CN117152849A - Novel method for identifying identities of underground weak characteristic personnel of coal mine - Google Patents
Novel method for identifying identities of underground weak characteristic personnel of coal mine Download PDFInfo
- Publication number
- CN117152849A CN117152849A CN202311298611.4A CN202311298611A CN117152849A CN 117152849 A CN117152849 A CN 117152849A CN 202311298611 A CN202311298611 A CN 202311298611A CN 117152849 A CN117152849 A CN 117152849A
- Authority
- CN
- China
- Prior art keywords
- personnel
- image
- underground
- identification
- coal mine
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 51
- 239000003245 coal Substances 0.000 title claims abstract description 28
- 238000001514 detection method Methods 0.000 claims abstract description 14
- 238000009499 grossing Methods 0.000 claims abstract description 13
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 238000005457 optimization Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims abstract description 7
- 238000010801 machine learning Methods 0.000 claims abstract description 6
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 5
- 238000005286 illumination Methods 0.000 claims description 20
- 238000000605 extraction Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 4
- 230000019771 cognition Effects 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 230000008014 freezing Effects 0.000 claims description 3
- 238000007710 freezing Methods 0.000 claims description 3
- 238000013508 migration Methods 0.000 claims description 3
- 230000005012 migration Effects 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000013526 transfer learning Methods 0.000 claims description 3
- 238000003860 storage Methods 0.000 abstract description 6
- 238000004458 analytical method Methods 0.000 abstract description 4
- 230000010354 integration Effects 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 9
- 238000012549 training Methods 0.000 description 4
- 230000009977 dual effect Effects 0.000 description 2
- 239000000428 dust Substances 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000010883 coal ash Substances 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000036544 posture Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000452 restraining effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- G06V40/50—Maintenance of biometric data or enrolment thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/096—Transfer learning
-
- 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
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of 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
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a novel method for identifying the identity of a weak characteristic person in a coal mine, which comprises the following steps: acquiring video stream data of underground weak characteristic personnel of the coal mine through a detection camera; processing the video stream data, intercepting an underground personnel image, and constructing an underground weak characteristic personnel re-identification data set; preprocessing the dataset image, including image enhancement and noise cancellation; establishing a personnel re-identification model by using a machine learning algorithm, wherein the personnel re-identification model comprises a convolutional neural network and label smoothing and optimization; updating the personnel re-identification model periodically by using an incremental learning method so as to adapt to new underground weak characteristic personnel image data; and (5) carrying out identity re-identification on the underground weak characteristic personnel, and outputting an identification result. The application can improve the re-identification capability of underground weak characteristic personnel of the coal mine, effectively reduce the requirement on storage, improve the system efficiency, realize data integration and analysis and improve the level of safety management.
Description
Technical Field
The application belongs to the technical field of computer vision, and particularly relates to a novel method for identifying the identities of underground weak characteristic personnel of a coal mine.
Background
The accurate positioning and identity recognition technology of mine personnel can effectively early warn illegal behaviors, suppress underground overman production and avoid or reduce coal mine accidents. However, the current underground personnel monitoring research mostly only stops at the detection stage, mainly locates intervals, and lacks classification and identification of individual identities of interval group personnel; the identity confirmation method based on the common face recognition still has limitation in application aspects such as the identity recognition of illegal persons due to special environments such as low illumination, strong light interference and high dust of mines, wearing of labor protection articles of underground persons, face coal ash and the like.
The underground personnel re-identification technology is used as an important means for confirming personnel identity information, and a liberation scheme is provided for the identity identification of the personnel with weak characteristics in the mine. However, the internal structure of the mine is complex in environment, is influenced by the environment such as low illumination and high dust fog in the pit, is difficult to acquire visual images, cannot acquire personnel information in place, and is an important problem to be solved at present.
Disclosure of Invention
The application aims to provide a novel method for identifying the identity of a weak characteristic person in a coal mine, which can improve the re-identification capability of the weak characteristic person in the coal mine, effectively reduce the requirement for storage, improve the system efficiency, realize data integration and analysis and improve the level of safety management.
In order to achieve the purpose, the application provides a novel method for identifying the identities of weak characteristic personnel in a coal mine, which comprises the following steps:
step one: acquiring video stream data of underground weak characteristic personnel of the coal mine through a detection camera;
step two: processing the video stream data, intercepting an underground personnel image, and constructing an underground weak characteristic personnel re-identification data set;
step three: preprocessing the dataset image, including image enhancement and noise cancellation;
step four: establishing a personnel re-identification model by using a machine learning algorithm, wherein the personnel re-identification model comprises a convolutional neural network and label smoothing and optimization;
step five: updating the personnel re-identification model periodically by using an incremental learning method so as to adapt to new underground weak characteristic personnel image data;
step six: and (5) carrying out identity re-identification on the underground weak characteristic personnel, and outputting an identification result.
As a further scheme of the application: processing the video stream data and constructing a downhole weak characteristic personnel re-identification data set, which comprises the following steps:
pulling a video stream from a network camera by using OpenCV and GSstreamer to provide real-time video preview and acquisition functions;
data acquisition is carried out in a multithreading mode;
and generating a candidate frame by using a target detection algorithm Faster-RCNN, and intercepting a personnel image.
As a further scheme of the application: the preprocessing of the image data is to enhance the image data by using a dual authentication type generation countermeasure network, and comprises a generator for enhancing the low-illumination image and an identifier for judging the authenticity of the enhanced image.
As a further scheme of the application: the generator is used for extracting and upsampling the characteristics of the image data;
the feature extraction part comprises convolution and maximum pooling operation, so that parameters required to be trained are reduced, and feature graphs with different scales are obtained;
and in the up-sampling part, the image restoration of different sizes is completed through deconvolution, and finally the enhanced image is output.
As a further scheme of the application: the discriminator is to further process the enhanced image of the generator, and adopts a global discriminator and a local discriminator:
the global discriminator is used for improving the self-adaptive capacity of the global low-illumination image and improving the illumination intensity in the global range;
the local discriminator randomly selects local small blocks from the enhanced image and the normal bright illumination image to judge true and false, so that the cognition capability of the model on the illumination non-uniform image is enhanced, and the illumination is improved in a local range.
As a further scheme of the application: the method for establishing the personnel re-identification model by using the machine learning algorithm specifically comprises the following steps:
carrying out feature extraction on the personnel image subjected to input pretreatment by utilizing a RestNet-50, wherein the feature extraction comprises global features and local features;
obtaining the complementarity scores of the global and local features by using a k nearest neighbor method;
and performing label smoothing and label optimization by using the feature complementarity score so as to reduce image noise and enhance the recognition capability of the features.
As a further scheme of the application: performing label smoothing and label optimization by using the feature complementarity score to reduce image noise and enhance the recognition capability of the features, and specifically comprising the following steps:
calculating Jaccard similarity between k nearest neighbors of the global feature and the local feature to measure mutual similarity degree, and obtaining feature complementarity score;
optimizing each local feature tag according to the corresponding feature complementarity score using tag smoothing;
by means of prediction of the local features, the global features are guided to capture reliable fine-grained information from the local features, and the tags are perfected by means of more reliable information.
As a further scheme of the application: the incremental learning method is used for updating the personnel re-identification model regularly, and specifically comprises the following steps:
continuously collecting new underground weak characteristic personnel image data, including new personnel, new postures, new environmental conditions and the like;
for the newly acquired data, preprocessing steps similar to the initial data set are carried out, so that the quality and consistency of the data are ensured;
combining the new data with the initial data set, and updating the model by using a transfer learning method;
and evaluating the updated model, and using the test data set to verify the accuracy and the performance so as to guide the further model optimization and adjustment.
As a further scheme of the application: the process of updating the model by the migration learning method comprises the following steps:
determining whether to freeze the weight of the initial model according to the number and diversity of samples of the new task, wherein if the samples are fewer, the freezing parameters can prevent overfitting;
the newly added layer is trained by using a smaller learning rate, so that the knowledge learned by the initial model is prevented from being destroyed, and a slightly larger learning rate is adopted for the fine tuning layer.
Compared with the prior art, the application has the following beneficial effects:
1. the application improves the re-identification capability of underground weak feature personnel of the coal mine, and can more accurately identify and match the identity of the underground weak feature personnel by adopting a label smoothing and label optimizing method based on the feature complementarity score;
2. the application has lower storage requirement, the traditional method generally needs to store a large amount of training data and model parameters, and the application uses an incremental learning method, so that the model can be updated without retraining the whole model, thereby reducing the storage requirement and improving the system efficiency;
3. the application can realize data integration and analysis, upload the re-identified personnel data to the platform for integration and analysis, provide scientific data support for coal mine safety management, and improve the safety management level.
Drawings
Fig. 1 shows a schematic diagram of the working principle provided by the embodiment of the application.
Fig. 2 shows a schematic diagram of an image preprocessing network according to an embodiment of the present application.
Fig. 3 is a schematic diagram of a downhole personnel re-identification network structure according to an embodiment of the present application.
Fig. 4 is a schematic diagram showing a detection result of person identification re-identification in the embodiment of the application.
Detailed Description
The application is further illustrated by the following examples.
Fig. 1 shows a schematic diagram of the working principle provided by the embodiment of the application.
A new method for identifying the identities of weak characteristic personnel in the underground coal mine comprises the following steps:
step one: acquiring video stream data of underground weak characteristic personnel of the coal mine through a detection camera;
wherein, the detection camera needs to have high resolution; a wide viewing angle is required; the ability to adapt to complex environments is needed; the method needs to have stronger instantaneity; the underground coal mine weak characteristic personnel identification system has strong durability and can work stably and reliably in an underground environment, so that accurate and heavy identification of underground coal mine weak characteristic personnel is ensured.
Step two: processing the video stream data, intercepting an underground personnel image, and constructing an underground weak characteristic personnel re-identification data set, wherein the method specifically comprises the following steps of:
pulling a video stream from a network camera by using OpenCV and GSstreamer to provide real-time video preview and acquisition functions;
the data acquisition is carried out in a multithreading mode, so that the running efficiency and stability are improved;
and generating a candidate frame by using a target detection algorithm Faster-RCNN, and intercepting a personnel image.
In the embodiment of the application, the underground coal mine weak characteristic personnel re-identification data set uses videos provided by a Chen four-floor coal mine video control system, 20000 pictures are extracted from the videos, a target detection fast-RCNN algorithm is used for carrying out target detection to generate candidate frames, and personnel images are intercepted. Finally, 63852 images of 2000 pedestrians captured by 10 cameras are obtained, wherein the training set comprises 28394 images of 1000 pedestrians, the test set comprises 31110 images of 1000 pedestrians and interferents, and the query set comprises 4348 images of 1000 pedestrians. Fig. 4 is a schematic diagram showing a human intrusion detection result in an embodiment of the present application.
Step three: preprocessing the dataset image, including image enhancement and noise cancellation;
specific: image enhancement of the image data using a dual authentication generation countermeasure network includes a generator for enhancing a low-light image and an authenticator for performing true and false discrimination of the enhanced image. Fig. 2 is a schematic diagram of an image preprocessing network according to an embodiment of the present application.
Furthermore, the generator adopts a U-net network structure, and performs feature extraction and up-sampling on the image data;
the feature extraction part comprises convolution and maximum pooling operation, so that parameters required to be trained are reduced, and feature graphs with different scales are obtained;
and in the up-sampling part, the image restoration of different sizes is completed through deconvolution, and finally the enhanced image is output.
The mathematical model of the generator is expressed as:
x f ＝G(I x )
wherein I is x X is the true low-illumination image input f To enhance the post-image, G generates a mapping function. In order to keep the essential information of the original image to the maximum extent, a local image area with a fixed size is selected from the real low-light image to strengthen the local image area on the basis that the generator initially acquires the enhanced image. The real low-light image and the enhanced image are subjected to feature extraction by utilizing a VGG-16 model which is trained in advance on the ImageNet, and the self-feature retention loss is used for restraining the distance between the real low-light image and the enhanced image.
Further, the discriminator is to further process the enhanced image of the generator, and adopts a global discriminator and a local discriminator at the same time:
the global discriminator is used for improving the self-adaptive capacity of the global low-illumination image and improving the illumination intensity in the global range;
the local discriminator randomly selects local small blocks from the enhanced image and the normal bright illumination image to judge true and false, so that the cognition capability of the model on the illumination non-uniform image is enhanced, and the illumination is improved in a local range.
After the image is input into the discriminator, the characteristic extraction is carried out firstly, then the characteristic is summarized through the full-connection layer, and finally the authenticity is judged through the output value.
Step four: the method comprises the following steps of establishing a personnel re-identification model by using a machine learning algorithm, wherein the personnel re-identification model comprises a Convolutional Neural Network (CNN) and label smoothing and optimizing:
carrying out feature extraction on the personnel image subjected to input pretreatment by utilizing a RestNet-50, wherein the feature extraction comprises global features and local features;
obtaining the complementarity scores of the global and local features by using a k nearest neighbor method;
performing label smoothing and label optimization by using the feature complementarity score to reduce image noise and enhance the recognition capability of the features, and specifically comprising the following steps:
calculating Jaccard similarity between k nearest neighbors of the global feature and the local feature to measure mutual similarity degree, and obtaining feature complementarity score;
optimizing each local feature tag according to the corresponding feature complementarity score using tag smoothing;
by means of prediction of the local features, the global features are guided to capture reliable fine-grained information from the local features, and the tags are perfected by means of more reliable information. Fig. 3 shows a schematic diagram of the downhole personnel re-identification network structure according to the embodiment.
In the embodiment of the application, the underground weak characteristic personnel re-identification network is realized by a Pytorch framework, a learned model is converted into an ONNX format, and then the ONNX model is converted into a TensorRT model to realize model acceleration reasoning so as to facilitate real-time reasoning after being deployed to terminal equipment, the terminal equipment is applied to an actual working scene, a C++ rewriting model preprocessing and post-processing part is used for model deployment, and a Cuda is used for preprocessing to accelerate image scaling and model reasoning speed.
Step five: the incremental learning method is used for updating the personnel re-identification model regularly so as to adapt to new underground weak characteristic personnel image data, and the method specifically comprises the following steps:
continuously collecting new underground weak characteristic personnel image data;
for the newly acquired data, preprocessing steps similar to the initial data set are carried out, so that the quality and consistency of the data are ensured;
combining the new data with the initial data set, and updating the model by using a transfer learning method;
and evaluating the updated model, and using the test data set to verify the accuracy and the performance so as to guide the further model optimization and adjustment.
Further, the process of updating the model by the migration learning method comprises the following steps:
determining whether to freeze the weight of the initial model according to the number and diversity of samples of the new task, wherein if the samples are fewer, the freezing parameters can prevent overfitting;
the newly added layer is trained by using a smaller learning rate, so that the knowledge learned by the initial model is prevented from being destroyed, and a slightly larger learning rate is adopted for the fine tuning layer.
In the embodiment of the application, an initial personnel re-identification model is used as a basic model, and the model is pre-trained and fine-tuned on a large amount of training data, so that the model has preliminary identification capability.
New downhole weak feature personnel image data is continuously obtained over time. The new data may contain situations where the previous model has not been touched or has not been fully adapted. To improve the accuracy and robustness of the model, it is necessary to incorporate these new data into the training process of the model.
Combining the new data with the basic model, updating the model in a repeated iteration mode, gradually improving the performance of the model, keeping the timeliness of the model, and ensuring that the model keeps high accuracy when processing the new underground weak characteristic personnel image data.
Step six: and (5) carrying out identity re-identification on the underground weak characteristic personnel, and outputting an identification result. Fig. 4 shows a schematic diagram of a person identification re-identification detection result in an embodiment of the present application, where the first column is a person image to be queried, and the last ten columns are person images with top ten query results.
It should be understood that, although the steps in the flowcharts of the embodiments of the present application are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in various embodiments may include multiple sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor do the order in which the sub-steps or stages are performed necessarily performed in sequence, but may be performed alternately or alternately with at least a portion of the sub-steps or stages of other steps or other steps.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above-described embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above-described embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
Claims (9)
1. The new method for identifying the identities of the underground weak characteristic personnel of the coal mine is characterized by comprising the following steps of:
step one: acquiring video stream data of underground weak characteristic personnel of the coal mine through a detection camera;
step two: processing the video stream data, intercepting an underground personnel image, and constructing an underground weak characteristic personnel re-identification data set;
step three: preprocessing the dataset image, including image enhancement and noise cancellation;
step four: establishing a personnel re-identification model by using a machine learning algorithm, wherein the personnel re-identification model comprises a convolutional neural network and label smoothing and optimization;
step five: updating the personnel re-identification model periodically by using an incremental learning method so as to adapt to new underground weak characteristic personnel image data;
step six: and (5) carrying out identity re-identification on the underground weak characteristic personnel, and outputting an identification result.
2. The new method for identifying the identity of the underground weak character personnel of the coal mine according to claim 1, which is characterized by processing the video stream data and constructing an underground weak character personnel re-identification data set, and specifically comprises the following steps:
pulling a video stream from a network camera by using OpenCV and GSstreamer to provide real-time video preview and acquisition functions;
data acquisition is carried out in a multithreading mode;
and generating a candidate frame by using a target detection algorithm Faster-RCNN, and intercepting a personnel image.
3. The method according to claim 1, wherein the preprocessing of the image data is image enhancement of the image data by using a dual-authentication generation countermeasure network, and the method comprises a generator for enhancing a low-illumination image and an identifier for performing true and false discrimination on the enhanced image.
4. A new method for identifying weak character personnel under coal mine according to claim 3, wherein the generator is used for extracting and upsampling the characteristics of the image data;
the feature extraction part comprises convolution and maximum pooling operation, so that parameters required to be trained are reduced, and feature graphs with different scales are obtained;
and in the up-sampling part, the image restoration of different sizes is completed through deconvolution, and finally the enhanced image is output.
5. A new method for identifying weak personnel in a coal mine according to claim 3, wherein the identifier is to further process the enhanced image of the generator, and a global identifier and a local identifier are adopted at the same time:
the global discriminator is used for improving the self-adaptive capacity of the global low-illumination image and improving the illumination intensity in the global range;
the local discriminator randomly selects local small blocks from the enhanced image and the normal bright illumination image to judge true and false, so that the cognition capability of the model on the illumination non-uniform image is enhanced, and the illumination is improved in a local range.
6. The new method for identifying the identity of the underground weak characteristic personnel of the coal mine according to claim 1, which is characterized by utilizing a machine learning algorithm to establish a personnel re-identification model, and specifically comprising the following steps:
carrying out feature extraction on the personnel image subjected to input pretreatment by utilizing a RestNet-50, wherein the feature extraction comprises global features and local features;
obtaining the complementarity scores of the global and local features by using a k nearest neighbor method;
and performing label smoothing and label optimization by using the feature complementarity score so as to reduce image noise and enhance the recognition capability of the features.
7. The method for recognizing the identity of the weak feature personnel in the underground coal mine according to claim 6, wherein the tag smoothing and the tag optimizing are performed by using the feature complementarity score so as to reduce image noise and enhance the recognition capability of the features, and the method specifically comprises the following steps:
calculating Jaccard similarity between k nearest neighbors of the global feature and the local feature to measure mutual similarity degree, and obtaining feature complementarity score;
optimizing each local feature tag according to the corresponding feature complementarity score using tag smoothing;
by means of prediction of the local features, the global features are guided to capture reliable fine-grained information from the local features, and the tags are perfected by means of more reliable information.
8. The new method for identifying the identity of the underground weak characteristic personnel of the coal mine according to claim 1, which is characterized by periodically updating the personnel re-identification model by using an incremental learning method, and specifically comprises the following steps:
continuously collecting new underground weak characteristic personnel image data;
for the newly acquired data, preprocessing steps similar to the initial data set are carried out, so that the quality and consistency of the data are ensured;
combining the new data with the initial data set, and updating the model by using a transfer learning method;
and evaluating the updated model, and using the test data set to verify the accuracy and the performance so as to guide the further model optimization and adjustment.
9. The new method for identifying the identity of the underground weak character personnel of the coal mine according to claim 8, wherein the process of updating the model by the migration learning method is as follows:
determining whether to freeze the weight of the initial model according to the number and diversity of samples of the new task, wherein if the samples are fewer, the freezing parameters can prevent overfitting;
the newly added layer is trained by using a smaller learning rate, so that the knowledge learned by the initial model is prevented from being destroyed, and a slightly larger learning rate is adopted for the fine tuning layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311298611.4A CN117152849A (en) | 2023-10-09 | 2023-10-09 | Novel method for identifying identities of underground weak characteristic personnel of coal mine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311298611.4A CN117152849A (en) | 2023-10-09 | 2023-10-09 | Novel method for identifying identities of underground weak characteristic personnel of coal mine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117152849A true CN117152849A (en) | 2023-12-01 |
Family
ID=88884326
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311298611.4A Pending CN117152849A (en) | 2023-10-09 | 2023-10-09 | Novel method for identifying identities of underground weak characteristic personnel of coal mine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117152849A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111985601A (en) * | 2019-05-21 | 2020-11-24 | 富士通株式会社 | Data identification method for incremental learning |
| CN114882441A (en) * | 2022-05-30 | 2022-08-09 | 郑州大学 | Pedestrian re-identification method and system based on deep learning |
| CN116152858A (en) * | 2023-02-28 | 2023-05-23 | 常州大学 | Unsupervised pedestrian re-identification method based on double cross attention mechanism |
| CN116363498A (en) * | 2022-05-12 | 2023-06-30 | 河海大学 | Loess plateau terrace automatic identification method based on deep migration learning |
-
2023
- 2023-10-09 CN CN202311298611.4A patent/CN117152849A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111985601A (en) * | 2019-05-21 | 2020-11-24 | 富士通株式会社 | Data identification method for incremental learning |
| CN116363498A (en) * | 2022-05-12 | 2023-06-30 | 河海大学 | Loess plateau terrace automatic identification method based on deep migration learning |
| CN114882441A (en) * | 2022-05-30 | 2022-08-09 | 郑州大学 | Pedestrian re-identification method and system based on deep learning |
| CN116152858A (en) * | 2023-02-28 | 2023-05-23 | 常州大学 | Unsupervised pedestrian re-identification method based on double cross attention mechanism |
Non-Patent Citations (5)
| Title |
|---|
| 刘国华主编: "《机器视觉技术》", 30 November 2021, 华中科技大学出版社, pages: 301 - 302 * |
| 吴心筱等: "《迁移学习基础及应用》", 31 May 2021, 北京理工大学出版社, pages: 99 * |
| 孙彦景等: "《联合DD-GAN和全局特征的井下人员重识别方法》", 《 西安电子科技大学学报》, 8 September 2021 (2021-09-08), pages 1 - 11 * |
| 王泽楷著: "《产品管理与运营系列丛书 AI产品经理 方法技术与实战》", 31 October 2022, 机械工业出版社, pages: 222 - 224 * |
| 王溢琴: "《基于深度学习的遥感图像语义分割方法研究》", 31 December 2022, 科学技术文献出版社, pages: 58 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111080620B (en) | Road disease detection method based on deep learning | |
| CN111626123B (en) | Video data processing method, device, computer equipment and storage medium | |
| CN110414507B (en) | License plate recognition method and device, computer equipment and storage medium | |
| CN112766195B (en) | Electrified railway bow net arcing visual detection method | |
| Zhang et al. | Bridge damage detection using a single-stage detector and field inspection images | |
| CN111539320B (en) | Multi-view gait recognition method and system based on mutual learning network strategy | |
| Hou et al. | A deep learning method for pavement crack identification based on limited field images | |
| CN112966665A (en) | Pavement disease detection model training method and device and computer equipment | |
| CN104978567A (en) | Vehicle detection method based on scenario classification | |
| CN104751485A (en) | GPU adaptive foreground extracting method | |
| CN111191535A (en) | Pedestrian detection model construction method based on deep learning and pedestrian detection method | |
| CN104463104A (en) | Fast detecting method and device for static vehicle target | |
| CN108847031A (en) | Traffic behavior monitoring method, device, computer equipment and storage medium | |
| CN112419261A (en) | Visual acquisition method and device with abnormal point removing function | |
| CN112418033A (en) | Landslide slope surface segmentation and identification method based on mask rcnn neural network | |
| CN117197746A (en) | Safety monitoring system and method based on deep learning | |
| CN111402185B (en) | Image detection method and device | |
| CN107862314B (en) | Code spraying identification method and device | |
| CN117496570A (en) | Face recognition method and system based on multi-scale convolution | |
| CN116416270B (en) | Method and device for monitoring underground working surface leakage water | |
| CN117152849A (en) | Novel method for identifying identities of underground weak characteristic personnel of coal mine | |
| CN114120056B (en) | Small target identification method, device, electronic equipment, medium and product | |
| CN112487967A (en) | Scenic spot painting behavior identification method based on three-dimensional convolution network | |
| Soundarya et al. | Deep Learning based Automatic Vehicle Number Plate Recognition for Vehicle Authentication and Intelligent Security System | |
| Akbar et al. | License Plate Identification using Machine Learning Techniques |
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 |