CN112464897A - Electric power operator screening method and device - Google Patents
Electric power operator screening method and device Download PDFInfo
- Publication number
- CN112464897A CN112464897A CN202011478342.6A CN202011478342A CN112464897A CN 112464897 A CN112464897 A CN 112464897A CN 202011478342 A CN202011478342 A CN 202011478342A CN 112464897 A CN112464897 A CN 112464897A
- Authority
- CN
- China
- Prior art keywords
- image
- face
- similarity
- recognized
- face feature
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a device for discriminating electric power operators, wherein the method comprises the following steps: receiving a face image to be recognized; carrying out image preprocessing on a face image to be recognized to generate a plurality of image matrixes to be recognized with different scales; respectively inputting a plurality of image matrixes to be recognized into a target human face feature extraction model, and outputting human face feature data corresponding to each image matrix to be recognized; the target human face feature extraction model is generated through a preset model training process, and human face feature data comprise human face feature similarity; and determining whether the personnel corresponding to the face image to be recognized has the work qualification or not based on the comparison result of whether the similarity of each face feature reaches the preset threshold value or not. The accuracy of personnel selection under the complex environment is effectively improved, and the potential safety hazard is prevented.
Description
Technical Field
The invention relates to the technical field of personnel identification, in particular to a method and a device for discriminating electric power operating personnel.
Background
With the rapid development of social economy and the continuous improvement of technological level, in recent years, the electric power industry in China is rapidly developed, and in the process of electric power production, strict requirements are imposed on operating personnel, and whether the on-site operating personnel have work qualification needs to be examined.
In the real-time operation process, whether qualified personnel come in or go out of the operation site needs to be monitored, if the unqualified personnel are found, an alarm is given in time, and images of the personnel are recorded. However, in the process of power generation, field operators work in the field, the field illumination condition changes infrequently, the working environment varies widely, the working position changes constantly, and the human body posture changes constantly, which puts severe requirements on face detection and recognition algorithms, and especially when the face of an operator occurs: when the problems of large-angle change, expression change, large-area shielding, large-range size change and the like exist, the recognition rate of the existing face video monitoring technology can be greatly reduced, the accuracy of discrimination of personnel is reduced, and potential safety hazards can be caused.
Disclosure of Invention
The invention provides a method and a device for discriminating electric power operating personnel, which solve the technical problems that the discrimination accuracy of personnel is reduced and potential safety hazards are possibly caused due to the fact that the recognition rate of the existing face video monitoring technology is reduced due to the complex environment.
The invention provides a method for discriminating electric power operators, which comprises the following steps:
receiving a face image to be recognized;
carrying out image preprocessing on the face image to be recognized to generate a plurality of image matrixes to be recognized with different scales;
respectively inputting the image matrixes to be recognized into a target face feature extraction model, and outputting face feature data corresponding to each image matrix to be recognized; the target human face feature extraction model is generated through a preset model training process, and the human face feature data comprise human face feature similarity;
and determining whether the personnel corresponding to the face image to be recognized has the work qualification or not based on the comparison result of whether the face feature similarity reaches the preset threshold or not.
Optionally, the step of performing image preprocessing on the face image to be recognized to generate a plurality of image matrixes to be recognized with different scales includes:
cutting the face image to be recognized to obtain a plurality of first image matrixes with different scales;
respectively executing image turning operation on the plurality of first image matrixes according to preset requirements to obtain a plurality of second image matrixes;
and respectively executing image normalization operation and image normalization operation on the second image matrixes to generate a plurality of to-be-identified image matrixes with different scales.
Optionally, the model training process comprises:
respectively acquiring human face images in various historical electric power operation scenes;
carrying out image preprocessing on the face image to generate a plurality of historical face image matrixes with different scales;
inputting a plurality of historical face image matrixes into a preset initial face feature extraction model to obtain a plurality of historical face image similarities;
determining a face comprehensive loss function based on the similarity of a plurality of historical face images and a preset weight;
and training the initial face feature extraction model by adopting the face comprehensive loss function to obtain a target face feature extraction model.
Optionally, the preset weight includes a first weight and a second weight, and the step of determining the face synthetic loss function based on the similarity of the plurality of historical face images and the preset weight includes:
selecting a first similarity and a second similarity from the plurality of historical face image similarities;
determining a first class similarity accumulated value of the historical face image matrix based on the first similarity and the first weight;
determining a second type similarity accumulated value of the historical face image matrix based on the second similarity and the second weight;
and determining the face comprehensive loss function according to the first type of similarity accumulated value and the second type of similarity accumulated value.
Optionally, the step of determining whether the person corresponding to the face image to be recognized has a work qualification based on the comparison result of whether each face feature similarity reaches a preset threshold includes:
if any one of the face feature similarities reaches a preset threshold value, determining that the personnel corresponding to the face image to be recognized has the work qualification;
and if the similarity of each face feature does not reach the preset threshold, determining that the personnel corresponding to the face image to be recognized do not have the work qualification.
The invention also provides a device for discriminating the electric power operating personnel, which comprises:
the image receiving module is used for receiving a face image to be recognized;
the image preprocessing module is used for preprocessing the face image to be recognized to generate a plurality of image matrixes to be recognized with different scales;
the human face feature data output module is used for respectively inputting the image matrixes to be recognized into a target human face feature extraction model and outputting human face feature data corresponding to each image matrix to be recognized; the target human face feature extraction model is generated through a preset model training module, and the human face feature data comprise human face feature similarity;
and the work qualification judging module is used for determining whether the personnel corresponding to the face image to be recognized has work qualification or not based on the comparison result of whether the face feature similarity reaches the preset threshold value or not.
Optionally, the image preprocessing module includes:
the image cutting submodule is used for cutting the face image to be recognized to obtain a plurality of first image matrixes with different scales;
the image turning sub-module is used for respectively executing image turning operation on the plurality of first image matrixes according to preset requirements to obtain a plurality of second image matrixes;
and the image normalization submodule is used for respectively performing image normalization operation and image normalization operation on the second image matrixes to generate a plurality of to-be-identified image matrixes with different scales.
Optionally, the model training module comprises:
the historical face image acquisition submodule is used for respectively acquiring face images in various historical electric power operation scenes;
the image preprocessing submodule is used for preprocessing the face image to generate a plurality of historical face image matrixes with different scales;
the historical face image similarity determining submodule is used for inputting a plurality of historical face image matrixes into a preset initial face feature extraction model to obtain a plurality of historical face image similarities;
a comprehensive loss function determining submodule for determining a comprehensive loss function of the face based on the similarity of a plurality of historical face images and a preset weight;
and the model training submodule is used for training the initial face feature extraction model by adopting the face comprehensive loss function to obtain a target face feature extraction model.
Optionally, the preset weights include a first weight and a second weight, and the comprehensive loss function determining sub-module includes:
the similarity obtaining unit is used for selecting a first similarity and a second similarity from a plurality of historical face image similarities;
the first accumulation unit is used for determining a first type of similarity accumulated value of the historical human face image matrix based on the first similarity and the first weight;
the second accumulation unit is used for determining a second type of similarity accumulation value of the historical human face image matrix based on the second similarity and the second weight;
and the face comprehensive loss function determining unit is used for determining the face comprehensive loss function according to the first type of similarity accumulated value and the second type of similarity accumulated value.
Optionally, the work qualification determining module includes:
the first judgment submodule is used for determining that the personnel corresponding to the face image to be recognized has the working qualification if any one of the face feature similarity reaches a preset threshold value;
and the second judging submodule is used for determining that the personnel corresponding to the face image to be recognized does not have the working qualification if the face feature similarity does not reach the preset threshold value.
According to the technical scheme, the invention has the following advantages:
the method comprises the steps of receiving a face image to be recognized, carrying out image preprocessing on the face image to be recognized to obtain a plurality of image matrixes to be recognized with different scales, respectively inputting the plurality of image matrixes to be recognized into a target face feature extraction model to output face feature similarity corresponding to the image matrixes to be recognized of each scale, and finally determining whether personnel corresponding to the face image to be recognized have work qualification or not based on a comparison result of whether the face feature similarity is larger than a preset threshold value or not. Therefore, the technical problems that the accuracy of screening personnel is reduced and potential safety hazards are possibly caused due to the fact that the recognition rate of the existing face video monitoring technology is reduced due to the complex environment are solved, the accuracy of screening the personnel in the complex environment is effectively improved, and the potential safety hazards are prevented.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without inventive exercise.
Fig. 1 is a flowchart illustrating steps of a method for screening electric power operators according to an embodiment of the present invention;
fig. 2 is a flowchart illustrating steps of a method for screening electric power operators according to a second embodiment of the present invention;
FIG. 3 is a flowchart illustrating a calculation of a face synthetic loss function according to a second embodiment of the present invention;
fig. 4 is a schematic structural diagram of an electric power operation monitoring platform according to a third embodiment of the present invention;
fig. 5 is a block diagram of a structure of an electric power operator screening apparatus according to a fourth embodiment of the present invention.
Detailed Description
The embodiment of the invention provides a method and a device for screening electric power operating personnel, which are used for solving the technical problems that the identification rate of the existing face video monitoring technology is reduced due to complex environment, so that the accuracy of personnel screening is reduced, and potential safety hazards are possibly caused.
In order to make the objects, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the embodiments described below are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, fig. 1 is a flowchart illustrating steps of a method for screening an electric power operator according to an embodiment of the present invention.
The invention provides a method for discriminating electric power operators, which comprises the following steps:
the face image to be recognized refers to an image in which the face of the operator exists in the power operation scene.
In the embodiment of the invention, corresponding monitoring equipment such as a camera is usually arranged in the power operation scene to ensure the safety of the power operation scene, and in order to determine whether the operator has corresponding work qualification, a target detection algorithm is adopted to acquire a face image to be recognized from a monitoring video stream to determine that the image of the face of the target operator exists.
Optionally, the face image to be recognized may also be periodically acquired at the predetermined period as the acquisition time.
It should be noted that, in particular, the target detection algorithm used for acquiring the face image to be recognized from the video stream is not limited in the embodiment of the present invention.
102, performing image preprocessing on the face image to be recognized to generate a plurality of image matrixes to be recognized with different scales;
after the face image to be recognized is acquired, image preprocessing can be performed on the face image to be recognized to obtain, for example, image matrixes to be recognized under different illumination and different scenes and at different angles, and in order to further improve the detection accuracy, image preprocessing can be performed on the face image to be recognized at different scales to acquire a plurality of image matrixes to be recognized at different scales.
103, respectively inputting a plurality of image matrixes to be recognized into a target human face feature extraction model, and outputting human face feature data corresponding to each image matrix to be recognized;
the target human face feature extraction model is generated through a preset model training process, and the human face feature data comprise human face feature similarity;
in the specific implementation, a target face feature extraction model is generated through a model training process, a plurality of image matrixes to be recognized with different scales are respectively input into the target face feature extraction model, and face feature data corresponding to each image matrix to be recognized are output.
It is worth mentioning that the face feature data includes, but is not limited to, the following features: face feature similarity, face image confidence, face frame coordinates and the like.
And 104, determining whether the personnel corresponding to the face image to be recognized has the work qualification or not based on the comparison result of whether the face feature similarity reaches the preset threshold or not.
After the face feature similarity is obtained, the face feature similarity is compared with a preset threshold value, whether the face feature similarity can reach the preset threshold value is judged, and therefore whether the personnel corresponding to the face image to be recognized have the work qualification is determined.
In the embodiment of the invention, the human face image to be recognized is received, the image preprocessing is carried out on the human face image to be recognized to obtain a plurality of image matrixes to be recognized with different scales, the image matrixes to be recognized are respectively input into a target human face feature extraction model to output the human face feature similarity corresponding to the image matrixes to be recognized of each scale, and finally, whether personnel corresponding to the human face image to be recognized have work qualification or not is determined based on a comparison result of whether the human face feature similarity is larger than a preset threshold value or not. Therefore, the technical problems that the accuracy of screening personnel is reduced and potential safety hazards are possibly caused due to the fact that the recognition rate of the existing face video monitoring technology is reduced due to the complex environment are solved, the accuracy of screening the personnel in the complex environment is effectively improved, and the potential safety hazards are prevented.
Referring to fig. 2, fig. 2 is a flowchart illustrating steps of a method for screening electric power operators according to a second embodiment of the present invention.
The invention provides a method for discriminating electric power operators, which comprises the following steps:
in the embodiment of the invention, the received face image to be recognized is cut to obtain the image with the face of the person, and in order to further improve the accuracy of image recognition, the face image of the person can be further extracted from a plurality of different scales to obtain a plurality of first image matrixes with different scales.
after the first image matrixes with different scales are obtained, the picture can be turned left and right or turned up and down through a flip function of opencv, so that a plurality of second image matrixes suitable for picture recognition are output.
And 204, respectively executing image normalization operation and image normalization operation on the second image matrixes to generate a plurality of to-be-identified image matrixes with different scales.
After the second image matrix is obtained, the second image matrix can be subjected to zero-mean value standard processing to realize image standardization operation, then the image standardization operation is carried out on the image matrix after the image standardization, and the image matrix is converted into a floating point type matrix of [ -1, 1] so as to obtain the image matrixes to be identified with different sizes.
The zero mean specification process refers to taking each variable in the image matrix minus their mean.
The image normalization operation refers to subtracting their mean from each variable in the image matrix and dividing by the standard deviation.
the target human face feature extraction model is generated through a preset model training process, and the human face feature data comprise human face feature similarity;
optionally, the model training process may include the following sub-steps S1-S5:
s1, respectively acquiring face images in various historical electric power operation scenes;
s2, carrying out image preprocessing on the face image to generate a plurality of historical face image matrixes with different scales;
s3, inputting the plurality of historical face image matrixes into a preset initial face feature extraction model to obtain a plurality of historical face image similarities;
optionally, the initial face feature extraction model is constructed by a backbone neural network composed of SE-resenext Module, sense Module and the like. And a data enhancement technology and an anchor design strategy are fused to improve the accuracy of face detection under different scales, backgrounds and illumination.
In the embodiment of the invention, the face images in various historical electric power operation scenes, such as the face images in the electric power operation scenes with different illumination or different angles, can be respectively obtained, the face images are subjected to image preprocessing, namely, after the images are subjected to operations such as standardization, normalization and the like, historical face image matrixes with various different scales are generated, and a plurality of historical face image matrixes are input into a preset initial face feature extraction model to obtain a plurality of historical face image similarities.
The specific implementation process of the image preprocessing is the same as that in the step 202-204, and is not described herein again.
S4, determining a face comprehensive loss function based on the similarity of the plurality of historical face images and a preset weight;
further, the preset weights include a first weight and a second weight, and step S4 may include the following sub-steps:
selecting a first similarity and a second similarity from the plurality of historical face image similarities;
determining a first class similarity accumulated value of the historical face image matrix based on the first similarity and the first weight;
determining a second type similarity accumulated value of the historical face image matrix based on the second similarity and the second weight;
and determining the face comprehensive loss function according to the first type of similarity accumulated value and the second type of similarity accumulated value.
As shown in fig. 3, in one example of the present invention, the face synthesis loss function is calculated by the following steps:
the first similarity S1 is applied to a first input of the first multiplier block 301, a first weight α 1 of the first similarity is applied to a first input of the first multiplier block 301,
a second similarity S2 is provided to a first input of the second multiplier module 302, a second weight α 2 of the second similarity is provided to a second input of the second multiplier module 302,
The setpoint module 303 sends the value-1 to the 3 rd input of the second multiplier module 302;
the first multiplier module 301 receives the value obtained by multiplying the first similarity by the weight coefficient thereof to the input end of the first exponent calculation module 304, and the second multiplier module 302 receives the value obtained by multiplying the second similarity by the weight coefficient thereof to the input end of the second exponent calculation module 305, so as to complete the exponent operation of the two types of similarities;
the output of the first index calculation module 304 is connected to the input end of the first accumulation calculation module 306 for performing the accumulation calculation of the picture first similarity full matrix value, and the output of the second index calculation module 305 is connected to the input end of the second accumulation calculation module 307 for performing the accumulation calculation of the picture second similarity full matrix value;
the output of the first accumulation calculation module 306 is connected to the first input terminal of the 3 rd multiplier module 308, and the output of the second accumulation calculation module 307 is connected to the second input terminal of the 3 rd multiplier module 308;
the output of the 3 rd multiplier block is coupled to a first input of the first adder block 309 and the setpoint block 310 sends the value +1 to a second input of the first adder block 309;
the integrated value calculated by the first adder module 309 is sent to the first natural logarithm module 311, and the logarithm operation is performed on the calculated integrated value, so that the face integrated loss function of each graph similarity is calculated.
And S5, training the initial face feature extraction model by adopting the face comprehensive loss function to obtain a target face feature extraction model.
In another example of the present invention, a human face comprehensive loss function may be substituted into a back propagation algorithm to adjust the overall parameters of the initial human body clothing feature matrix extraction network, and the output of the initial human body clothing feature matrix extraction network is calculated and the tracking target value is continuously calculated until the tracking target value is equal to the optimized target value, thereby completing the overall parameter adjustment.
The back propagation algorithm may be, but is not limited to, a self-optimizing estimation method for a dual-constrained target neural network.
And step 206, determining whether the personnel corresponding to the face image to be recognized has the work qualification or not based on the comparison result of whether each face feature similarity reaches a preset threshold or not.
In one example of the present invention, step 206 may further include the following sub-steps:
if any one of the face feature similarities reaches a preset threshold value, determining that the personnel corresponding to the face image to be recognized has the work qualification;
and if the similarity of each face feature does not reach the preset threshold, determining that the personnel corresponding to the face image to be recognized do not have the work qualification.
In the embodiment of the invention, the face image to be recognized is received, the image preprocessing is carried out on the face image to be recognized to obtain a plurality of image matrixes to be recognized with different scales, the plurality of image matrixes to be recognized are respectively input into the target face feature extraction model to output the face feature similarity corresponding to the image matrixes to be recognized of each scale, and finally, whether personnel corresponding to the face image to be recognized have work qualification or not is determined based on the comparison result of whether the face feature similarity is larger than the preset threshold value or not. Therefore, the technical problems that the accuracy of screening personnel is reduced and potential safety hazards are possibly caused due to the fact that the recognition rate of the existing face video monitoring technology is reduced due to the complex environment are solved, the accuracy of screening the personnel in the complex environment is effectively improved, and the potential safety hazards are prevented.
Referring to fig. 4, fig. 4 is a schematic structural diagram of an electric power operation monitoring platform according to a third embodiment of the present invention.
The intelligent recognition system comprises a physical resource layer 401, a scheduling management layer 402, a training environment layer 403 and a service application layer 404, ensures that an intelligent recognition algorithm can normally and stably run, is transplanted to an edge side terminal, and provides a full life cycle management function of model training, prediction, evaluation and deployment for a user.
The physical resource layer 401 includes heterogeneous computing hardware (CPU, GPU), storage, network devices, and security devices.
The scheduling management layer 402 is developed based on kubernets and docker, and includes cluster management, resource virtualization, task scheduling, and the like.
The training environment layer 403 is a service provided in the form of docker, and includes machine learning/deep learning environments such as a mainstream learning framework TensorFlow, pytorrech, Caffe, scimit-spare, XGBoost, and the like, and integrates an interactive code debugging notebook such as JupyterHub, and an MPI parallel programming interface. And the system running environment and the learning environment perform iterative management of versions through a docker warehouse.
The business application layer 404 includes data processing, data tagging, model training, and model publishing.
The model training module is based on various machine learning and deep learning training environments, configures parameters through a pre-written training script and then suspends training. The whole training process is automatically completed by a pipeline (pipeline) built in a background, and model production is developed around the processes of data processing, data labeling, training and model management. The model training is carried out through docker presetting learning environments such as TensorFlow, PyTorch, Caffe, scimit-lean, XGboost and the like. By using the task scheduling system, a user can submit a learning task code to the cluster, the task management system allocates resources for the user according to the quota of the user, creates an environment specified by the user, adds the learning task into the task queue, and runs the learning program when the resources are free. Users can submit codes by one key to generate distributed tasks, and development cost and resource occupation are greatly reduced.
Referring to fig. 5, fig. 5 is a block diagram illustrating a structure of an electric power operator screening apparatus according to a fourth embodiment of the present invention.
The invention also provides a device for discriminating the electric power operating personnel, which comprises:
the image receiving module 501 is configured to receive a face image to be recognized;
the image preprocessing module 502 is configured to perform image preprocessing on the face image to be recognized, so as to generate a plurality of image matrixes to be recognized with different scales;
a face feature data output module 503, configured to input the multiple image matrices to be recognized into a target face feature extraction model, and output face feature data corresponding to each image matrix to be recognized; the target face feature extraction model is generated by a preset model training module 504, and the face feature data includes face feature similarity;
and a working qualification judging module 505, configured to determine whether the person corresponding to the face image to be recognized has a working qualification based on a comparison result of whether each face feature similarity reaches a preset threshold.
Optionally, the image preprocessing module 502 includes:
the image cutting submodule is used for cutting the face image to be recognized to obtain a plurality of first image matrixes with different scales;
the image turning sub-module is used for respectively executing image turning operation on the plurality of first image matrixes according to preset requirements to obtain a plurality of second image matrixes;
and the image normalization submodule is used for respectively performing image normalization operation and image normalization operation on the second image matrixes to generate a plurality of to-be-identified image matrixes with different scales.
Optionally, the model training module 504 includes:
the historical face image acquisition submodule is used for respectively acquiring face images in various historical electric power operation scenes;
the image preprocessing submodule is used for preprocessing the face image to generate a plurality of historical face image matrixes with different scales;
the historical face image similarity determining submodule is used for inputting a plurality of historical face image matrixes into a preset initial face feature extraction model to obtain a plurality of historical face image similarities;
a comprehensive loss function determining submodule for determining a comprehensive loss function of the face based on the similarity of a plurality of historical face images and a preset weight;
and the model training submodule is used for training the initial face feature extraction model by adopting the face comprehensive loss function to obtain a target face feature extraction model.
Optionally, the preset weights include a first weight and a second weight, and the comprehensive loss function determining sub-module includes:
the similarity obtaining unit is used for selecting a first similarity and a second similarity from a plurality of historical face image similarities;
the first accumulation unit is used for determining a first type of similarity accumulated value of the historical human face image matrix based on the first similarity and the first weight;
the second accumulation unit is used for determining a second type of similarity accumulation value of the historical human face image matrix based on the second similarity and the second weight;
and the face comprehensive loss function determining unit is used for determining the face comprehensive loss function according to the first type of similarity accumulated value and the second type of similarity accumulated value.
Optionally, the work qualification determining module 505 includes:
the first judgment submodule is used for determining that the personnel corresponding to the face image to be recognized has the working qualification if any one of the face feature similarity reaches a preset threshold value;
and the second judging submodule is used for determining that the personnel corresponding to the face image to be recognized does not have the working qualification if the face feature similarity does not reach the preset threshold value.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. An electric power operator screening method characterized by comprising:
receiving a face image to be recognized;
carrying out image preprocessing on the face image to be recognized to generate a plurality of image matrixes to be recognized with different scales;
respectively inputting the image matrixes to be recognized into a target face feature extraction model, and outputting face feature data corresponding to each image matrix to be recognized; the target human face feature extraction model is generated through a preset model training process, and the human face feature data comprise human face feature similarity;
and determining whether the personnel corresponding to the face image to be recognized has the work qualification or not based on the comparison result of whether the face feature similarity reaches the preset threshold or not.
2. The electric power operator screening method according to claim 1, wherein the step of performing image preprocessing on the face image to be recognized to generate a plurality of image matrixes to be recognized with different scales includes:
cutting the face image to be recognized to obtain a plurality of first image matrixes with different scales;
respectively executing image turning operation on the plurality of first image matrixes according to preset requirements to obtain a plurality of second image matrixes;
and respectively executing image normalization operation and image normalization operation on the second image matrixes to generate a plurality of to-be-identified image matrixes with different scales.
3. The electric power operator screening method according to claim 1, wherein the model training process includes:
respectively acquiring human face images in various historical electric power operation scenes;
carrying out image preprocessing on the face image to generate a plurality of historical face image matrixes with different scales;
inputting a plurality of historical face image matrixes into a preset initial face feature extraction model to obtain a plurality of historical face image similarities;
determining a face comprehensive loss function based on the similarity of a plurality of historical face images and a preset weight;
and training the initial face feature extraction model by adopting the face comprehensive loss function to obtain a target face feature extraction model.
4. The electric power operator screening method according to claim 3, wherein the preset weight includes a first weight and a second weight, and the step of determining the comprehensive loss function of the face based on the similarity of the plurality of historical face images and the preset weight includes:
selecting a first similarity and a second similarity from the plurality of historical face image similarities;
determining a first class similarity accumulated value of the historical face image matrix based on the first similarity and the first weight;
determining a second type similarity accumulated value of the historical face image matrix based on the second similarity and the second weight;
and determining the face comprehensive loss function according to the first type of similarity accumulated value and the second type of similarity accumulated value.
5. The method for discriminating electric power operators according to claim 1, wherein the step of determining whether the person corresponding to the face image to be recognized has the work qualification based on the comparison result of whether the similarity of each face feature reaches a preset threshold includes:
if any one of the face feature similarities reaches a preset threshold value, determining that the personnel corresponding to the face image to be recognized has the work qualification;
and if the similarity of each face feature does not reach the preset threshold, determining that the personnel corresponding to the face image to be recognized do not have the work qualification.
6. An electric power operator screening apparatus, characterized by comprising:
the image receiving module is used for receiving a face image to be recognized;
the image preprocessing module is used for preprocessing the face image to be recognized to generate a plurality of image matrixes to be recognized with different scales;
the human face feature data output module is used for respectively inputting the image matrixes to be recognized into a target human face feature extraction model and outputting human face feature data corresponding to each image matrix to be recognized; the target human face feature extraction model is generated through a preset model training module, and the human face feature data comprise human face feature similarity;
and the work qualification judging module is used for determining whether the personnel corresponding to the face image to be recognized has work qualification or not based on the comparison result of whether the face feature similarity reaches the preset threshold value or not.
7. The electric power operator screening apparatus according to claim 6, wherein the image preprocessing module includes:
the image cutting submodule is used for cutting the face image to be recognized to obtain a plurality of first image matrixes with different scales;
the image turning sub-module is used for respectively executing image turning operation on the plurality of first image matrixes according to preset requirements to obtain a plurality of second image matrixes;
and the image normalization submodule is used for respectively performing image normalization operation and image normalization operation on the second image matrixes to generate a plurality of to-be-identified image matrixes with different scales.
8. The electric power operator screening apparatus according to claim 6, wherein the model training module includes:
the historical face image acquisition submodule is used for respectively acquiring face images in various historical electric power operation scenes;
the image preprocessing submodule is used for preprocessing the face image to generate a plurality of historical face image matrixes with different scales;
the historical face image similarity determining submodule is used for inputting a plurality of historical face image matrixes into a preset initial face feature extraction model to obtain a plurality of historical face image similarities;
a comprehensive loss function determining submodule for determining a comprehensive loss function of the face based on the similarity of a plurality of historical face images and a preset weight;
and the model training submodule is used for training the initial face feature extraction model by adopting the face comprehensive loss function to obtain a target face feature extraction model.
9. The electric power operator screening apparatus according to claim 8, wherein the preset weight includes a first weight and a second weight, and the synthetic loss function determination submodule includes:
the similarity obtaining unit is used for selecting a first similarity and a second similarity from a plurality of historical face image similarities;
the first accumulation unit is used for determining a first type of similarity accumulated value of the historical human face image matrix based on the first similarity and the first weight;
the second accumulation unit is used for determining a second type of similarity accumulation value of the historical human face image matrix based on the second similarity and the second weight;
and the face comprehensive loss function determining unit is used for determining the face comprehensive loss function according to the first type of similarity accumulated value and the second type of similarity accumulated value.
10. The electric power operator screening apparatus according to claim 3, wherein the work qualification determination module includes:
the first judgment submodule is used for determining that the personnel corresponding to the face image to be recognized has the working qualification if any one of the face feature similarity reaches a preset threshold value;
and the second judging submodule is used for determining that the personnel corresponding to the face image to be recognized does not have the working qualification if the face feature similarity does not reach the preset threshold value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011478342.6A CN112464897B (en) | 2020-12-15 | 2020-12-15 | Electric power operator screening method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011478342.6A CN112464897B (en) | 2020-12-15 | 2020-12-15 | Electric power operator screening method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112464897A true CN112464897A (en) | 2021-03-09 |
| CN112464897B CN112464897B (en) | 2021-09-24 |
Family
ID=74804765
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011478342.6A Active CN112464897B (en) | 2020-12-15 | 2020-12-15 | Electric power operator screening method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112464897B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030086593A1 (en) * | 2001-05-31 | 2003-05-08 | Chengjun Liu | Feature based classification |
| CN104517104A (en) * | 2015-01-09 | 2015-04-15 | 苏州科达科技股份有限公司 | Face recognition method and face recognition system based on monitoring scene |
| CN108182397A (en) * | 2017-12-26 | 2018-06-19 | 王华锋 | A kind of multiple dimensioned face verification method of multi-pose |
| CN108509828A (en) * | 2017-02-28 | 2018-09-07 | 深圳市朗驰欣创科技股份有限公司 | A kind of face identification method and face identification device |
-
2020
- 2020-12-15 CN CN202011478342.6A patent/CN112464897B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030086593A1 (en) * | 2001-05-31 | 2003-05-08 | Chengjun Liu | Feature based classification |
| CN104517104A (en) * | 2015-01-09 | 2015-04-15 | 苏州科达科技股份有限公司 | Face recognition method and face recognition system based on monitoring scene |
| CN108509828A (en) * | 2017-02-28 | 2018-09-07 | 深圳市朗驰欣创科技股份有限公司 | A kind of face identification method and face identification device |
| CN108182397A (en) * | 2017-12-26 | 2018-06-19 | 王华锋 | A kind of multiple dimensioned face verification method of multi-pose |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112464897B (en) | 2021-09-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111008640A (en) | Image recognition model training and image recognition method, device, terminal and medium | |
| CN110135476A (en) | A kind of detection method of personal safety equipment, device, equipment and system | |
| CN112541393A (en) | Transformer substation personnel detection method and device based on deep learning | |
| CN115376130A (en) | Method, device, equipment and medium for generating electric power construction site identification model | |
| CN115565101A (en) | Production safety abnormity identification method and device, electronic equipment and storage medium | |
| CN111476349A (en) | Model testing method and server | |
| CN112464897B (en) | Electric power operator screening method and device | |
| CN112528855B (en) | Electric power operation dressing standard identification method and device | |
| CN113536842B (en) | Electric power operation personnel safety dressing identification method and device | |
| CN113627298A (en) | Training method of target detection model and method and device for detecting target object | |
| CN116311082B (en) | Wearing detection method and system based on matching of key parts and images | |
| CN117726322A (en) | Intelligent management method and system for probe test equipment | |
| CN115641607A (en) | Method, device, equipment and storage medium for detecting wearing behavior of power construction site operator | |
| CN116798074A (en) | Stranger identification method and device and electronic equipment | |
| CN114998275A (en) | State recognition method of target object and training method of deep learning model | |
| CN114067360A (en) | Pedestrian attribute detection method and device | |
| CN116778583A (en) | Method, device and equipment for discriminating compliance of worker behaviors | |
| CN112949606A (en) | Method and device for detecting wearing state of industrial garment, storage medium and electronic device | |
| CN116189028B (en) | Image recognition method, device, electronic equipment and storage medium | |
| CN112528996A (en) | Picture processing method, device, equipment, storage medium and computer program product | |
| CN112990045B (en) | Method and apparatus for generating image change detection model and image change detection | |
| CN112101266B (en) | Multi-ARM-based motion recognition model distributed reasoning method | |
| CN115641642A (en) | Behavior recognition method and device based on multi-semantic multi-modal information | |
| CN117651000A (en) | Chaotic engineering test system and method applied to heterogeneous environments of cloud-on-cloud and cloud-off | |
| CN117853789A (en) | Image detection method and device |
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 |