CN117876972B - Workshop safety supervision method and system based on internet of things perception - Google Patents
Workshop safety supervision method and system based on internet of things perception Download PDFInfo
- Publication number
- CN117876972B CN117876972B CN202410277674.XA CN202410277674A CN117876972B CN 117876972 B CN117876972 B CN 117876972B CN 202410277674 A CN202410277674 A CN 202410277674A CN 117876972 B CN117876972 B CN 117876972B
- Authority
- CN
- China
- Prior art keywords
- workshop
- safety
- principal component
- human
- data
- 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
- 238000000034 method Methods 0.000 title claims abstract description 42
- 230000008447 perception Effects 0.000 title claims abstract description 19
- 239000013598 vector Substances 0.000 claims abstract description 67
- 230000006399 behavior Effects 0.000 claims abstract description 57
- 230000009467 reduction Effects 0.000 claims abstract description 23
- 210000000988 bone and bone Anatomy 0.000 claims abstract description 16
- 238000000605 extraction Methods 0.000 claims abstract description 16
- 238000007781 pre-processing Methods 0.000 claims abstract description 16
- 238000005457 optimization Methods 0.000 claims abstract description 14
- 238000000513 principal component analysis Methods 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 10
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 239000011159 matrix material Substances 0.000 claims description 39
- 230000000737 periodic effect Effects 0.000 claims description 18
- 238000012544 monitoring process Methods 0.000 claims description 15
- 238000001228 spectrum Methods 0.000 claims description 12
- 238000012417 linear regression Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 9
- 230000002159 abnormal effect Effects 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 6
- 238000006116 polymerization reaction Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000002123 temporal effect Effects 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 abstract 1
- 230000009471 action Effects 0.000 description 3
- 230000036541 health Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the field of internet of things perception, in particular to a workshop safety supervision method and system based on internet of things perception, wherein the method comprises safety threshold management, data acquisition and preprocessing, workshop data characteristic and human behavior characteristic extraction, human behavior characteristic dimension reduction, data and human behavior safety prediction, model evaluation analysis and optimization; according to the invention, the space-time feature vectors of human bones are extracted by using the GCN and LSTM models, the features are subjected to dimension reduction by using the self-adaptive principal component analysis algorithm based on self-adaptive learning rate optimization, the robustness, the interpretability and the data processing efficiency of the models are increased, so that the human behavior classification prediction model can be better applied to a workshop safety supervision system, and the safety coefficient and the performance of the workshop safety supervision are increased.
Description
Technical Field
The invention belongs to the field of internet of things sensing, and particularly relates to a workshop safety supervision method and system based on internet of things sensing.
Background
Workshop safety supervision refers to safety management and supervision of an industrial workshop so as to ensure the safety of working environment and the health of staff, prevent and reduce accidents, and protect the life safety and health of workers and ensure the safe operation of the workshop. However, the existing workshop safety supervision system cannot conduct safety prediction on the workshop environment according to the change trend of workshop environment data, and in the existing supervision system, the robustness of a human behavior prediction model is poor, the prediction rate is low, and timeliness and accuracy of workshop human behavior safety supervision cannot be met.
Disclosure of Invention
Aiming at the situation, in order to overcome the defects of the prior art, the invention provides a workshop safety supervision method and a system based on the internet of things perception, and aims at solving the problem that the existing workshop safety supervision system cannot safely predict the workshop environment according to the change trend of workshop environment data; aiming at the problems that in the existing supervision system, the robustness of a human behavior prediction model is poor, the prediction speed is low, and timeliness and accuracy of workshop human behavior safety supervision cannot be met, the invention uses a GCN model and an LSTM model to extract space feature vectors of human bones, and uses a self-adaptive principal component analysis algorithm based on self-adaptive learning rate optimization to reduce dimensions of the space feature vectors of the human bones, so that the robustness, the interpretability and the data processing efficiency of the model are improved, and finally uses a softmax function to classify and predict the space feature vectors of the human bones after dimension reduction, so that the human behavior classification prediction model can be better applied to a workshop safety supervision system, and the safety coefficient and the performance of workshop safety supervision are increased.
The technical scheme adopted by the invention is as follows: the invention provides a workshop safety supervision method based on internet of things perception, which specifically comprises the following steps:
Step S1: safety threshold management, namely setting safety thresholds of workshop temperature, workshop humidity and workshop gas concentration;
Step S2: workshop data acquisition, namely acquiring workshop temperature, workshop humidity, workshop gas concentration and human behavior images in a workshop in real time by using a temperature sensor, a humidity sensor, a gas concentration sensor and a high-definition camera;
Step S3: data preprocessing, namely alarming missing values, abnormal values and outliers of workshop temperature, workshop humidity and workshop gas concentration, judging whether the data are real potential safety hazard data or error data, continuously sending out the alarm until the data return to normal if the data are the real potential safety hazard data, removing the error data to obtain a preprocessed data set, adjusting the size of a collected human behavior image in the workshop to be uniform and fixed, and carrying out noise reduction processing on the image to obtain the preprocessed human behavior image set;
Step S4: extracting data set characteristics and predicting data safety, extracting trend characteristics of the preprocessed data set by using a first derivative, extracting periodic characteristics of the preprocessed data set by using a frequency domain analysis method, and carrying out safety prediction and alarm on workshop temperature, workshop humidity and workshop gas concentration according to the trend characteristics and the periodic characteristics, sending out early warning signals and storing early warning information;
Step S5: extracting features of the image set, and extracting features of the human behavior image set by using a human feature extraction model to obtain space-time feature vectors of human bones;
step S6: feature dimension reduction, namely performing dimension reduction on the space-time feature vector of the human skeleton by using an adaptive principal component analysis algorithm based on adaptive learning rate optimization to obtain the space-time feature vector of the human skeleton after dimension reduction;
Step S7: classifying behaviors, namely classifying and predicting space-time feature vectors of the human bones after dimension reduction by using a softmax function to obtain human behavior classification prediction results;
Step S8: judging human behaviors, judging whether potential safety hazards exist in human behaviors according to human behavior classification prediction results, and if the potential safety hazards exist, sending out early warning signals and storing early warning information;
step S9: and (3) evaluating and optimizing the model, evaluating the human body characteristic extraction model and the self-adaptive principal component analysis algorithm based on self-adaptive learning rate optimization, and optimizing and improving the human body characteristic extraction model and the self-adaptive principal component analysis algorithm based on the self-adaptive learning rate optimization according to feedback and requirements in practical application.
Further, step S4 specifically includes the following steps:
step S41, extracting trend characteristics of the preprocessed data set by using the first derivative;
Step S42, extracting periodic characteristics of the preprocessed data set by using a frequency domain analysis method based on Fourier transform;
Further, step S42 specifically includes the following steps:
Step S421: performing Fourier transform on the preprocessed data set to obtain frequency domain data, frequency components and signal strength of the frequency components in the frequency domain data of the preprocessed data set;
Step S422: extracting amplitude spectrums of different frequency components according to signal intensity of the different frequency components in frequency domain data;
step S423: extracting amplitude spectrum characteristics, extracting peak values in the amplitude spectrum, and mapping the peak values to corresponding frequency components;
Step S424: mapping the amplitude spectrum peak value to a corresponding frequency component for performing inverse Fourier transform, and performing feature extraction to obtain periodic features of the preprocessed data set;
step S43: respectively normalizing trend characteristics and periodic characteristics of the preprocessed data set;
Step S44: performing linear regression model training by using the trend characteristics and the periodic characteristics of the standardized data set to obtain a trained linear regression model;
step S45: and predicting the temperature, the humidity and the gas concentration of the workshop by using a trained linear regression model to obtain predicted values of the temperature, the humidity and the gas concentration of the workshop, and sending out an early warning alarm when the predicted values exceed safety thresholds of the temperature, the humidity and the gas concentration of the workshop.
Further, step S5 specifically includes the following steps:
step S51: extracting human skeleton key points and position information and confidence information of the human skeleton key points from human behavior images by using a CNN model;
step S52: processing the position information and the confidence information of key points of the human skeleton by using a GCN model to obtain a space feature vector of the human skeleton;
further, step S52 specifically includes the following steps:
Step S521: creating nodes, namely defining each human skeleton key point as a node of the graph, and taking the position information and the confidence information of the human skeleton key points as feature vectors of the nodes;
step S522: constructing edges, determining connection relations between nodes according to skeleton structures, constructing an adjacency matrix, constructing a two-dimensional adjacency matrix according to the connection relations between the nodes, and defining position elements of the adjacency matrix according to the connection relations between the nodes;
step S523: judging the symmetry of the adjacent matrix according to the symmetry of the human body structure, judging whether the elements at the (i, j) and (j, i) positions of the adjacent matrix are equal, if not, removing the error data and reconstructing the adjacent matrix to obtain a final edge;
Step S524: combining the nodes and the edges to form a graph structure of the whole human body;
Step S525: initializing GCN layers, and carrying out convolution operation on the feature vector of the node and the feature vector of the neighbor node by each GCN layer to obtain an updated feature vector;
Step S526: initializing GCN layers, updating the feature vector of the node layer by layer, taking the output of each layer as the input of the next layer until reaching the last GCN layer, and obtaining the feature vector of the fusion polymerization context information;
Step S527: carrying out average pooling operation on the feature vectors fused with the polymerization context information to obtain a space feature vector with global features of each node;
step S53: and performing time-based feature extraction on the spatial feature vector of the human skeleton by using the LSTM model to obtain the spatial and temporal feature vector of the human skeleton.
Further, step S6 specifically includes the following steps:
Step S61: defining the set of all the space-time feature vectors of human bones as And randomly screening partial vectors therefrom to define principal component subsets；
Step S62: setting the initial learning rate asSetting a learning rate threshold;
Step S63: principal component subset Updating, setting an updating round number threshold value, and carrying out updating on the main component subsetUpdating to obtain final principal component subset；
Further, step S63 specifically includes the following steps:
Step S631: projection using the principal component subset to obtain a residual of the principal component subset And transpose the matrix of the current principal component subset to obtain a transposed matrix of the principal component subsetWhereinRepresented as update principal component subsetA wheel;
Step S632: updating a matrix, and updating a transposed matrix of the principal component by using residual calculation, wherein the formula is as follows:
；
In the method, in the process of the invention, Denoted as the firstThe transpose matrix of the principal component subset of the round update,Representing updated firstThe transpose matrix of the principal component subset of the round update,For the initial rate of learning to be the same,Is the firstThe residuals of the principal component subsets of the round update,Is the residual errorIs a transposed matrix of (a);
Step S633: the principal component subsets are superimposed to form the first The product of the principal component subset of the wheel update and the learning rate is the same as the firstAdding the principal component subsets of the round updates to obtain the firstThe principal component subset for +1 round of updating is given by:
In the method, in the process of the invention, Denoted as the firstA principal component subset of the wheel updates;
step S634, dynamically adjusting the learning rate according to the change amounts of the updated principal component subset and the updated principal component subset before updating;
step S635 of Principal component subset for wheel updateNormalizing;
Step S636, repeating steps S631 to S635 until the number of update rounds or learning rate reaches the set threshold value to obtain the final principal component subset ；
Step S64, data conversion, which is to collect all space-time characteristic vectors of human bonesProjection onto final principal component subsetAnd obtaining the space-time characteristic vector of the human skeleton after dimension reduction.
The invention provides a workshop safety supervision system based on internet of things perception, which comprises a sensor and monitoring module, a data acquisition module, a data preprocessing module, a safety judgment module, a user operation interface module and a safety early warning module;
The sensor and monitoring module comprises a high-definition monitoring camera, a temperature sensor, a humidity sensor and a gas concentration sensor, and is used for monitoring the temperature of a workshop, the humidity of the workshop, the gas concentration of the workshop and human behavior images in the workshop;
the data acquisition module acquires the workshop temperature, the workshop humidity, the workshop gas concentration and the human behavior image in the workshop, which are acquired by the sensor and the monitoring module, and sends all data to the data preprocessing module;
The data preprocessing module alarms on the missing value, the abnormal value and the outlier of the workshop temperature, the workshop humidity and the workshop gas concentration, judges that the data are real potential safety hazard data or error data, removes the error data, and simultaneously performs preprocessing of uniformly fixing the size and reducing noise on human behavior images in the workshop;
The safety judgment module is used for carrying out safety prediction judgment on the workshop temperature, the workshop humidity, the workshop gas concentration and the human behavior in the workshop, sending an early warning signal to the safety early warning module and sending early warning information to the user operation interface module;
The user operation interface module is responsible for setting safety thresholds of workshop temperature, workshop humidity and workshop gas concentration, and simultaneously receiving early warning information from the safety judgment module and storing the information;
the safety early warning module is responsible for receiving the early warning signal from the safety judgment module and sending out early warning alarm.
The beneficial results obtained by adopting the scheme of the invention are as follows:
(1) Aiming at the problem that the existing workshop safety supervision system cannot safely predict the workshop environment according to the change trend of the workshop environment data, the invention uses a linear regression prediction model based on first derivative and frequency domain conversion to safely predict the workshop environment, judges, processes and alarms the missing value, abnormal value and outlier of the workshop environment data, reduces the influence of the missing value, abnormal value and outlier of the workshop environment data on a prediction model, and increases the safety coefficient;
(2) Aiming at the problems that the robustness of a human behavior prediction model in the existing supervision system is poor, the prediction speed is low, and the timeliness and accuracy of workshop human behavior safety supervision cannot be met, the invention uses a GCN model and an LSTM model to extract the spatial feature vector of human bones, and uses a self-adaptive principal component analysis algorithm based on self-adaptive learning rate optimization to reduce the dimension of the spatial feature vector of the human bones, thereby increasing the robustness, the interpretability and the data processing efficiency of the model, and finally uses a softmax function to classify and predict the spatial feature vector of the human bones after dimension reduction, so that the human behavior classification prediction model can be better applied to the workshop safety supervision system, and the safety coefficient and the performance of workshop safety supervision are increased.
Drawings
FIG. 1 is a schematic flow chart of a workshop safety supervision method based on the Internet of things perception;
fig. 2 is a flow chart of step S4;
fig. 3 is a flow chart of step S42;
fig. 4 is a flow chart of step S5;
Fig. 5 is a flow chart of step S52;
FIG. 6 is a flow chart of step S6;
fig. 7 is a flow chart of step S63;
FIG. 8 is a schematic diagram of a workshop safety supervision system based on Internet of things perception;
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all embodiments of the invention; all other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, the workshop safety supervision method based on internet of things perception provided by the invention specifically includes the following steps:
Step S1: safety threshold management, namely setting safety thresholds of workshop temperature, workshop humidity and workshop gas concentration;
Step S2: workshop data acquisition, namely acquiring workshop temperature, workshop humidity, workshop gas concentration and human behavior images in a workshop in real time by using a temperature sensor, a humidity sensor, a gas concentration sensor and a high-definition camera;
Step S3: data preprocessing, namely alarming missing values, abnormal values and outliers of workshop temperature, workshop humidity and workshop gas concentration, judging whether the data are real potential safety hazard data or error data, continuously sending out the alarm until the data return to normal if the data are the real potential safety hazard data, removing the error data to obtain a preprocessed data set, adjusting the size of a collected human behavior image in the workshop to be uniform and fixed, and carrying out noise reduction processing on the image to obtain the preprocessed human behavior image set;
Step S4: extracting data set characteristics and predicting data safety, extracting trend characteristics of the preprocessed data set by using a first derivative, extracting periodic characteristics of the preprocessed data set by using a frequency domain analysis method, and carrying out safety prediction and alarm on workshop temperature, workshop humidity and workshop gas concentration according to the trend characteristics and the periodic characteristics, sending out early warning signals and storing early warning information;
Step S5: extracting features of the image set, and extracting features of the human behavior image set by using a human feature extraction model to obtain space-time feature vectors of human bones;
step S6: feature dimension reduction, namely performing dimension reduction on the space-time feature vector of the human skeleton by using an adaptive principal component analysis algorithm based on adaptive learning rate optimization to obtain the space-time feature vector of the human skeleton after dimension reduction;
Step S7: classifying behaviors, namely classifying and predicting space-time feature vectors of the human bones after dimension reduction by using a softmax function to obtain human behavior classification prediction results;
Step S8: judging human behaviors, judging whether potential safety hazards exist in human behaviors according to human behavior classification prediction results, and if the potential safety hazards exist, sending out early warning signals and storing early warning information;
step S9: and (3) evaluating and optimizing the model, evaluating the human body characteristic extraction model and the self-adaptive principal component analysis algorithm based on self-adaptive learning rate optimization, and optimizing and improving the human body characteristic extraction model and the self-adaptive principal component analysis algorithm based on the self-adaptive learning rate optimization according to feedback and requirements in practical application.
In a second embodiment, referring to fig. 2 and 3, the present embodiment is based on the above embodiment, and step S4 specifically includes the following steps:
step S41, extracting trend characteristics of the preprocessed data set by using the first derivative;
Step S42, extracting periodic characteristics of the preprocessed data set by using a frequency domain analysis method based on Fourier transform, and specifically comprises the following steps:
Step S421: performing Fourier transform on the preprocessed data set to obtain frequency domain data, frequency components and signal strength of the frequency components in the frequency domain data of the preprocessed data set;
Step S422: extracting amplitude spectrums of different frequency components according to signal intensity of the different frequency components in frequency domain data;
step S423: extracting amplitude spectrum characteristics, extracting peak values in the amplitude spectrum, and mapping the peak values to corresponding frequency components;
Step S424: mapping the amplitude spectrum peak value to a corresponding frequency component for performing inverse Fourier transform, and performing feature extraction to obtain periodic features of the preprocessed data set;
step S43: respectively normalizing trend characteristics and periodic characteristics of the preprocessed data set;
Step S44: performing linear regression model training by using the trend characteristics and the periodic characteristics of the standardized data set to obtain a trained linear regression model;
step S45: and predicting the temperature, the humidity and the gas concentration of the workshop by using a trained linear regression model to obtain predicted values of the temperature, the humidity and the gas concentration of the workshop, and sending out an early warning alarm when the predicted values exceed safety thresholds of the temperature, the humidity and the gas concentration of the workshop.
By executing the operation, the method and the system aim at the problem that the existing workshop safety supervision system cannot safely predict the workshop environment according to the change trend of the workshop environment data.
Referring to fig. 4 and 5, the third embodiment is based on the above embodiment, and step S5 specifically includes the following steps:
Step S51: extracting human skeleton key points and position information and confidence information of the human skeleton key points from a human behavior image by using a basic CNN model;
Step S52: the GCN model is used for processing the position information and the confidence information of the key points of the human skeleton to obtain the space feature vector of the human skeleton, and the method specifically comprises the following steps:
Step S521: creating nodes, namely defining each human skeleton key point as a node of the graph, and taking the position information and the confidence information of the human skeleton key points as feature vectors of the nodes;
step S522: constructing edges, determining connection relations between nodes according to skeleton structures, constructing an adjacency matrix, constructing a two-dimensional adjacency matrix according to the connection relations between the nodes, and defining position elements of the adjacency matrix according to the connection relations between the nodes;
step S523: judging the symmetry of the adjacent matrix according to the symmetry of the human body structure, judging whether the elements at the (i, j) and (j, i) positions of the adjacent matrix are equal, if not, removing the error data and reconstructing the adjacent matrix to obtain a final edge;
Step S524: combining the nodes and the edges to form a graph structure of the whole human body;
Step S525: initializing the GCN layers to be 50 layers, and carrying out convolution operation on the feature vector of the node and the feature vector of the neighbor node by each GCN layer to obtain an updated feature vector;
Step S526: initializing GCN layers, updating the feature vector of the node layer by layer, taking the output of each layer as the input of the next layer until reaching the last GCN layer, and obtaining the feature vector of the fusion polymerization context information;
Step S527: carrying out average pooling operation on the feature vectors fused with the polymerization context information to obtain a space feature vector with global features of each node;
step S53: and performing time-based feature extraction on the spatial feature vector of the human skeleton by using the LSTM model to obtain the spatial and temporal feature vector of the human skeleton.
In a fourth embodiment, referring to fig. 6 and 7, the present embodiment is based on the above embodiment, and step S6 specifically includes the following steps:
Step S61: defining the set of all the space-time feature vectors of human bones as And randomly screening the quarter vectors from the above as principal component subsets；
Step S62: setting the initial learning rate asSetting a learning rate threshold;
Step S63: principal component subset Updating, setting the threshold value of the number of updating rounds as 10, and dividing the main components into sub-groupsUpdating to obtain final principal component subsetThe method specifically comprises the following steps:
Step S631: projection using the principal component subset to obtain a residual of the principal component subset And transpose the matrix of the current principal component subset to obtain a transposed matrix of the principal component subsetWhereinRepresented as update principal component subsetA wheel;
Step S632: updating a matrix, and updating a transposed matrix of the principal component by using residual calculation, wherein the formula is as follows:
；
In the method, in the process of the invention, Denoted as the firstThe transpose matrix of the principal component subset of the round update,Representing updated firstThe transpose matrix of the principal component subset of the round update,For the initial rate of learning to be the same,Is the firstThe residuals of the principal component subsets of the round update,Is the residual errorIs a transposed matrix of (a);
Step S633: the principal component subsets are superimposed to form the first The product of the principal component subset of the wheel update and the learning rate is the same as the firstAdding the principal component subsets of the round updates to obtain the firstThe principal component subset for +1 round of updating is given by:
In the method, in the process of the invention, Denoted as the firstA principal component subset of the wheel updates;
step S634, dynamically adjusting the learning rate according to the change amounts of the updated principal component subset and the updated principal component subset before updating;
step S635 of Principal component subset for wheel updateNormalizing;
Step S636, repeating steps S631 to S635 until the number of update rounds or learning rate reaches the set threshold value to obtain the final principal component subset ；
Step S64, data conversion, which is to collect all space-time characteristic vectors of human bonesProjection onto final principal component subsetAnd obtaining the space-time characteristic vector of the human skeleton after dimension reduction.
By executing the operation, aiming at the problems that the robustness of the human behavior prediction model in the existing supervision system is poor, the prediction speed is low, and the timeliness and accuracy of workshop human behavior safety supervision cannot be met, the invention uses the GCN model and the LSTM model to extract the space feature vector of the human skeleton, uses the self-adaptive principal component analysis algorithm based on self-adaptive learning rate optimization to reduce the dimension of the space feature vector of the human skeleton, increases the robustness, the interpretability and the data processing efficiency of the model, and finally uses the softmax function to classify and predict the space feature vector of the human skeleton after dimension reduction, so that the human behavior classification prediction model can be better applied to the workshop safety supervision system, and the safety coefficient and the performance of the workshop safety supervision are increased.
Fifth embodiment, referring to fig. 8, the workshop safety supervision system based on internet of things perception provided by the invention comprises a sensor and monitoring module, a data acquisition module, a data preprocessing module, a safety judgment module, a user operation interface module and a safety early warning module;
The sensor and monitoring module comprises a high-definition monitoring camera, a temperature sensor, a humidity sensor and a gas concentration sensor, and is used for monitoring the temperature of a workshop, the humidity of the workshop, the gas concentration of the workshop and human behavior images in the workshop;
the data acquisition module acquires the workshop temperature, the workshop humidity, the workshop gas concentration and the human behavior image in the workshop, which are acquired by the sensor and the monitoring module, and sends all data to the data preprocessing module;
The data preprocessing module alarms on the missing value, the abnormal value and the outlier of the workshop temperature, the workshop humidity and the workshop gas concentration, judges that the data are real potential safety hazard data or error data, removes the error data, and simultaneously performs preprocessing of uniformly fixing the size and reducing noise on human behavior images in the workshop;
The safety judgment module is used for carrying out safety prediction judgment on the workshop temperature, the workshop humidity, the workshop gas concentration and the human behavior in the workshop, sending an early warning signal to the safety early warning module and sending early warning information to the user operation interface module;
The user operation interface module is responsible for setting safety thresholds of workshop temperature, workshop humidity and workshop gas concentration, and simultaneously receiving early warning information from the safety judgment module and storing the information;
the safety early warning module is responsible for receiving the early warning signal from the safety judgment module and sending out early warning alarm.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
The invention and its embodiments have been described above with no limitation, and the actual construction is not limited to the embodiments of the invention as shown in the drawings. In summary, if one of ordinary skill in the art is informed by this disclosure, a structural manner and an embodiment similar to the technical solution should not be creatively devised without departing from the gist of the present invention.
Claims (8)
1. The workshop safety supervision method based on the internet of things perception is characterized by comprising the following steps of:
Step S1: safety threshold management, namely setting safety thresholds of workshop temperature, workshop humidity and workshop gas concentration;
Step S2: workshop data acquisition, namely acquiring workshop temperature, workshop humidity, workshop gas concentration and human behavior images in a workshop in real time by using a temperature sensor, a humidity sensor, a gas concentration sensor and a high-definition camera;
Step S3: data preprocessing, namely alarming missing values, abnormal values and outliers of workshop temperature, workshop humidity and workshop gas concentration, judging whether the data are real potential safety hazard data or error data, continuously sending out the alarm until the data return to normal if the data are the real potential safety hazard data, removing the error data to obtain a preprocessed data set, adjusting the size of a collected human behavior image in the workshop to be uniform and fixed, and carrying out noise reduction processing on the image to obtain the preprocessed human behavior image set;
Step S4: extracting data set characteristics and predicting data safety, extracting trend characteristics of the preprocessed data set by using a first derivative, extracting periodic characteristics of the preprocessed data set by using a frequency domain analysis method, and carrying out safety prediction and alarm on workshop temperature, workshop humidity and workshop gas concentration according to the trend characteristics and the periodic characteristics, sending out early warning signals and storing early warning information;
Step S5: extracting features of the image set, and extracting features of the human behavior image set by using a human feature extraction model to obtain space-time feature vectors of human bones;
step S6: feature dimension reduction, namely performing dimension reduction on the space-time feature vector of the human skeleton by using an adaptive principal component analysis algorithm based on adaptive learning rate optimization to obtain the space-time feature vector of the human skeleton after dimension reduction;
Step S7: classifying behaviors, namely classifying and predicting space-time feature vectors of the human bones after dimension reduction by using a softmax function to obtain human behavior classification prediction results;
Step S8: judging human behaviors, judging whether potential safety hazards exist in human behaviors according to human behavior classification prediction results, and if the potential safety hazards exist, sending out early warning signals and storing early warning information;
step S9: and (3) evaluating and optimizing the model, evaluating the human body characteristic extraction model and the self-adaptive principal component analysis algorithm based on self-adaptive learning rate optimization, and optimizing and improving the human body characteristic extraction model and the self-adaptive principal component analysis algorithm based on the self-adaptive learning rate optimization according to feedback and requirements in practical application.
2. The method for supervising workshop safety based on internet of things perception according to claim 1, wherein the method comprises the following steps: step S4, specifically comprising the following steps:
step S41, extracting trend characteristics of the preprocessed data set by using the first derivative;
Step S42, extracting periodic characteristics of the preprocessed data set by using a frequency domain analysis method based on Fourier transform;
step S43: respectively normalizing trend characteristics and periodic characteristics of the preprocessed data set;
Step S44: performing linear regression model training by using the trend characteristics and the periodic characteristics of the standardized data set to obtain a trained linear regression model;
step S45: and predicting the temperature, the humidity and the gas concentration of the workshop by using a trained linear regression model to obtain predicted values of the temperature, the humidity and the gas concentration of the workshop, and sending out an early warning alarm when the predicted values exceed safety thresholds of the temperature, the humidity and the gas concentration of the workshop.
3. The method for supervising workshop safety based on internet of things perception according to claim 1, wherein the method comprises the following steps: step S5, specifically comprising the following steps:
step S51: extracting human skeleton key points and position information and confidence information of the human skeleton key points from human behavior images by using a CNN model;
step S52: processing the position information and the confidence information of key points of the human skeleton by using a GCN model to obtain a space feature vector of the human skeleton;
step S53: and performing time-based feature extraction on the spatial feature vector of the human skeleton by using the LSTM model to obtain the spatial and temporal feature vector of the human skeleton.
4. The method for supervising workshop safety based on internet of things perception according to claim 1, wherein the method comprises the following steps: step S6, specifically comprising the following steps:
Step S61: defining the set of all the space-time feature vectors of human bones as And randomly screening partial vectors therefrom is defined as principal component subset/>；
Step S62: setting the initial learning rate asSetting a learning rate threshold;
Step S63: principal component subset Updating, setting an updating round number threshold value, and carrying out/>, on the main component subsetUpdating to obtain the final principal component subset/>；
Step S64, data conversion, which is to collect all space-time characteristic vectors of human bonesProjection to final principal component subset/>And obtaining the space-time characteristic vector of the human skeleton after dimension reduction.
5. The method for supervising workshop safety based on internet of things perception according to claim 2, wherein the method comprises the following steps: step S42, specifically comprising the following steps:
Step S421: performing Fourier transform on the preprocessed data set to obtain frequency domain data, frequency components and signal strength of the frequency components in the frequency domain data of the preprocessed data set;
Step S422: extracting amplitude spectrums of different frequency components according to signal intensity of the different frequency components in frequency domain data;
step S423: extracting amplitude spectrum characteristics, extracting peak values in the amplitude spectrum, and mapping the peak values to corresponding frequency components;
Step S424: mapping the amplitude spectrum peak value to a corresponding frequency component for performing inverse Fourier transform, and performing feature extraction to obtain the periodic feature of the preprocessed data set.
6. The internet of things-based workshop safety supervision method according to claim 3, wherein: step S52 specifically includes the following steps:
Step S521: creating nodes, namely defining each human skeleton key point as a node of the graph, and taking the position information and the confidence information of the human skeleton key points as feature vectors of the nodes;
step S522: constructing edges, determining connection relations between nodes according to skeleton structures, constructing an adjacency matrix, constructing a two-dimensional adjacency matrix according to the connection relations between the nodes, and defining position elements of the adjacency matrix according to the connection relations between the nodes;
step S523: judging the symmetry of the adjacent matrix according to the symmetry of the human body structure, judging whether the elements at the (i, j) and (j, i) positions of the adjacent matrix are equal, if not, removing the error data and reconstructing the adjacent matrix to obtain a final edge;
Step S524: combining the nodes and the edges to form a graph structure of the whole human body;
Step S525: initializing GCN layers, and carrying out convolution operation on the feature vector of the node and the feature vector of the neighbor node by each GCN layer to obtain an updated feature vector;
Step S526: initializing GCN layers, updating the feature vector of the node layer by layer, taking the output of each layer as the input of the next layer until reaching the last GCN layer, and obtaining the feature vector of the fusion polymerization context information;
Step S527: and carrying out average pooling operation on the feature vectors fused with the polymerization context information to obtain the space feature vector with global features of each node.
7. The method for supervising the workshop safety based on the internet of things perception according to claim 4, wherein the method comprises the following steps: step S63, specifically includes the following steps:
Step S631: projection using the principal component subset to obtain a residual of the principal component subset And transpose the matrix of the current principal component subset to obtain a transposed matrix/>, of the principal component subsetWherein/>Expressed as update principal component subset/>A wheel;
Step S632: updating a matrix, and updating a transposed matrix of the principal component by using residual calculation, wherein the formula is as follows:
；
In the method, in the process of the invention, Expressed as/>Transposed matrix of principal component subset for round update,/>Represents updated/>Transposed matrix of principal component subset for round update,/>For initial learning rate,/>For/>Residual of principal component subset of round update,/>Is residual/>Is a transposed matrix of (a);
Step S633: the principal component subsets are superimposed to form the first The product of the principal component subset of the round update and the learning rate is the same as the/>Principal component subset addition of the round update to get the/>The principal component subset for +1 round of updating is given by:
；
In the method, in the process of the invention, Expressed as/>A principal component subset of the wheel updates;
step S634, dynamically adjusting the learning rate according to the change amounts of the updated principal component subset and the updated principal component subset before updating;
step S635 of Principal component subset of round updates/>Normalizing;
Step S636, repeating steps S631 to S635 until the number of update rounds or learning rate reaches the set threshold value to obtain the final principal component subset 。
8. The workshop safety supervision system based on the internet of things perception is used for realizing the workshop safety supervision method based on the internet of things perception according to any one of claims 1 to 7, and is characterized by comprising a sensor and monitoring module, a data acquisition module, a data preprocessing module, a safety judging module, a user operation interface module and a safety early warning module;
The sensor and monitoring module comprises a high-definition monitoring camera, a temperature sensor, a humidity sensor and a gas concentration sensor, and is used for monitoring the temperature of a workshop, the humidity of the workshop, the gas concentration of the workshop and human behavior images in the workshop;
the data acquisition module acquires the workshop temperature, the workshop humidity, the workshop gas concentration and the human behavior image in the workshop, which are acquired by the sensor and the monitoring module, and sends all data to the data preprocessing module;
The data preprocessing module alarms on the missing value, the abnormal value and the outlier of the workshop temperature, the workshop humidity and the workshop gas concentration, judges that the data are real potential safety hazard data or error data, removes the error data, and simultaneously performs preprocessing of uniformly fixing the size and reducing noise on human behavior images in the workshop;
The safety judgment module is used for carrying out safety prediction judgment on the workshop temperature, the workshop humidity, the workshop gas concentration and the human behavior in the workshop, sending an early warning signal to the safety early warning module and sending early warning information to the user operation interface module;
The user operation interface module is responsible for setting safety thresholds of workshop temperature, workshop humidity and workshop gas concentration, and simultaneously receiving early warning information from the safety judgment module and storing the information;
the safety early warning module is responsible for receiving the early warning signal from the safety judgment module and sending out early warning alarm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410277674.XA CN117876972B (en) | 2024-03-12 | 2024-03-12 | Workshop safety supervision method and system based on internet of things perception |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410277674.XA CN117876972B (en) | 2024-03-12 | 2024-03-12 | Workshop safety supervision method and system based on internet of things perception |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117876972A CN117876972A (en) | 2024-04-12 |
| CN117876972B true CN117876972B (en) | 2024-05-07 |
Family
ID=90579728
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410277674.XA Active CN117876972B (en) | 2024-03-12 | 2024-03-12 | Workshop safety supervision method and system based on internet of things perception |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117876972B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10605470B1 (en) * | 2016-03-08 | 2020-03-31 | BrainofT Inc. | Controlling connected devices using an optimization function |
| CN112005198A (en) * | 2018-01-25 | 2020-11-27 | 脸谱科技有限责任公司 | Hand state reconstruction based on multiple inputs |
| JP2022043974A (en) * | 2020-09-04 | 2022-03-16 | 株式会社日立製作所 | Behavior recognition apparatus, learning apparatus, and behavior recognition method |
| CN115268287A (en) * | 2022-07-20 | 2022-11-01 | 清华大学 | Intelligent home comprehensive experiment system and data processing method |
| CN116311374A (en) * | 2023-03-27 | 2023-06-23 | 淮阴工学院 | Method and system for identifying and early warning abnormal behaviors of workers in chemical plant |
-
2024
- 2024-03-12 CN CN202410277674.XA patent/CN117876972B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10605470B1 (en) * | 2016-03-08 | 2020-03-31 | BrainofT Inc. | Controlling connected devices using an optimization function |
| CN112005198A (en) * | 2018-01-25 | 2020-11-27 | 脸谱科技有限责任公司 | Hand state reconstruction based on multiple inputs |
| JP2022043974A (en) * | 2020-09-04 | 2022-03-16 | 株式会社日立製作所 | Behavior recognition apparatus, learning apparatus, and behavior recognition method |
| CN115268287A (en) * | 2022-07-20 | 2022-11-01 | 清华大学 | Intelligent home comprehensive experiment system and data processing method |
| CN116311374A (en) * | 2023-03-27 | 2023-06-23 | 淮阴工学院 | Method and system for identifying and early warning abnormal behaviors of workers in chemical plant |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117876972A (en) | 2024-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6759474B2 (en) | Vessel automatic tracking methods and systems based on deep learning networks and average shifts | |
| Surya | Risk analysis model that uses machine learning to predict the likelihood of a fire occurring at a given property | |
| CN112287816B (en) | Dangerous work area accident automatic detection and alarm method based on deep learning | |
| CN107872644B (en) | Video monitoring method and device | |
| US10914608B2 (en) | Data analytic engine towards the self-management of complex physical systems | |
| US9245235B2 (en) | Integrated approach to model time series dynamics in complex physical systems | |
| CN109714324B (en) | User network abnormal behavior discovery method and system based on machine learning algorithm | |
| CN101315701B (en) | Movement destination image partition method | |
| US11631306B2 (en) | Methods and system for monitoring an environment | |
| CN102469302B (en) | Background model learning system for lighting change adaptation utilized for video surveillance | |
| CN107463898A (en) | The stage performance abnormal behavior monitoring method of view-based access control model sensing network | |
| CN104933542B (en) | A kind of logistic storage monitoring method based on computer vision | |
| CN109241870B (en) | Coal mine underground personnel identity identification method based on gait identification | |
| CN115223049A (en) | Knowledge distillation and quantification technology for large model compression of electric power scene edge calculation | |
| CN117876972B (en) | Workshop safety supervision method and system based on internet of things perception | |
| Subramanian et al. | Fusion for the detection of dependent signals using multivariate copulas | |
| CN117318981A (en) | Network attack prediction method, system, electronic equipment and storage medium | |
| Bhushan et al. | Incremental principal component analysis based outlier detection methods for spatiotemporal data streams | |
| CN116738302A (en) | Method for detecting anomalies in time-series data generated by an infrastructure device in a network | |
| CN116152735A (en) | Security monitoring dangerous source identification system and method based on edge cloud computing | |
| CN113537240B (en) | Deformation zone intelligent extraction method and system based on radar sequence image | |
| Akyol | A comprehensive comparison study of traditional classifiers and deep neural networks for forest fire detection | |
| CN114373142A (en) | Pedestrian falling detection method based on deep learning | |
| CN106791647A (en) | A kind of hydroelectric power plant's condition monitoring system and method based on video intelligent identification | |
| Nazir et al. | Low complexity object detection with background subtraction for intelligent remote monitoring |
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 |