitoring method based on YOLOV5 | |
| CN105095856A (en) | Method for recognizing human face with shielding based on mask layer | |
| CN111414813A (en) | Dangerous driving behavior identification method, device, equipment and storage medium | |
| CN112329719B (en) | Behavior recognition method, behavior recognition device and computer-readable storage medium | |
| CN114898466B (en) | Intelligent factory-oriented video action recognition method and system | |
| CN112419268A (en) | Method, device, equipment and medium for detecting image defects of power transmission line | |
| CN113298077A (en) | Transformer substation foreign matter identification and positioning method and device based on deep learning | |
| CN111862065A (en) | Power transmission line diagnosis method and system based on multitask deep convolutional neural network | |
| CN114494103A (en) | Defect detection method and device | |
| CN115830399A (en) | Classification model training method, apparatus, device, storage medium, and program product | |
| CN111091104A (en) | Target object protection detection method, device, equipment and storage medium | |
| CN117115158B (en) | Defect detection method and device based on deep contrast learning | |
| CN118038555A (en) | Dangerous behavior detection method for chemical enterprise personnel based on improvement YOLOv8 | |
| CN116189063B (en) | Key frame optimization method and device for intelligent video monitoring | |
| CN111199172A (en) | Terminal screen recording-based processing method and device and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |